Loading…

Sign up or log in to bookmark your favorites and sync them to your phone or calendar.

Wednesday, September 25
 

13:00

Early Registration Open
Wednesday September 25, 2019 13:00 - 18:00
Lobby
  • YouTube Link -
 
Thursday, September 26
 

09:00

Expo Hall Open
Thursday September 26, 2019 09:00 - 18:30
Expo Hall
  • YouTube Link -

09:00

Registration Open
Thursday September 26, 2019 09:00 - 18:30
Lobby
  • YouTube Link -

09:00

Coat Check
Thursday September 26, 2019 09:00 - 19:00
Lower Level

10:00

Welcome from 2019 OCP Foundation Board Chair
Speakers
avatar for Mark Roenigk

Mark Roenigk

OCP Chairman of the Board, Facebook


Thursday September 26, 2019 10:00 - 10:10
Auditorium Theatre

10:10

OCP 2019 Highlights
Speakers
avatar for Rocky Bullock

Rocky Bullock

CEO, OCP
Rocky has more than 40 years of experience in the technology industry with expertise in professional services, systems integration, product design, new product introductions, technology development, sales, operations, manufacturing and logistics. He has held executive and senior manager... Read More →


Thursday September 26, 2019 10:10 - 10:15
Auditorium Theatre

10:15

2019 Project Highlights and Accomplishments
Speakers
avatar for Bill Carter

Bill Carter

CTO, OCP
Bill is CTO for the Open Compute Project Foundation where he is a visionary and pathfinder on this open source journey. Bill works will the entire OCP community and many other software groups, standards organizations, and industry consortiums to bring together people from around the... Read More →


Thursday September 26, 2019 10:15 - 10:30
Auditorium Theatre

10:30

Facebook: Challenges and Opportunities of Architecting AI Systems at Facebook Datacenter at Scale
The rapid growth of AI, fueled by improvements in algorithms and availability of more data, has stressed all major components of the traditional datacenter infrastructure, such as compute, memory, storage and fabric. To contain the unsustainable growth of the infrastructure requires designing and adapting innovative AI platforms. In this talk we will describe major challenges which arise when building such AI platforms, both for inference and training. Those challenges include the need to accommodate constantly evolving workloads, while maintaining high-performance, energy efficiency and programability. We will also talk about opportunities for OCP community to address these challenges at various levels of open-spec AI hardware and open-source software, and provide some concrete examples of Facebook solutions which resulted in recent OCP contribution.

Speakers
avatar for Misha Smelyanskiy

Misha Smelyanskiy

Director, AI System SW/HW, Facebook


Thursday September 26, 2019 10:30 - 10:55
Auditorium Theatre

10:55

11:00

11:05

Operationalizing a Global, Circular IT Industry is Both Our Opportunity and Imperative
The global IT industry is responsible for 4% of global emissions and on track to double by 2025.  The OCP community is doing a great job driving advances in operational energy efficiency.  But this is only part of the puzzle, one third of the puzzle in fact.  To truly assess and address the environmental impact of data center equipment, it is imperative to include both pre-use (embodied) and post-EOL costs of the massive scaling of our collective infrastructure.  In this session we will explore the true total cost of equipment, and the opportunity represented by circular data centers to redefine lifetimes, and maximize both financial value and sustainability.  Operationalizing circular data centers means enabling a global, circular IT hardware industry, and to do so catalyzes both financial and environmental opportunity, and democratized access to growth.   We’ll talk about what this entails, what’s happening now, and what is at stake.  

Speakers
avatar for Ali Fenn

Ali Fenn

Chief Innovation Officer, ITRenew
Ali Fenn is the Chief Innovation Officer at ITRenew, where she leads circular data center initiatives, including the development of modeling and analytics platforms, market development, business model innovation, and new services.  Ali has been building enterprise and cloud technology... Read More →


Thursday September 26, 2019 11:05 - 11:30
Auditorium Theatre

11:30

Open Systems Firmware 2021 Initiatve
Speakers
avatar for Bill Carter

Bill Carter

CTO, OCP
Bill is CTO for the Open Compute Project Foundation where he is a visionary and pathfinder on this open source journey. Bill works will the entire OCP community and many other software groups, standards organizations, and industry consortiums to bring together people from around the... Read More →


Thursday September 26, 2019 11:30 - 11:35
Auditorium Theatre

11:35

Marketplace Momentum & Next Phase
Speakers
avatar for Steve Helvie

Steve Helvie

VP of Channels, OCP


Thursday September 26, 2019 11:35 - 11:40
Auditorium Theatre

11:40

EDGE Project Update and showcase of Edge New Products
Speakers
avatar for Bill Carter

Bill Carter

CTO, OCP
Bill is CTO for the Open Compute Project Foundation where he is a visionary and pathfinder on this open source journey. Bill works will the entire OCP community and many other software groups, standards organizations, and industry consortiums to bring together people from around the... Read More →


Thursday September 26, 2019 11:40 - 11:45
Auditorium Theatre

11:45

11:50

12:00

Lunch
Thursday September 26, 2019 12:00 - 13:30
Expo Hall
  • YouTube Link -

13:00

Open Domain-Specific Architecture (ODSA) Project Overview
The Open Domain-Specific Architecture aims to develop an open architectural interface to support chiplet-based design for domain-specific architectures. Domain-specific architectures are programmable high-performance products designed to execute high-intensity workloads, such as machine learning, networking and storage, efficiently. Chiplet-based designs implement integrated products by partitioning a design across multiple die instead of a single monolithic design. The ODSA aims to define an open interface such that chiplets from multiple vendors that support the interface can be assembled into domain-specific products. The ODSA started as an informal industry working group in November, 2018 and became an OCP sub-project in March, 2019.This talk will introduce the audience to the ODSA and its three workstreams: (a) Interface Definition; (b) Business and workflow; (c) Reference implementations. The talk will review recent results developed by the group and their relevance and significance to the technical goals of the group. The talk will highlight a new analysis of PHY interconnect technologies and design process learned from the PoC. Talks in the session by other speakers will cover technical results in each workstream in greater detail. The talk will also cover areas where the ODSA group is looking for assistance from new participants.

Speakers
avatar for Bapi Vinnakota

Bapi Vinnakota

Engineer, Netronome
Bapi is the Director of Silicon Architecture Program Management and OCP ODSA Workgroup Lead at Netronome. Bapi taught at the University of Minnesota, Twin Cities, immediately after a Ph.D. at Princeton. He joined Intel through an acquisition and was the architect of a flow processor... Read More →


Thursday September 26, 2019 13:00 - 13:10
Auditorium Theatre

13:10

Open Accelerator Infrastructure (OAI) Project Overview
Artificial Intelligence (AI) applications are rapidly evolving and producing an explosion of new types of hardware accelerators for Machine Learning (ML), Deep Learning (DL), and High-Performance Computing (HPC). Different implementations target similar requirements for power/cooling, robustness, serviceability, configuration, programming, management and debug, as well as inter-module communication to scale up and input/output bandwidth to scale out. We need an open infrastructure to intercept rapid innovation in artificial intelligence. OAI is where open accelerator infrastructure meets open artificial intelligence. In this group we define and build a modularly interoperable infrastructure around OAM.

Speakers
avatar for Whitney Zhao

Whitney Zhao

Hardware Engineer, Facebook Inc.
Whitney is a seasoned hardware engineer leading AI/ML system design in Facebook. Whitney has led multiple hardware generations ranging from general purpose 2S system such as Tioga Pass to ML JBOG Big Basin systems, all of which have been contributed to OCP. She has been driving multiple... Read More →


Thursday September 26, 2019 13:10 - 13:20
Auditorium Theatre

13:20

Open System Firmware (OSF) Project Overview
Hear from the Project Lead on how this open source firmare project enables the industry to use a Linux or EFI core to provide platform initialization.

Speakers
avatar for Ron Minnich

Ron Minnich

Software Engineer, Google
Ron Minnich is the inventor of LinuxBIOS, now known as coreboot. He is a member of the Technical Steering Committee for LinuxBoot, as well as co-leader of the Open Systems Firmware project at the Open Compute Project. He has worked in firmware for 40 years and in the open source... Read More →


Thursday September 26, 2019 13:20 - 13:30
Auditorium Theatre

13:30

openEDGE Project Overview
A review of the current state of the openEDGE project, including new contributions, new initiatives, and exploring various deployment opportunities for openEDGE.

Speakers
avatar for Mike Moore

Mike Moore

Regional Product Manager, Nokia
Over 20 years of R&D experience in both the Wireless and Defense Industries prior to transitioning into a customer facing sales and support role.  Now the Western Hemisphere Regional Product Manager for Nokia’s AirFrame Data Center Portfolio, with a focus on creating highly available... Read More →


Thursday September 26, 2019 13:30 - 13:40
Auditorium Theatre

13:30

Building and Operating an OCP Data Center at Small Scale
It is well recognised that the OCP style of data center is both energy and resource efficient with many operational advantages over traditional data center arrangements. This workshop highlights many aspects of the build and operation of a data center that is inspired by OCP, which uses OCP IT hardware. It will touch upon the operational energy savings of using direct fresh air in a climate that has a very large range of temperatures from -30 degC to 25 degC and no centralised UPS all at a scale that would be suitable for containers. The IT resources are operated and monitored using a complete suit of opensource tools and environments all of which enables this workshop to run an interactive demonstration of the capabilities of Opensource on OCP IT hardware at the small scale. The workshop will also introduce the research and development that RISE has undertaken with partners to create holistic thermal management control.

Speakers
avatar for Jon Summers

Jon Summers

Scientific Leader in Data Centers, Research Institutes of Sweden
My main area of research these days is on thermal and energy management of the digital infrastructure. I have worked in a research environment focused on heat transfer from the chip to the chiller in the backdrop of using real IT equipment, which lead me to join RISE in the north... Read More →


Thursday September 26, 2019 13:30 - 13:55
Emerald Room
  • YouTube Link -http://www.youtube.com/watch?v=_CxVWMKB04Q

13:40

Open RMC Project Overview
In Q4 2018, OCP established the OpenRMC subproject to architect and implement an open-source rack management controller. The OpenRMC will allow the rack to be managed via a Redfish interface. As a fast cycle deliverable, OpenRMC has selected the RMC on an OpenEdge platform as a target platform. This presentation will present the project’s progress and review its collateral.

Speakers
avatar for John Leung

John Leung

Principal Engineer, Intel Corporation
John is a Principal Engineer at Intel Corporation in the Data Center Group.  John has been in the computer industry for over 37 years: architecting, designing and coding.  At Intel, John is currently focused on the manageability interfaces for Intel® Rack Scale Design and server... Read More →


Thursday September 26, 2019 13:40 - 13:50
Auditorium Theatre

13:50

OCP NIC 3.0 Technology and Ecosystem Update
OCP NIC 3.0 specification has been a follow-on to the OCP Mezz 2.0 specification. This has been a collective effort of over 14 industry partners who all showcased their contributions during OCP Global Summit in March 2019 in the OCP Experience Area. The community has strongly rallied around this specification and we see strong adoption in the industry across OEMs, ODMs, system suppliers and NIC suppliers. In this session we will provide a general update on the OCP NIC 3.0 that will cover technology and ecosystem status and current on-going compliance and inter-operability efforts. We are motivated to engage the EU community to contribute to this effort via future enhancements and options that makes sense from EU industry perspective.

Speakers
avatar for Damien  Chong

Damien Chong

Hardware Engineer, Facebook


Thursday September 26, 2019 13:50 - 14:00
Auditorium Theatre

14:00

Executive Track
Thursday September 26, 2019 14:00 - 14:15
G102/G103
  • YouTube Link -

14:00

Networking Executive Track - presented by Facebook
As network traffic demands from applications (Live, Video, Messenger, machine learning, and more) continues to grow, new challenges are emerging in power, density, and scale-up / scale-out design paradigms.
 
This talk will cover approaches to networking as they relate to support for 200G and 400G optical modules, and movement into new Facebook network topologies. It will also address Facebook’s experiences with Minipack in production (Minipack is Facebook’s latest fabric switch, introduced at the March 2019 OCP Global Summit). This talk will also discuss forward looking ideas related to network monitoring at scale, which are critical for switch element design, topology optimization, and workload distribution.

Speakers
avatar for Hans-Juergen Schmidtke

Hans-Juergen Schmidtke

Director, Network Hardware Engineering, Facebook
Hans-Juergen Schmidtke is Director of Engineering in Facebook's InfrastructureFoundation team based in Menlo Park, CA. In this role he is part of a team that isEngineering Facebook's Network Infrastructure. In previous roles at Juniper Networks,Hans-Juergen Schmidtke was VP of Product Management in the CTO office, VP ofAdvanced Technologies, and Chief Architect for Converged Core. In these roles he wasresponsible for the architecture of SDN, data center, acce... Read More →


Thursday September 26, 2019 14:00 - 14:25
Auditorium Theatre

14:00

14:20

Bringing Intelligence to Storage with Samsung SmartSSD - presented by Samsung
The rise of data demands innovations in storage architectures to overcome traditional bottlenecks, and further improve cost and performance efficiency to provide platforms for future data storage and processing. While NVMe and NVMe-oF are unleashing the true performance of SSDs and are now becoming mainstream, the future calls for more revolutionary changes in the overall storage architectures. In this session, Samsung discusses its new SmartSSD technology, which integrates data processing into an all-in-one SSD solution with disruptive scalability and latency for large-scale AI/ML workloads.

Speakers

Thursday September 26, 2019 14:20 - 14:35
G102/G103

14:30

Executive Track
Thursday September 26, 2019 14:30 - 14:55
Auditorium Theatre
  • YouTube Link -

14:30

Openstack & OCP
EnterCloudSuite European is a public cloud platform based on Openstack snd partially OCP equipment to run memory- and CPU-optimised instances. Specifically, the platform is built upon refurbished Windmill servers in a non-redundant architecture: single SSD, single NIC, single power feed and single switch. As a provider, OCP infrastructure gives us the ability to provide low-cost resources for those cloud-native application which don’t need any kind of server-side redundancy, and to offer more capacity to the end customer. OCP also standardize and simplify the components, making spare parts management and server replacement easier. Additionally, customers using these type of instances can take advantage of lower prices compared to standard instances and increased scalability, in a spot instance use case (like SKY Italy does). This also perfectly fits into a Kubernetes use case, where you spawn multiple replicas of your services across nodes and then if one of those became unavailable kubernetes will handle it and respawn the container on another available node.

Speakers
MC

Mariano Cunietti

Head of Cloud, Irideos Spa


Thursday September 26, 2019 14:30 - 14:55
Emerald Room

14:40

Empower New Data Age with new Open Compute Evolution - presented by Inspur
Inspur is a leading global data center and cloud computing solutions provider. From 2016, Inspur has been actively participating into the OCP community with its expertise in product innovation and datacenter solution enablement. Nowadays, with the best practices from the engagement with hyperscaling DC customers, Inspur has open sourced several new products across compute node, storage to AI workload. 
Alan will introduce these new portfolio and articulate the benefits from these OCP based products to the broad cloud customers.

Speakers
avatar for Alan Chang

Alan Chang

Deputy GM of Server Product Line, Inspur


Thursday September 26, 2019 14:40 - 14:55
G102/G103

15:00

Open up your potential with HPE ProLiant - presented by HPE
Speakers
avatar for Tom Lattin

Tom Lattin

Vice President ProLiant and Cloudline Systems, Hewlett Packard Enterprise


Thursday September 26, 2019 15:00 - 15:15
G102/G103

15:00

ZT Systems Open Accelerator Infrastructure (OAI) Server Solution - presented by ZT Systems
The server industry has been built on the foundation of standards. Harnessing the performance benefit of new innovations through standards leads to faster, more efficient product development with improved manufacturability, serviceability, reliability, and cost.  With the rapid evolution of AI, there has been an explosion of hardware accelerations for Machine Learning (ML), Deep Learning (DL), and High-Performance computing (HPC) including GPUs, FPGAs, ASICs, NPUs, etc. Building server solutions for each hardware accelerator is not feasible due to development costs and time-to-market considerations. Open Accelerator Infrastructure is an open standards approach to realize the performance benefits of the various hardware accelerators with a single system implementation. The system details of the ZT Systems Open Accelerator Infrastructure (OAI) Server Solution will be reviewed during this session.

Speakers
RM

Raymond Miles

VP of Field Engineering, ZT Systems


Thursday September 26, 2019 15:00 - 15:25
Auditorium Theatre

15:00

Approach of OCP Adoption and How to Operate our OCP Infrastructure in Yahoo! JAPAN
Since 2016, we have adopted more than 4,000 OCP servers in our Japan DC. How do we approach OCP deal for adoption and how do we operate OCP server and other components in our infrastructure. The metrics are explained that are useful for comparing OCP server and EIA19. Also our the data center has changed by adopting OCP server than ever. Also talk about Why is it changed? What is it changed?.

Speakers
avatar for Kazuhide Fujimi

Kazuhide Fujimi

Server Infrastructure Architect, Yahoo Japan Corporation
Kazuhide Fujimi is a Server Infrastructure Architect at Yahoo! Japan and has responsibility for procurement, testing, QA and troubleshooting of server hardware. Yahoo! Japan is one of the famous CSP in Japan and has several data centers in Japan as on-premise Infrastructure envir... Read More →


Thursday September 26, 2019 15:00 - 15:25
Emerald Room

15:20

Architecting for the Zettabyte Era with Zoned Storage - presented by Western Digital
Zoned Storage is designed to help data center architects and application developers prepare for the future zettabytes of data predicted to be created from video, IoT/edge data, or AI/ML workloads. Zoned Storage, Western Digital’s open source initiative, represents an opportunity to reap the benefits of combining SMR/ZNS technologies under a unified architecture at zettabyte scale. Eyal Bek will talk about how developers can utilize this ecosystem to take advantage of ZNS and SMR technologies, ensuring access to highest capacities and seamless integration, and do the following:
· Leverage open source and standards
· Enable performance efficiencies by serializing workloads into zones
· Simplify firmware and provide the host with system-level intelligence for data placement
· Ensure access to highest capacities and better performance
Get started now for a competitive advantage using Zoned Storage across both SMR HDDs and ZNS SSDs. Use purpose-built storage for competitive TCO now and into the future.

Speakers
avatar for Hans Holmberg

Hans Holmberg

Technologist, Western Digital
Hans is a Technologist at Western Digital, working on emerging storage solutions and driving the zoned block device ecosystem forward. He has a long history of working with flash and has architected and implemented solutions for everything from raw NAND to Open-Channel and NVMe. Hans... Read More →


Thursday September 26, 2019 15:20 - 15:35
G102/G103

15:30

Executive Track, 25 mins.
Thursday September 26, 2019 15:30 - 15:55
Auditorium Theatre
  • YouTube Link -

15:30

Immersion Cooling for Research and Colocation - A case study in cooperation and partnership between OCP members
Submer Immersion Cooling released the first commercially available OCP compatible Immersion Cooling system in March 2019 at the Open Compute Global summit in San José, California, supporting 21” hardware and a flexible bus-bar systems supporting multi-zone 12v & 48v configurations. Since then, Submer has been actively partnering with other OCP members such as 2CRSI, LiteOn, Gigabyte and more and has been presenting and deploying full OCP solutions for European public research centres, colocation data centres (also leveraging refreshed OCP vanity free hardware) bringing unprecedented efficiencies and economic advantages. In this session, Submer will cover two case studies (colocation and research) where OCP was not originally being considered but managed to make a case for ease of deployment, operation and much more. Hear about the experiences partnering with OCP partners and introducing 3rd parties to OCP.

Speakers
DP

Daniel Pope

CEO, Submer Immersion Cooling


Thursday September 26, 2019 15:30 - 15:55
Emerald Room
  • YouTube Link -http://www.youtube.com/watch?v=SBNswlL62dI

15:40

From Bare Metal toward OCP Solutions - presented by MiTAC + CircleB
OCP solution adoption has been steadily growing through its great CAPEX and OPEX saving benefits in newly built datacenters meanwhile MiTAC’s ESA kit has continuously extended such benefits to a legacy datacenter and EIA-310 racks, including several world-class CSPs and Telco companies.
While collaborating with different customers and various open foundations, we, MiTAC and Circle B would like to present, by how we integrate different OCP products elements – Compute nodes, storages, switches, ESA kits toward different solution level of products and applications, such as the next generation Central Office with the SDN service also provides an optimized solution for the telecom carriers.
We believe with OCP several solution case sharing, it could help more and more potential OCP solution adopters quickly and efficiently implement OCP solutions.

Speakers
avatar for Menno Kortekaas

Menno Kortekaas

Technical Director, Circle B
Providing OCP based solutions to European datacenters
HC

Hancock Chang

System Lead, MiTAC


Thursday September 26, 2019 15:40 - 15:55
G102/G103

16:00

SONiC Together - presented by Mellanox
SONiC is a game changer for data center networks as it is the one truly open source network operating system that Tier 1 & 2 clouds and large enterprises can now standardize on.  The main holdup for some has been the lack of official support – but that changes now.  Mellanox Technologies is offering full ASIC to Protocol (A2P) support for the community edition SONiC.

Speakers
DI

David Iles

Sr. Director Ethernet Switching, Mellanox


Thursday September 26, 2019 16:00 - 16:15
G102/G103

16:00

Exploring Lightning JBoF Configurations for Different Use Cases
The OCP Lightning JBoF is a versatile platform that is capable of supporting multiple chassis level configurations. At Facebook, we deploy Lightning in a simple form, connecting the storage server (aka head node) with a x16 PCIe connection to the Lightning tray. While this configuration fulfills our production needs, we also realize other potential OCP adoptions could have different requirements in terms of IO workloads, compute resource and data availability. In this talk, we will discuss different configurations of Lightning to meet these needs, leveraging the PCIe switch features such as Dynamic Partitioning. Facebook has explored these options when we investigated to migrate more applications to Lightning. We feel it is important to share these findings with the entire OCP community. We hope that this talk will enlighten the audience and help them build solutions around OCP Lightning that could cater for their use cases.

Speakers
WZ

Wei Zhang

Software Engineer, Facebook


Thursday September 26, 2019 16:00 - 16:25
Emerald Room

16:00

Executive Panel: Collaborative Silicon Development
This panel will share their viewpoints on the significant changes expected in the business and operational workflow to move to a chiplet-based design flow. Panelists are all in an operational/test role on current flows, or is a chiplet/tool/IP supplier to support these flows.

Moderators
avatar for Bapi Vinnakota

Bapi Vinnakota

Engineer, Netronome
Bapi is the Director of Silicon Architecture Program Management and OCP ODSA Workgroup Lead at Netronome. Bapi taught at the University of Minnesota, Twin Cities, immediately after a Ph.D. at Princeton. He joined Intel through an acquisition and was the architect of a flow processor... Read More →

Thursday September 26, 2019 16:00 - 17:00
Auditorium Theatre

16:20

From Cloud to Edge: Bring the CSP proven solution to Edge data center - presented by Wiwynn
Wiwynn has been actively involved in the Open Compute Project (OCP) and works with world-class data centers to define, design and deliver the most suitable IT equipment for workload optimization and best TCO (total cost of ownership). We understand the importance of operational efficiency in the data center and the benefits that good hardware design can bring to the data center.
Now, we are extending our product portfolio from Hyper Scale to Edge applications.
We hope to bring the existing advantage and benefit from our experience to edge data center.
This talk, Wiwynn will presents the most leading DC and edge solution, and how to move from AI to the Edge.

Speakers
SL

Steven Lu

Vice President of Product Management and Marketing, Wiwynn


Thursday September 26, 2019 16:20 - 16:35
G102/G103

16:30

How OCP servers help unleash stranded flash capacity: Deep dive on disaggregation architecture trends and OCP solutions
This session will focus on three important topics: 1. How OCP servers enable efficient implementation of large scale-out NoSQL databases. 2. How disaggregation helps such large database implementations to be efficient and highly performant. 3. How NVMe SSDs are evolving to provide significant performance throughput at a lower latency. In traditional deployments with NVMe SSDs directly attached to the processor, it is difficult to achieve balanced system configuration resulting in stranded flash performance and capacity. In disaggregated storage, flash is shared among pool of compute nodes using dynamic binding of storage resources based on performance, capacity requirements. However, carving out storage from the flash pool and exposing over network introduces latency. In this session we outline different network protocols, latency implications and industry trends to deliver low latency disaggregated flash storage over the network using OCP platforms and RocksDB.

Speakers
AR

Anjaneya Reddy Chagam

Senior Principal Engineer and Chief Software-Defined Storage Architect, Intel
Anjaneya “Reddy” Chagam is a Senior Principal Engineer and Chief Software-Defined Storage Architect in Intel’s Data Center Group. He is responsible for developing software-defined storage strategy, architecture, and technology initiatives. He is a member of the Technical Steering... Read More →
JS

Jay Shenoy

VP Technology, Hyve
MW

Manoj Wadekar

Storage Architect, Facebook


Thursday September 26, 2019 16:30 - 16:55
Emerald Room

16:40

Bringing Intelligence to Storage with Samsung SmartSSD - presented by Samsung
The rise of data demands innovations in storage architectures to overcome traditional bottlenecks, and further improve cost and performance efficiency to provide platforms for future data storage and processing. While NVMe and NVMe-oF are unleashing the true performance of SSDs and are now becoming mainstream, the future calls for more revolutionary changes in the overall storage architectures. In this session, Samsung discusses its new SmartSSD technology, which integrates data processing into an all-in-one SSD solution with disruptive scalability and latency for large-scale AI/ML workloads.

Thursday September 26, 2019 16:40 - 16:55
G102/G103

17:00

Opening Night Happy Hour
Thursday September 26, 2019 17:00 - 18:30
Expo Hall
  • YouTube Link -
 
Friday, September 27
 

08:00

Registration Open
Friday September 27, 2019 08:00 - 12:00
Lobby
  • YouTube Link -

08:00

Expo Hall Open
Friday September 27, 2019 08:00 - 16:30
Expo Hall
  • YouTube Link -

08:00

Coat Check
Friday September 27, 2019 08:00 - 16:30
Lower Level
  • YouTube Link -

09:00

Open Tape for Open Compute
Recently developed open-source software components that facilitate tape integration into open compute and cloud storage environments are described.  Workload trends and economic aspects motivating the use of tape are also discussed.

Speakers
RH

Robert Haas

IBM Zurich
avatar for Slavisa Sarafijanovic

Slavisa Sarafijanovic

IBM Research Zurich
AI for storage: similarity search for finding semantically similar data; data access prediction for proactive caching. Tape integration with file and object storage.
MA

Mark A. Lantz

IBM Zurich


Friday September 27, 2019 09:00 - 09:15
Auditorium Theatre

09:00

Motivation for ODSA/Fit into OCP
The ODSA is a new and exciting project under the OCP Server subgroup. With the decline of Dennard’s scaling and evident slow down to crawl of Moore’s Law, a new approach is needed within industry to continue developing cost effective solutions for AI/ML/Networking use cases which take advantage of existing developments in the standard areas such as I/O interfaces while allowing for innovations in specific areas such as accelerators. ODSA enable this approach for a continued innovation. OCP is the de-facto open source hardware initiative for data center and infrastructure, and now with ODSA being a part of OCP, we can expect exciting fast solutions at system level center around ODSA-based modules. In this talk, we will look at some analogous use cases such as OAM and how similar approach can be leveraged by the industry to promote and enable ODSA engagement and adoption among developers and users alike.

Speakers
DJ

Dharmesh Jani

OCP Co-Chair and FB OCP Lead, Facebook


Friday September 27, 2019 09:00 - 09:20
G103

09:00

An Architecture for a Highly Available Disaggregation NOS built upon Open Networking Linux
Service providers have more stringent requirements for their network devices than operators running data centers. As well as needing support for more complex control plane protocols, such as IP/MPLS, the devices need to be able to continue routing traffic when either (or both) data-plane and control-plane. In this talk, Metaswitch will introduce an open architecture for a portable, disaggregated network operating system that can meet the High Availability requirements of service providers and other network operators. This HA architecture is based on Open Networking Linux and leverages other open networking technologies

Speakers
avatar for Pat Moore

Pat Moore

Metaswitch


Friday September 27, 2019 09:00 - 09:25
G102

09:00

Advancements to Improving Cloud System Up-Time with Runtime Firmware Upgrade
Cloud systems demands high availability. Today most of the system firmware upgrade requires reboot. This session will provide advancements in the firmware area to upgrade the system components without reboot. This session will also highlight the OS requirements for both control plan and data plane to make the firmware upgrade without reboot possible. The session will also highlight the changes needed in standard bodies to facilitate the firmware upgrade without reboot.

Speakers
avatar for Mallik Bulusu

Mallik Bulusu

Principal Firmware Engineering Manager, Microsoft Corporation
Mallik Bulusu is a Principal Firmware Engineering Manager (Azure) at Microsoft Corporation. He has been specializing in design and development of system firmware using UEFI paradigm since 1999. Mallik holds 60+ patents on novel advances in system firmware design for Server and Mobile... Read More →
avatar for Murugasamy Nachimuthu

Murugasamy Nachimuthu

Sr. Principal Engineer, Intel Corporation
Murugasamy (Sammy) Nachimuthu is a lead cloud solutions architect, focusing on improving cloud service provider infrastructure. Sammy joined Intel in 1995, he played key role in architecture of Rack Scale Design (RSD) architecture, Intel® Optane™ DC Persistent Memory, PCIe, Memory... Read More →


Friday September 27, 2019 09:00 - 09:25
G104

09:00

Driving transformation in Telcos infrastructures through OCP adoption
Fixed/mobile telco operators, and cloud providers strive to face the challenging demands coming from the evolution of IP networks (e.g. huge bandwidth requirements, integration of billions of devices and millions of services in the cloud). At ONF, we are committed to help service providers to overcome such challenges leveraging open source software and disaggregated hardware. For over three years, ONF has been building open-source solutions on top of OCP accepted hardware and software. ONF wants to introduce these solutions to the OCP audience. In the first part of the talk, we want to present Trellis, SEBA and Stratum and elaborate how OCP plays an important role in our ecosystem. These are just few examples of how ONF leverages OCP components - Mobile access network can be also supported on the same hardware infrastructure. Working in field deployments (and at scale) means also facing unique problems that you never see during the development process. Software and hardware issues can occur, sometimes all together and such issues are also unpredictable. The second part of the talk shares with the OCP community pitfalls and lessons learnt coming from deployments experience of our open source projects. Trellis is a multi-purpose leaf-spine fabric which builds an extensive features set on top of OCP accepted hardware from multiple ODMs with ASICs coming from different vendors, as well as OCP accepted software like ONIE and ONL. SEBA (or SDN Enabled Broadband Access) is a self-contained pod solution for disaggregated Passive Optical Networks (PONs) where the reference implementation is built on an OCP approved specification for AT&T's 16 ports XGS-PON vOLT. Stratum is an open source silicon-independent switch operating system. It builds an open, minimal production-ready distribution for white box switches. Stratum uses both ONIE and ONL to provide boot and platform support on network devices and can leverage open interfaces like SAI and SDKLT to program switches' ASICs.

Speakers
avatar for Pier Luigi Ventre

Pier Luigi Ventre

Member Of Technical Staff, Open Networking Foundation
Pier Luigi Ventre is a Member of Technical Staff at Open Networking Foundation where he is working on Trellis - the leading open-source multi-purpose leaf-spine fabric. Before joining ONF, Pier Luigi Ventre worked at CNIT (National Inter-University Consortium for Telecommunications... Read More →


Friday September 27, 2019 09:00 - 09:25
G105

09:00

Rack for Telco Central Office
At the regional summit in 2018 KPN presented the challenges a Telco has when they want to deploy a pre-populated, pre-cabled rack into a legacy Telco location. Together with OCP platinum member Schneider Electric, KPN has been working on a solution for these challenges and will jointly present their ideas to the OCP community in this talk.

Speakers
avatar for Michel Geensen

Michel Geensen

Principal Architect, KPN
Michel Geensen has 20+ years of experience in Telco IP Networks. At the moment he is working as an Principal Architect in KPN's Technology Office. KPN is the incumbent Telecom Operator from the Netherlands (a very large Telco in a very small country).As a part of his current job... Read More →


Friday September 27, 2019 09:00 - 09:25
Emerald Room

09:00

Datacenter-ready Secure Control Module (DC-SCM)
A Catalyst for the Modular Building Block Architecture (MBA): An update on the progress made since OCP Global Summit on Datacenter-ready Secure Control Module and its Interface (DC-SCM & DC-SCI)

Speakers
avatar for Siamak Tavallaei

Siamak Tavallaei

Principal Architect, Microsoft
Siamak Tavallaei is a Principal Architect at Microsoft’s Azure division. Collaborating with industry partners, he drives a number of initiatives in research, design, and deployment of hardware for Microsoft’s cloud-scale services such as Azure, Machine Learning, Bing, Office 365... Read More →


Friday September 27, 2019 09:00 - 09:25
G107

09:15

Demonstration of Silicon Photonics (de)multiplexor Using the CORNERSTONE Platform
The Photonic Integrated Circuit (PIC) industry is predicted to serve a data centre market worth US $19 Billion within 5 years, with silicon photonics forecast to exhibit a Compound Annual Growth Rate (CAGR) of 44% [1]. Silicon photonics can leverage the existing CMOS infrastructure for high volume production of systems such as transceivers and sensors, but within the silicon photonics ecosystem, there is currently a gap in the rapid prototyping services available for SME’s and fabless companies, alike. CORNERSTONE is a silicon photonics foundry service based in the UK which addresses this gap in the ecosystem by offering a versale rapid prototyping foundry service based on industry compatible processes, centered around a DUV scanner, to provide prototyping chips via a multi-project-wafer (MPW) mechanism whilst enabling a seamless transition to high volume foundries. A commonly discussed issue with silicon photonics devices stems from the strong thermo-opc effect of the silicon itself giving rise to high temperature sensitivity of devices, particularly resonant structures. Here, we present a silicon photonics bidirectional angled multimode interferometer (BAMMI) to largely eliminate the temperature dependent performance of a (de)mulplexor.

Speakers
CL

Callum Littlejohns

University of Southampton
GR

Graham Reed

University of Southampton


Friday September 27, 2019 09:15 - 09:30
Auditorium Theatre

09:25

The Open Domain-Specific Architecture: A Technical Introduction
The Open Domain-Specific Architecture aims to develop an open architectural interface to support chiplet-based design for domain-specific architectures. Domain-specific architectures are programmable high-performance products designed to execute high-intensity workloads, such as machine learning, networking and storage, efficiently. Chiplet-based designs implement integrated products by partitioning a design across multiple die instead of a single monolithic design. The ODSA aims to define an open interface such that chiplets from multiple vendors that support the interface can be assembled into domain-specific products. The ODSA started as an informal industry working group in November, 2018 and became an OCP sub-project in March, 2019.This talk will introduce the audience to the ODSA and its three workstreams: (a) Interface Definition; (b) Business and workflow; (c) Reference implementations. The talk will review recent results developed by the group and their relevance and significance to the technical goals of the group. The talk will highlight a new analysis of PHY interconnect technologies and design process learned from the PoC. Talks in the session by other speakers will cover technical results in each workstream in greater detail. The talk will also cover areas where the ODSA group is looking for assistance from new participants.

Speakers
avatar for Bapi Vinnakota

Bapi Vinnakota

Engineer, Netronome
Bapi is the Director of Silicon Architecture Program Management and OCP ODSA Workgroup Lead at Netronome. Bapi taught at the University of Minnesota, Twin Cities, immediately after a Ph.D. at Princeton. He joined Intel through an acquisition and was the architect of a flow processor... Read More →


Friday September 27, 2019 09:25 - 09:45
G103

09:30

Experimental Evaluation of Cooling Performance for Liquid Immersion with Bubble Assisted Natural Convection
A bubble-assisted natural convection method achieves both the high cooling limit of the forced convection method in which the refrigerant is convected by the pump and the feature of the natural convection method achieving a low PUE of around 1.02. In our previous report, the characteristics were examined mainly based on simulation, but this time the potential of this cooling method was evaluated mainly from an experimental approach. Specifically, a fluorocarbon as a refrigerant and a silicone oil which has a low global warming potential and is expected to exhibit the similar cooling performance. In the natural convection method, the lower the viscosity of the refrigerant, the higher the cooling efficiency. However, the bubble assistance effect was more remarkable as the viscosity of the refrigerant becomes higher. As a result, even with silicone oil, a high cooling effect was demonstrated. In addition, with any of the refrigerants, high cooling properes of PUE 1.02-1.04 were demonstrated with bubble assistance. That is, when selecting a refrigerant, it is not necessary to consider only physical properes relating to convection such as viscosity as an evaluation value, and other physical properes such as cost and global warming coefficient can be considered as evaluation values. These results indicate that the bubble assisted convection technology exhibits promising potential for a practically useful cooling technology.

Speakers
HK

Hideo Kubo

Fujitsu
MM

Morito Matsuoka

Osaka University
KM

Kazuhiro Matsuda

Osaka University


Friday September 27, 2019 09:30 - 09:45
Auditorium Theatre

09:30

SONiC – Reliability, Manageability and Extensibility
SONiC has evolved a lot since last Regional Summit. It is a building block for an extensible modern network. We will present a few key area of evolving: warm reboot for reliability, error handling, L2/L3 capability, manage-ability.

Speakers
BG

Ben Gale

Technical Director, Broadcom Inc
Ben has worked in networking and NOS development for 25 years, with 3Com, LVL7 and Broadcom. He is the primary Broadcom technical point-of-contact into the SONiC community.
XL

Xin Liu

Principal Product Manager, Microsoft
Xin has been worked in the cloud computing for more than 5 years. She is the community lead for SONiC and SAI projects. Xin has been with Microsoft since she got her Ph.D in computer science from NEU. She has experience of speaking at INFOCOM, OCP, Networking workshops and all sorts... Read More →
GL

Guohan Lu

Principal Software Engineer Manager, Microsoft


Friday September 27, 2019 09:30 - 09:55
G102

09:30

Arm SBBR and ServerReady Program
This session presents the Arm server firmware standard SBBR (Server Base Boor Requirements) and its relevance to the Arm ServerReady Program including the testing and certification. It will also discuss the Arm server hardware, security and server management standards and their relevance to OCP and their dependency on OCP standards.

Speakers
DW

Dong Wei

Fellow, Arm


Friday September 27, 2019 09:30 - 09:55
G104

09:30

A Distributed Disaggregated Chassis Model for Building a Large Scale-Out Forwarding System
In recent years, the Leaf-Spine CLOS topology has been widely adopted in the Cloud Data Centers as the scale-out topology for interconnecting compute and storage. This interconnection paradigm takes advantage of whitebox switches built on merchant silicon and offers disaggregation of the procurement of the hardware forwarding element from the control software of these switches. The Distributed Disaggregate Chassis model for building a scale-out forwarding system is based on these principles: 1. White box hardware modules built on Merchant forwarding Silicon. 2. Disaggregation of procurement of hardware and software. 3. Achieve a unified system design that employs a limited number of modules to scale from small to a very large system operating as a single logical L2/L3 router. In this talk, AT&T is presenting a system specification for the DDC-FS and the white box hardware components that make up the DDC-FS to address the WAN routing use case.

Speakers
avatar for Janet Peng

Janet Peng

Director Member Technical Staff, AT&T
Janet Peng is Director Member of  Technical Staff at AT&T Labs, Domain 2.0 Architecture and Design, responsible for delivering the architecture and design for AT&T network infrastructure.  She is leading open disaggregated white box initiatives at AT&T to transform traditional... Read More →


Friday September 27, 2019 09:30 - 09:55
G105

09:30

An Introduction to ACS Liquid Cooling Cold Plate Requirements
The OCP Advanced Cooling Solution (ACS) Cold Plate work stream are generating a Liquid Cooling Cold Plate Requirements document. This document focuses on liquid cooling definitions, cooling component selections, and requirements for liquid cooled solutions. The document is specifically focused on rack manifold distributed cooling within the Technology Cooling System (TCS) fluid loop. This is the fluid loop from the Coolant Distribution Unit (CDU) to the rack, through the manifold and the IT equipment, and then back through the manifold to the CDU. The liquid cooling solutions for the IT equipment spans all levels of liquid cooling, from hybrid solutions where both air and liquid cooling is used to full liquid installations where only liquid cooling is used. The definitions and selection of liquid cooling components include TCS cooling loop components such as: the cooling liquid, CDU, rack manifold, quick disconnect couplings, and cold plates. The document stresses the importance of ensuring material compatibility between the cooling liquid and all the materials that are exposed to the cooling liquid, which are referred to as wetted materials, to minimize any long term risks of corrosion. The liquid cooling requirements are determined to address quality, safety, management, and monitoring of liquid cooled installations. The target is to have the requirements document approved by the OCP Incubation Committee before the Regional Summit Meeting, and to present the approved document at the Summit.

Speakers
JG

Jessica Gullbrand

Tech Lead Engineer, Intel Corporation


Friday September 27, 2019 09:30 - 09:55
Emerald Room

09:30

HW/SW Co-design for future AI platforms - Large memory unified training platform (Zion)
Last decade has seen remarkable progress in AI, made possible by accumulation of large amounts of data, improvements in algorithmic technologies, and availability of large amounts of compute and storage capabilities, made possible by Moore’s law. Ironically, this progress has coincided with the slowing down of Moore’s law. This ‘coincidence’ created a unique challenge and an opportunity for HW/SW co-design for deep learning, which implies simultaneous design and optimization of several aspects of hardware and software to achieve some set target, such as performance, performance per watt, or area. Deep learning has been particularly amenable to such co-design processes across various parts of the software and hardware stack, which has led to a variety of novel algorithms, numerical optimizations and AI hardware. In this talk, we overview and characterize deep learning models used in many Facebook social network services, present their computational characteristics of our models, and show number of opportunities for better co-design. We will also present how the co-design methodology has been applied specifically to Facebook next generation training platform Zion.

Speakers
avatar for Whitney Zhao

Whitney Zhao

Hardware Engineer, Facebook Inc.
Whitney is a seasoned hardware engineer leading AI/ML system design in Facebook. Whitney has led multiple hardware generations ranging from general purpose 2S system such as Tioga Pass to ML JBOG Big Basin systems, all of which have been contributed to OCP. She has been driving multiple... Read More →
avatar for Dheevatsa Mudigere

Dheevatsa Mudigere

Research Scientist, Facebook


Friday September 27, 2019 09:30 - 09:55
G107

09:45

Collaboration Between CPU and Near-Data Accelerators (NDAs) for Machine Learning Training
Near-data accelerators (NDAs) that are integrated with main memory have the potential for significant power and performance benefits. Fully realizing these benefits requires the large available memory capacity to be shared between the host and the NDAs in a way that avoids copying data and that enables collaborative processing. We discuss how to solve four major challenges that enable such ne- grain sharing. Our solutions offer a path for standardizing access to CPU-connected NDAs that use current commodity memory interfaces and enable efficient and effective co-processing of data with concurrent access by both a CPU and NDAs within its main memory.  We demonstrate our approach in a simulated system that consists of a multi-core CPU and NDA-enabled DDR4 memory modules. We show that our mechanisms enable effective and efficient concurrent access using a set of microbenchmarks, and then demonstrate the potential of the system for the important variance-reduced gradient (SVRG) algorithm.

Speakers
BC

Benjamin Cho

University of Texas at Austin
ME

Mattan Erez

University of Texas at Austin


Friday September 27, 2019 09:45 - 10:00
Auditorium Theatre

09:50

Bunch of Wires Interface
Multi-Chiplet system-in-package designs have recently received a lot of attention as a mechanism to combat high SoC design costs and to economically manufacture large ASICs. Multi-Chiplet designs require low-power area-efficient inter-Chiplet communication. Current technologies either extend on-chip high-wire count buses using silicon interposers or off-package serial buses over organic substrates. The former approach leads to expensive packaging. The latter usually too complex to design. In this paper, we propose a simple Bunch of Wires (BoW) interface that operates in three modes, Base, Fast and Turbo. This interface combines the ease of development of parallel interfaces with the low cost of organic substrates, while providing throughput efficiencies up to 1Tbps/mm and a power efficiency below 0.7pJ/bit.

Speakers
RF

Ramin Farjad

SVP of R&D, Aquantia


Friday September 27, 2019 09:50 - 10:10
G103

10:00

Clonezilla for ONIE (Open Network Install Environment): A Switch and Full Rack Storage Operations Enabler
The Clonezilla for ONIE (CzfO) enables network switch storage operations that previously did not exist in the open-source network community. CzfO supports network switch storage operaons at the paron or disk/sector level, which enables rapid iteration for Software Defined Networking (SDN) and other programmable networks. CzfO supports network switch storage operations on the switch, in the rack, and over the network to support different development and production requirements. CzfO supports low-level switch storage backup, recovery, snapshot, and any other switch storage operatoons, which also can avoid the limitation of a target switch OS. With CzfO, network switch and full rack storage operations can support with a single integrated package.

Speakers
JM

Joe Mambretti

Northwestern University
JH

Jim Hao Chen

Northwestern University
High Performance Network & Cloud
SJ

Steven J.H. Shiau

Northwestern University


Friday September 27, 2019 10:00 - 10:15
Auditorium Theatre

10:00

Swedish Modules - 300kw Product Submission Presentation
Swedish Modules will supply technical data regarding the construction of the Swedish Modules 300kW scalable OCP-MDC. Swedish Modules will also share the benefits of using the OCP-MDC together with OCP gear to create a sustainable and efficient roll out of the infrastructure for the digital society. Furthermore, how the use of OCP- MDC solutions can easy up the deployment of OCP hardware outside the Hyperscale market by making it easy for small -and mid-size DC players to choose OCP and achieving the effects of using OCP.

Speakers
avatar for Roberto Söderhäll

Roberto Söderhäll

Chief Business Development Officer, Swedish Modules
Roberto is a veteran in working with innovation within the Data Center Industry. Spending more than two decades in the Data Center Industry, Roberto has been challenging the Status Quo of the Data Center Industry. As a speaker in many events his been proclaiming the Data Center as... Read More →


Friday September 27, 2019 10:00 - 10:25
G106

10:00

SONiC Ansible Test Automation
Scalable and repeatable testing is one of the founding building blocks of a successful product development cycle in today’s modern EOM business. SONiC’s astonishing growth rate in its feature set forces EOM HW vendors, such as Edgecore, to employ a highly-scalable test infra, supporting a rapid development environment while ensuring product reliability. SONiC, in a sense, challenges Edgecore’s traditional testing methodologies, requiring our teams to create an automated test management system in which testcases from multiple vendors must be run to validate interoperability of our HW along with products from multiple vendors, reducing regression issues, and enabling a fully automated CI/CD environment.

Speakers
avatar for Taskin Ucpina

Taskin Ucpina

Senior Director of SW Development, Edgecore
Taskin Ucpinar is a Senior Director at Edgecore , where he leads and coordinates Edgecore’s software development activities for next generation opensource projects, including SONiC and vOLTHA.Taskin has worked for small to large organizations, including Cisco, Symantec, Webroot... Read More →


Friday September 27, 2019 10:00 - 10:25
G102

10:00

Intel® Firmware Support Package (FSP) + EDK II for Cloud
This session will focus on creating BIOS using Intel® Firmware Support Package (FSP) and EFI Developer Kit II (EDK II) for Cloud platforms. This session also provide Intel® FSP for Cloud module overview and interface for integrating with EDK II. We also target to have a demo of the Intel® FSP + EDK II.

Speakers
avatar for Murugasamy Nachimuthu

Murugasamy Nachimuthu

Sr. Principal Engineer, Intel Corporation
Murugasamy (Sammy) Nachimuthu is a lead cloud solutions architect, focusing on improving cloud service provider infrastructure. Sammy joined Intel in 1995, he played key role in architecture of Rack Scale Design (RSD) architecture, Intel® Optane™ DC Persistent Memory, PCIe, Memory... Read More →


Friday September 27, 2019 10:00 - 10:25
G104

10:00

Converged Access Switch Specification Update
This is an update (final spec) to the specification of CAS for RAN fronthaul transport. A draft was introduced at the Global summit.

Speakers
avatar for Sumithra Bhojan

Sumithra Bhojan

Principal Member of Tech Staff, AT&T
Sumithra Bhojan is a Principal Member of Technical Staff at AT&T with a focus on Technology Design and Architecture for Broadband Wireline Access systems and RAN Fronthaul Transport. In her current role in AT&T labs, she has been focused on AT&T’s vision of transitioning access... Read More →


Friday September 27, 2019 10:00 - 10:25
G105

10:00

OCP-ACS Door Heat Exchanger Specification Update
Door Heat Exchanger (HEx) is an on-going work stream under the Advanced Cooling Solutions (ACS) sub-project. Following-up on discussions since the global summit, this presentation is meant to outline updates to the specification and spur feedback from the community. Within the specification, areas of physical interfaces, performance/metrology, monitoring/control, serviceability and testing will be covered in detail.

Speakers
JF

John Fernandes

Thermal Engineer, Facebook Inc.


Friday September 27, 2019 10:00 - 10:25
Emerald Room

10:00

Increasing network throughput and reducing CPU utilization by offloading KTLS onto crypto-capable SmartNIC – a case study
Encrypting all connections is a critical priority for most hyperscalers, but there is a penalty for software/CPU based TLS. Offloading TLS onto a crypto-enabled SmartNIC can return up to 90% of the CPU cycles lost when attempting to encrypt all connections within a data center. We will show our test results when offloading TLS onto a SmartNIC when running open source security applications on OCP Yosemite v2 servers and compare throughput and CPU utilization for software based TLS, plain text, and SmartNIC offloaded TLS. Enabling line-rate network performance at 50Gb/s with secure connections is practical and enables enhanced security for users.

Speakers
avatar for Nic Viljoen

Nic Viljoen

Associate Director of Engineering, Netronome
Nic works on upstream solutions using SmartNICs, focusing on the interaction between SmartNICs and the Linux kernel using the upstream BPF JIT for the Netronome NFP processor.


Friday September 27, 2019 10:00 - 10:25
G107

10:15

Automated Meet-Me-Room
Software automation of physical fiber interconnects in multi-tenant data centers enables enterprises and carriers to remotely provision and test interconnect, on-demand, instantly and without human error. An automated Meet-Me-Room enables new services to be provisioned in an inherently open source environment supporting interoperability. Bandwidth variability, changing workflows, disasters, equipment failures, upgrades and expansions are effectively managed by this soware-controlled physical interconnect layer.

Speakers

Friday September 27, 2019 10:15 - 10:30
Auditorium Theatre

10:15

Accelerating Innovation Through Chiplets and the Advanced Interface Bus (AIB)
Chiplets, in particular silicon die interconnected and copackaged, have become mainstream in Systems-on-a-Chip (SoCs) from Intel, AMD and others, enabled by advanced packaging technologies. Chiplets provide an advantageous means to introduce new I/O and computation capabilities to SoCs, by avoiding a monolithic silicon redesign of the SoC. Broader adoption of chiplets into lower cost or lower volume products can be enabled by reducing the time and expense of die-to-die interface development and expanding access to advanced packaging technologies. Standardization to achieve chiplet interoperability is possible within technical and business bounds. In this presentation we will examine how the high density of advanced packaging technologies has enabled the wide parallel bus as a high bandwidth, lower power die-to-die interface. Wide parallel busses of 1,000 or more wires between die that are co-packaged using an advanced packaging technology can provide lower development cost, lower unit cost, lower latency and lower power than other interfaces. The Advanced Interface Bus (AIB) is an open wide parallel interface PHY standard for die-to-die communications. With the AIB logic design open sourced in early 2019, and development in progress on open source enablement of AIB physical design, development effort and expense for die-to-die interfacing is intended to drop dramatically. Interoperability is a key goal of an ecosystem of chiplets. At the PHY level (the electrical, signaling and mechanical interface), business and technical considerations will likely keep a plurality of PHY options. A variety of die-to-die PHYs is likely to continue forward in the same way that PCI Express, Ethernet and DDR coexist in a system. Still, interoperability at another level such as the link layer protocol is possible. In this presentation we will examine useful protocols between chiplets and recommend protocol interoperability paths forward.

Speakers
DK

David Kehlet

Research Scientist, Intel


Friday September 27, 2019 10:15 - 10:35
G103

10:30

10:30

OCP Critical Facilities Operations Framework
We are providing an OCP recommended common framework for how to operate datacenters in a secure, cost effective and energy-efficient manner. Two case studies are included during the presentation to show real-life use of the framework. The framework is aligned with the OCP Ready Facility Recognition Program, Process driven IT Service Management according to ITIL v4, Operational practices in line with the criteria of EN 50600, Uptime Institute Standard for Operational Sustainability, Facilities Management practices in line with the principles of EN 15221 as well as other industry experience and best practices, e.g. BICSI 009.

Speakers
avatar for Lucas Cardholm

Lucas Cardholm

Vice President Services, Coromatic Group
Mr Cardholm, LL.M. and MBA, is a renowned speaker and an international expert in the field of mission critical infrastructure and economics of security investments. He is appointed as independent expert to the European Commission FP6, FP7 and H2020. His works are published. LinkedIn... Read More →


Friday September 27, 2019 10:30 - 10:55
G106

10:30

MiniPack: Low power 100G fabric switch architecture, mechanical & thermal cooling design
MiniPack is Facebook 2nd generation 100G fabric switch solution which provides 128 100GE ports in a single chassis. It is a low power solution compared to its predecessors with only one switching element, and it has a modular architecture that supports multiple port interface types. Its innovative mechanical design brings modularity and better thermal performance. This workshop will go over hardware architecture, with details on mechanical & thermal design.

Speakers
SK

Sami Khan

Technical Program Manager, Facebook
avatar for Jimmy  Leung

Jimmy Leung

Mechanical Engineer, Facebook, Inc.


Friday September 27, 2019 10:30 - 10:55
G102

10:30

Open System Firmware From Scratch
Open System Firmware has been in discussion and development since early 2018. Much work has been done to prove the concept and give demos, but we need to do more to enable the broader ecosystem. This session is intended to walk people thru the process of downloading the necessary ingredients and building a firmware image that can boot an OCP system.

Speakers
avatar for Jonathan Zhang

Jonathan Zhang

Software Engineer, Facebook
Jonathan is passionate at architecting lower level system firmware/software to achieve computer system design goals. He used to work at software organization of hardware companies (silicon vendors), he is working at hardware organization of software company (hyper-scaler). He is thrilled... Read More →
avatar for David Hendricks

David Hendricks

Firmware Engineer, Facebook
David got his first taste of BIOS tweaking as a kid overclocking gaming PCs and was drawn into the coreboot world where he learned real firmware hacking. Since then he's developed, deployed, and advocated for open source firmware as part of Google's ChromeOS team and more recently... Read More →


Friday September 27, 2019 10:30 - 10:55
G104

10:30

OCP Access updates
This presentation will cover new contributions from Edgecore for Telco Access

Speakers
avatar for Jeffrey Catlin

Jeffrey Catlin

VP Technology, Edgecore Networks
Jeff is the Vice President of Technology at Edgecore Networks.  He is responsible for research and integration of technology into Edgecore's award winning Open Networking product portfolio.  Jeff is an industry recognized technology leader with more than 25 years of in-depth experience... Read More →


Friday September 27, 2019 10:30 - 10:55
G105

10:30

Cooling of OCP V2 - Door HEX Step 2
As a part of the ACS Door HEX group, we would like to present the Door HEX Step 2 compatible to the specification. Technical details, feasibility etc.

Speakers
avatar for Resul Altinkilic

Resul Altinkilic

Product Manager Global IT Key Accounts, Rittal
- Planning, control and monitoring of customized product & system solutions from development to market launch to end of life cycle- Edge Computing- High Performance Computing- High Performance Cooling- Open Compute Project


Friday September 27, 2019 10:30 - 10:55
Emerald Room

10:30

Leveraging OCP Yosemite servers to build more cost-efficient and performant AI Inferencing architectures
Production AI Inferencing has server, accelerator, and infrastructure requirements for optimal performance but often uses expensive and proprietary data center infrastructures. A scale-out Inferencing infrastructure design requires optimum use of general-purpose compute elements married with networking, security and storage elements. Leveraging the innovative Yosemite server design, we show how increasingly efficient, performant, secure and open designs can be achieved.

Speakers
avatar for Antonis Karalis

Antonis Karalis

Netronome


Friday September 27, 2019 10:30 - 10:55
G107

10:40

The Chiplet Design Exchange
Machine-readable chiplet data exchange is the key for design automation, especially for modular designs. Chiplet's need to be described in terms of their physical and electrical characteristics at the very least but more importantly the key architectural description can help with design planning. There are business challenges for smooth data sharing, IP protection, and information control and a standard will help in the creation of the ecosystem. Based on zGlue Chiplet Exchange Format (ZEF), we are surveying and specifying standard practices and formats of Chiplet Data Exchange (CDX).

Speakers

Friday September 27, 2019 10:40 - 11:00
G103

10:45

Towards Industrialized Performance and Quality Control for Open Compute Hardware
This work, as part of a thesis project, falls within the challenging context of HPC manufacturing. In this paper, we propose an automated and highly adaptable solution for HPC performance benchmarks and quality controls to work towards rapid performance gains and to ensure consistently high production quality. Here we present our early-stage results.


Friday September 27, 2019 10:45 - 11:00
Auditorium Theatre

11:00

Computational Aspects of Nano Memories
Authors have created 16 bits memory using nano memristors with a 4x4 crossbar array architecture capable of reading and writing the data. The memristor has a 5 nm thick TiO2 and oxygen poor TiO2-X layers sandwiched between platinum (metal) plates. These TiO2 region creates a doped region whose resistance is lower than the undoped region. The total thickness of the memristor is 10 nm of which 5 nm is considered as doped and the rest is undoped region. The memory is modeled, simulated and characterized in Matlab Simulink environment. The proposed crossbar memory architecture using memristor becomes denser than existing RAMs reducing the silicon area and the power consumption. Crossbar architectures are easier to fabricate and have good noise tolerance capabilities. The amount of power consumed grows linearly with the size of the crossbar matrix due to alternate current paths thus lowering the overall effective resistance. In this work, we are providing simulation results for four cases when the width of doped region is i) much less than, ii) half, iii) equal and iv) much larger than the total oxide thickness. The timing and ON resistance of the memristors crossbar are provided in the paper.

Speakers
SH

Sotoudeh Hamedi-Hagh

San Jose State University
LH

Lili He

San Jose State University


Friday September 27, 2019 11:00 - 11:15
Auditorium Theatre

11:00

The four OCP principles reshaping the data centre facility
Choosing OCP bring big benefits for the data centre. The revolution in the facility is happening but quite out of sight. The deployment of OPC hardware is increasing fast now, but where are the savings for the facility? The principles of OPC; Efficiency, Scale, Openness and Impact are also valid for the datacenter facility. The OCP hardware needs a facilities and also the other way around. Therefore discussion in a broader audience around future thinking, reshaping, collaboration and optimization for the data centre facility is key. 

Last regional summit in Amsterdam the four principles were presented which enables the benefits for the facility. As follow up, at the summit in San Jose, a panel discussion was held with representatives from Facebook. Nokia, Berkley, Salesforce and Royal HaskoningDHV with as moderator OCP lead Robert Bunger.

This session will recap:
- the 4 OCP principles
- the feedback from the OCP Global Summit panel discussion in San Jose
- looking a step ahead for the data centre footprint reduction by OCP thinking 

Stijn de Kruijf, working with Royal HaskoningDHV, a large and independent engineering company, is working on datacenter facility design for over 16 years. In this period he is involved in large colo-cation, hyperscale and wholesale datacenters.

Speakers
avatar for Stijn de Kruijf

Stijn de Kruijf

Data centre facility expert, Royal HaskoningDHV


Friday September 27, 2019 11:00 - 11:25
G106
  • YouTube Link -

11:00

CLOS design with Facebook Minipack
In March Facebook announced the Mini-pack, a switch capable of providing 128 100GE ports. Cumulus Linux runs on this model as well (https://cumulusnetworks.com/blog/first-power-facebooks-minipack/). In this presentation, I would like to cover if and how this switch model based on the Tomahawk3 chipset might change a typical CLOS (spine/leaf) design. How would someone build a scalable design with these switches, what are the capabilities and caveats in this. A few key points: - Scalability - Multi DC architectures - Availability zones and redundancy - Overlay networks (EVPN VxLAN) - Exit routing - Automation and troubleshooting

Speakers
AL

Andreas La Quiante

SE, Cumulus Networks
I like networking, especially open networking.


Friday September 27, 2019 11:00 - 11:25
G102

11:00

Open System Firmware support for multi socket OCP platforms - coreboot POC on TiogaPass and Skylake SP
Open Systems Firmware is gaining traction in cloud service provider deployments. There is an increasing interest in extending coreboot support for multi socket Xeon SP platforms. In this talk, we outline coreboot porting effort on Tioga Pass 2 Socket Xeon Skylake SP platform using Skylake SP Intel® Firmware Support Package. We will cover coreboot changes needed to support multi-socket OCP platforms and show working demo on OCP Tioga Pass board with coreboot.

Speakers
avatar for Jonathan Zhang

Jonathan Zhang

Software Engineer, Facebook
Jonathan is passionate at architecting lower level system firmware/software to achieve computer system design goals. He used to work at software organization of hardware companies (silicon vendors), he is working at hardware organization of software company (hyper-scaler). He is thrilled... Read More →
AR

Anjaneya Reddy Chagam

Senior Principal Engineer and Chief Software-Defined Storage Architect, Intel
Anjaneya “Reddy” Chagam is a Senior Principal Engineer and Chief Software-Defined Storage Architect in Intel’s Data Center Group. He is responsible for developing software-defined storage strategy, architecture, and technology initiatives. He is a member of the Technical Steering... Read More →


Friday September 27, 2019 11:00 - 11:25
G104

11:00

openEDGE Ecosystem Opportunity
A review of the current state of the openEDGE project, including new contributions, new initiatives, and exploring various deployment opportunities for openEDGE.

Speakers
avatar for Mike Moore

Mike Moore

Regional Product Manager, Nokia
Over 20 years of R&D experience in both the Wireless and Defense Industries prior to transitioning into a customer facing sales and support role.  Now the Western Hemisphere Regional Product Manager for Nokia’s AirFrame Data Center Portfolio, with a focus on creating highly available... Read More →


Friday September 27, 2019 11:00 - 11:25
G105

11:00

Eco-system Enabling of Liquid Cooling Ingredients Update
The continued demand for increased compute performance comes with an increase in server power and power density. The increased server power requires more efficient cooling capabilities compared to traditionally used air cooling. Direct liquid cooling using cold plates can provide the needed cooling capability, as well as fit within a small volume to allow for greater server and compute density in the rack. Even though liquid cooling is not a new cooling strategy for IT equipment, the industry is not ready to efficiently mass deploy liquid cooling solutions. Therefore, Intel is working to enable liquid cooling ingredients with industry partners. The intent is to create more available, lower risk, and faster time to market opportunity for OEMs to deploy liquid cooling solutions. Intel’s efforts are specifically focused on manifold distributed liquid cooling at the server rack level (i.e. the technology cooling system, TCS). The TCS is the fluid loop providing the cooling liquid from the Cooling Distribution Unit (CDU) through the rack manifold to the IT equipment cold plates, and then back again to the CDU through the rack manifold. The liquid cooling ingredients in the secondary loop included in Intel’s enabling efforts are: cooling liquid, CDU, interchangeable Universal Quick Disconnects (UQD), and cold plates. Intel’s enabling efforts, together with Intel’s strategy to enable these ingredients was first introduced at the OCP Global Summit in March 2019, and an update on the eco-system enabling status will be presented at the OCP Regional Summit.

Speakers
JG

Jessica Gullbrand

Tech Lead Engineer, Intel Corporation


Friday September 27, 2019 11:00 - 11:25
Emerald Room

11:05

ODSA PoC
Developing a proof-of-concept (PoC) for an emerging technology is important to learn what can and cannot be easily done, to reduce risk for future implementations, and lastly to convince skeptics that the technology is viable. This is true for the Open Domain-Specific Architecture (ODSA) effort where our PoC effort is trying to simultaneously address three dimensions of standardizing chiplet development. First, we need to address architectural issues like validating interface protocols, evaluating performance trade-offs, and developing software programming models. Second, we need to explore physical challenges like packaging issues, power distribution limitations, and high-speed I/O interfaces. Third, we need to expose the business issues that are unique to chiplets like how to share highly sensitive information on silicon design and business metrics that are exposed at the bare die-level. For the ODSA PoC workstream we have chosen to implement a module that can serve as a Smart Network Interface Card and also potentially as a Computation Storage Solution. This is an area where people are likely to create future domain-specific accelerators and represents a problem hard enough to expose real-world issues. It is also an area where members of the ODSA community have expertise and existing silicon components that can be leverage for creating a PoC. The physical implementation of the PoC involves a package design and a software development board created with discrete components. The PoC workstream focuses on a) architecture, b) physical implementation of package and board, c) bring up and testing. The board is architected in a modular manner such that the board design effort directly translates to the substrate design.

Speakers
QJ

Quinn Jacobson

Strategic Architect, Achronix


Friday September 27, 2019 11:05 - 11:25
G103

11:15

High Efficiency Data Center PSU Based on Multiphase Synchronous Buck Converter with Extended Duty Cycle
The high power consumption of data centers confronts the managers with both financial and environmental challenges. One solution to the power system optimization that deserves serious consideration is DC power, since the AC power must be converted to DC for use by IT system components. In this paper, a high-efficiency Power Supply Unit (PSU) based on multi-phase synchronous DC to DC buck converter with extended duty cycle designed and discussed which that decrease power consumption, heat losses and increase power transmission range to load in data centers.

Speakers
AM

Amir Mardani

Qazvin Azad University
AM

Alireza Malmali

Qazvin Asad University


Friday September 27, 2019 11:15 - 11:30
Auditorium Theatre

11:30

Modular Data Center Group Update
Modular construction is not a new concept, but has a lagging adoption rate in the data center market. Our unique places in the data center market created knowledge isolation within sales, engineering, and production groups .The OCP-MDC group was created to provide a venue for shared knowledge to modular builders, component/hardware manufacturers, and end users. Outline: In this presentation, attendees will learn: Problems Modular Aims to Solve Deployment Speed Consistent Quality Consistent Cost Opportunity for Continual Improvement Status of our current specification and opportunities for it to improve. Product submission status, and how to submit for approval to the marketplace. Future Plans Modular components of a data center Power Cooling Compute Gaps In OCP-MDC Offerings Call To Action - Join our group to shape the OCP Modular Data Center

Speakers
avatar for Roberto Söderhäll

Roberto Söderhäll

Chief Business Development Officer, Swedish Modules
Roberto is a veteran in working with innovation within the Data Center Industry. Spending more than two decades in the Data Center Industry, Roberto has been challenging the Status Quo of the Data Center Industry. As a speaker in many events his been proclaiming the Data Center as... Read More →
avatar for Rob Coyle

Rob Coyle

Director of Sales, PCX Corporation
As the son, grandson, and brother of carpenters, my education in construction began early. This background, combined with my passion for innovative technology solutions, led me to modular construction.The construction industry has been historically slow-to-change, yet the problems... Read More →


Friday September 27, 2019 11:30 - 11:55
G106
  • YouTube Link -

11:30

Devlink Linux
We will describe the specific Facebook operational challenges, the integration of Devlink, and the impact with respect to fabric health monitoring and provisioning in data center operations. Devlink is a Linux upstream provisioning, health reporting and recovery system that reduces significantly the disruptive debug efforts with real-time monitoring of PCIe devices’ health and self-healing capabilities. Devlink unified and centralized driver reports aims to replace existing sporadic vendor-specific tools and the complexities in integrating them in multi-vendor data centers. In addition devlink provides provisioning capabilities to Network Interface Cards (NICs).

Speakers

Friday September 27, 2019 11:30 - 11:55
G102

11:30

Passing System Configuration Data from Firmware to Kernel
Firmware relies on intimate system knowledge for memory initialization and training. However, there is currently no standard method for passing this data to the operating system. The OS kernel and other monitor software relies on platform-specific drivers, but these are often based on proprietary tables using confidential information. Intel currently publishes a proprietary BIOS ACPI Data Table (BDAT) to convey system, such as DDR4/DDR5 training margins. This session proposes moving this data to a format for use by any open source firmware project. The proposed method can be implemented by a complete UEFI stack, or low-level firmware based on PEI driver architecture. This session will also describe next steps for implementation, as well EDK II example code available from TianoCore.

Speakers
avatar for Natarajan V Sivagar

Natarajan V Sivagar

Software Engineer, Google Inc
avatar for Sarathy Jayakumar

Sarathy Jayakumar

Principal Engineer, Intel Corp


Friday September 27, 2019 11:30 - 11:55
G104

11:30

openEDGE Chassis Form Factor Development
2U high open edge chassis addresses the need for even more compact edge deployments where number of servers needed is between 1 to 3 server sleds. The chassis supports the same server designs as the already introduced 3U high open edge chassis, including the same power supplies and RMC module.

Speakers
MO

Mikko Ohvo

Product Manager, Nokia


Friday September 27, 2019 11:30 - 11:55
G105

11:30

Immersed Computing® optimised IT platforms
This session covers the way in which Asperitas designs IT platforms by maximising the benefits of OCP hardware and combining it with immersion technology.

Speakers
avatar for Rolf Brink

Rolf Brink

CEO, Asperitas
Product development for cloud architectures and datacentre infrastructure, combined with a passion for cleantech innovations like liquid cooling, are the foundation of the developments which drive Asperitas.Asperitas is a clean-tech company focused on greening the datacentre industry... Read More →



Friday September 27, 2019 11:30 - 11:55
Emerald Room

11:30

The Impact of Domain Specific Acceleration and Chiplet Technology on the Semiconductor and Electronics Business
As we reach the end of Moore’s Law, the semiconductor industry is looking for new ways to continue to deliver ever increasing processing capabilities to customers. Domain specific acceleration and chiplet technology are complementary technologies that show a lot of promise in this regard. This session will review the technical and business challenges that the microelectronics industry currently faces and how the ODSA workgroup is working to solve them. It will also review the new business models that must emerge to support this new advancement in microelectronics.

Speakers
SF

Sam Fuller

Director, NXP


Friday September 27, 2019 11:30 - 12:00
G103

12:00

Lunch
Friday September 27, 2019 12:00 - 13:00
Expo Hall
  • YouTube Link -

13:00

Open Accelerator Infrastructure Overview
As a standard module form factor to accommodate accelerators from different suppliers, we contributed the OCP Accelerator Module (OAM) spec to OCP in March 2019.  Then, we formed a subgroup and encouraged other OCP members to join a team effort to build a modularly interoperable infrastructure around OAM.  Many companies have joined.  Open Accelerator Infrastructure (OAI) subgroup operates under OCP Server Project. The OAI Track is a progress report.

Speakers
avatar for Siamak Tavallaei

Siamak Tavallaei

Principal Architect, Microsoft
Siamak Tavallaei is a Principal Architect at Microsoft’s Azure division. Collaborating with industry partners, he drives a number of initiatives in research, design, and deployment of hardware for Microsoft’s cloud-scale services such as Azure, Machine Learning, Bing, Office 365... Read More →


Friday September 27, 2019 13:00 - 13:15
G103

13:00

The Road to Minipack Operationalization
In this workshop we will talk about the road of Minipack operationalization, what were the challenges faced, how did we overcame then by automating our testing pipeline. We'll also discuss what are the benefits that this platform brings to the teams running the network at Facebook. 

Speakers
avatar for Tom Harbert

Tom Harbert

Production Network Engineer, Facebook
avatar for Guilherme Loch Waltrick Goes

Guilherme Loch Waltrick Goes

Production Network Engineer, Facebook


Friday September 27, 2019 13:00 - 13:25
G102

13:00

OCP's Open Rack Manager Controller subproject
In Q4 2018, OCP established the OpenRMC subproject to architect and implement an open-source rack management controller. The OpenRMC will allow the rack to be managed via a Redfish interface. As a fast cycle deliverable, OpenRMC has selected the RMC on an OpenEdge platform as a target platform. This presentation will present the project’s progress and review its collateral.

Speakers
avatar for John Leung

John Leung

Principal Engineer, Intel Corporation
John is a Principal Engineer at Intel Corporation in the Data Center Group.  John has been in the computer industry for over 37 years: architecting, designing and coding.  At Intel, John is currently focused on the manageability interfaces for Intel® Rack Scale Design and server... Read More →


Friday September 27, 2019 13:00 - 13:25
G104

13:00

openEDGE on Arm Architecture
In Akraino Edge Stack Radio Edge Cloud blueprint, multiple architecture needs to be supported for the hardware platform choosen based on OCP openEDGE. This session will give an introduction how Arm architecture can be enabled in OCP openEDGE.

Speakers
DW

Dong Wei

Fellow, Arm


Friday September 27, 2019 13:00 - 13:25
G105

13:00

The Intel Transparent Supply Chain
The globalization of technology development, manufacturing and distribution has led to growing industry concern that bad actors can compromise the security of a computing platform by introducing counterfeit parts, malicious logic, or other vulnerabilities through the supply chain. Intel, as a recognized leader in the areas of supply chain quality and conflict-free minerals, has begun to actively address computing hardware and firmware supply chain security needs with initial service and product offerings.

In this session, we will share Intel’s viewpoint on strengthening security assurance based on a foundation of greater transparency into the provenance of the components which make up a platform. We will also discuss our vision to achieve a shared, federated model with supply chain partners and the development of more robust solutions in the future, from initial component sourcing through platform build and then through the rest of the device lifecycle."

Speakers
avatar for Catherine Spence

Catherine Spence

Senior Principal Engineer, Intel
Catherine Spence is the Commercial Client Chief Segment Architect in Intel’s Business Client Platform team. Her focus is on the future strategy, direction, and architecture for specialized features for commercial devices, especially for Intel® vPro™ premium platforms. Catherine... Read More →


Friday September 27, 2019 13:00 - 13:25
G107

13:00

OCP Ready Program for Colocation Data Center Facilities & Hydro66 OCP Ready™ Facility recognition process experience
  1. The OCP Foundation developed the requirements and process for a colocation data center facility provider to become a listed Colo Solution Provider on the OCP Marketplace. The facility provider must perform a self-assessment of their facility using the site assessment scorecard, which is presented to and reviewed by the Data Center Facility Project Community and the OCP Incubation Committee. The facility provider must also meet all requirements set forth by the OCP Foundation. The pilot program has now been completed and the program was officially launched at the OCP Summit in March. The program is now in full swing to bring new colocation data center facilities on to the program, from all global regions and countries. This presentation will provide an update on the program, and feedback from Hydro66 a colocation facility located in Luleå Sweden, who will share their experiences in applying for and gaining the OCP Ready™ Facility status.

Speakers
avatar for Andy Long

Andy Long

Director Strategic Development, Hydro66
Making OCP compliant rackspace available in a 100% renewable low cost multi-tenant colocation datacentre
avatar for Mark Dansie

Mark Dansie

Data Centre Transformation Specialist - EMEA, OCP Foundation
Mark Dansie has been involved in the computing and IT infrastructure services industry for over 36 years and is the OCP Ready TM Representative - EMEA. His career experience in consultancy, project and programme management, spans all aspects of building, campus and data center infrastructure... Read More →


Friday September 27, 2019 13:00 - 13:50
G106

13:20

OAI-UBB: Universal baseboard for OCP Accelerator Module
Artificial Intelligence (AI) applications are rapidly evolving and producing an explosion of new types of hardware accelerators for Machine Learning (ML), Deep Learning (DL), and High-Performance Computing (HPC). Different implementations target similar requirements for power/cooling, robustness, serviceability, configuration, programming, management and debug, as well as inter-module communication to scale up and input/output bandwidth to scale out. We need an open infrastructure to intercept rapid innovation in artificial intelligence. OAI is where open accelerator infrastructure meets open artificial intelligence. We contributed the OAM design specification during 2019 OCP Global summit, in this workshop we will present our universal OAM baseboard design proposal from multiple hyperscalers and different OAM suppliers.

Speakers
avatar for Siamak Tavallaei

Siamak Tavallaei

Principal Architect, Microsoft
Siamak Tavallaei is a Principal Architect at Microsoft’s Azure division. Collaborating with industry partners, he drives a number of initiatives in research, design, and deployment of hardware for Microsoft’s cloud-scale services such as Azure, Machine Learning, Bing, Office 365... Read More →
avatar for Richard Ding

Richard Ding

AI System Architect, Baidu
Richard Ding is AI System Architect for heterogeneous computing in Technical Group of Baidu. He leads architecture design of Baidu's AI computing platform X-MAN, the high-performance parallel file system FAST-F, and the large-scale training cluster KongMing. His research focuses... Read More →
avatar for Whitney Zhao

Whitney Zhao

Hardware Engineer, Facebook Inc.
Whitney is a seasoned hardware engineer leading AI/ML system design in Facebook. Whitney has led multiple hardware generations ranging from general purpose 2S system such as Tioga Pass to ML JBOG Big Basin systems, all of which have been contributed to OCP. She has been driving multiple... Read More →


Friday September 27, 2019 13:20 - 13:40
G103

13:30

Enabling PIM card hot swapping in SONiC for Minipack
Minipack is a new building block switch based on Broadcom Tomahawk 3 ASIC, in which two types of PIM cards are currently supported 100GbE and 400GbE. Since SONiC doesn’t support dynamic port creation and deletion, we have to build different images with specific ASIC and port configurations to fit combination of the two types of cards. To resolve the problem, we propose and enable hot swapping on PIM cards by capturing card plug/unplug events on the fly and dynamically changing ASIC and port configurations.

Speakers
DL

Deming Liu

Principal Engineer, Edgecore Networks


Friday September 27, 2019 13:30 - 13:55
G102

13:30

OpenBMC: Updates on current status and Telemetry framework
OpenBMC is a Linux distro for Baseboard Management Controllers. It is an inclusive and active community. First half of this session will provide information about different collaborative workgroups, release cadence, key features in current and upcoming releases and how to get involved. Second half will zoom in on design and progress of Telemetry framework in OpenBMC given that BMC plays a critical role in ensuring reliability, and serviceability at cloud scale. Accurate, consistent, and timely telemetry can help in root-causing, mitigating or preventing interruption to ensure high availability.

Speakers
SD

Sagar Dharia

Microsoft
NL

Neeraj Ladkani

Senior Software Engineer, Microsoft


Friday September 27, 2019 13:30 - 13:55
G104

13:30

Ampere for the openEDGE
Speakers
SH

Shannon Hollas

Senior Director, Ampere


Friday September 27, 2019 13:30 - 13:55
G105

13:30

Experiences with Immersion Cooling in Alibaba Datacenter
With the improvement of computing performance, the power consumption is significantly increased. Increasing power consumption of server components brings enormous challenges to heat dissipation and power optimization. Immersion cooling completely breaking the air-cooling power dissipation bottleneck and also achieving extremely low PUE. This presentation introduces the problems encountered by Alibaba in the scale deployment of immersion cooling data center, solutions and considerations.

Speakers
YZ

Yangfan Zhong

Senior Engineer, Alibaba


Friday September 27, 2019 13:30 - 13:55
Emerald Room

13:30

Using Secure-Boot on OCP NIC controllers to prevent supply-chain and hardware cloning attacks
Datacenters HW attacks are a serious security concern. There is an obvious need to prevent supply-chain attacks which secure-boot enabled devices can address. Yet attackers can exploit security breaches by cloning legitimate cards. By combining both cloning-protection capabilities with secure boot, OCP NIC controller devices can prevent both attack methods.

Speakers
EW

Elad Wind

Mellanox
YI

Yuval Itkin

Distinguised Architect, Mellanox
Servers management and OCP server hardware.


Friday September 27, 2019 13:30 - 13:55
G107

13:45

OAM Reference Systems Joint Review
OAI (Open Accelerator Infrastructure) is where open accelerator infrastructure meets open artificial intelligence. In this year’s OCP Global Summit in San Jose, we contributed the OAM (Open Accelerator Module) design specification, and have since been working with multiple partners to enable this form factor. Now, we are building upon the surrounding Infrastructure and target at a common baseboard design proposal. In this workshop, we will present our reference system designs which have been developed based on the common baseboard design. These reference designs, presented by Inspur, Hyve Design Solutions, and ZT Systems, show how different connection topologies and rack dimensions can be supported.

Speakers
avatar for Mark Chubb

Mark Chubb

Director of Architecture, ZT Systems
Mark is the Director of Architecture at ZT Systems.  Mark is based in Seattle, WA and works closely with key Cloud Customers to conceptualize and develop Server and Rack solutions for Hyperscale computing needs.  Mark has over 20 years of experience in HW Design and Architecture... Read More →
avatar for Emilie Yang

Emilie Yang

Technical Program Manager, Hyve Design Solutions
Emilie is a senior Technical Program Manager at Hyve Design Solutions in Taiwan.  She contributed several OCP hardware and cooling designs in past decade.  Now, she continuously delivers various products through program management and technical expertise to OCP community and cu... Read More →
avatar for Adam Tin

Adam Tin

System Engineer, Inspur
System Engineer at Inspur based in Taiwan. Over 16 years of server hardware design and architecture. As system leader to provide various form factor of server solution for target customer in US and Europe.


Friday September 27, 2019 13:45 - 14:25
G103

14:00

PCX 90kw Modular Data Center Marketplace Addition
In anticipation of our MDC product being accepted to the marketplace. Our submission will comply with and is based on the 90kw MDC Specification PCX contributed to. PCX would take this opportunity to present the technical data about the construction and efficiency the 90kw OCP-MDC provides. Additionally, we would speak to the advantages OCP hardware has/provides in a module like ours. Lastly we would share opportunities where new OPEN based hardware (from other companies) could improve future MDCs.

Speakers
avatar for Rob Coyle

Rob Coyle

Director of Sales, PCX Corporation
As the son, grandson, and brother of carpenters, my education in construction began early. This background, combined with my passion for innovative technology solutions, led me to modular construction.The construction industry has been historically slow-to-change, yet the problems... Read More →


Friday September 27, 2019 14:00 - 14:25
G106

14:00

Programmable silicon use-cases in the disaggregated network
Hardware – software disaggregation powered by ONIE, baremetal switches and a new breed of network operating systems held a lot of initial promise to IT architects. However aside from potential cost savings, the types of networking solutions that could be built for the mass market remained the same as before. In this session, we will discuss deployments with programmable switch and NIC ASICs powered by P4. We will talk about the new use-cases that are being realized such as advanced routing and switching, next generation network packet brokers, dataplane telemetry and scalable in-network offload. We will discuss use-cases in context of specific disaggregated operating systems such as SONiC, Stratum and IP Infusion’s OcNOS.

Speakers
avatar for Arkadiy Shapiro

Arkadiy Shapiro

Product Line Manager, Barefoot Networks, an Intel Company
Arkadiy Shapiro is a Product Line Manager at Barefoot Networks, an Intel company. In this role, he is responsible for product strategy and marketing of Software Development Environment that enables customers to create programmable network solutions with Barefoot’s ASICs. In the... Read More →


Friday September 27, 2019 14:00 - 14:25
G102

14:00

Inspur RMC solution facilitates the datacenter management
Inspur has been devoting to the simplification of data center management for years. Based on its experiences and collaboration with hyper scale DC owners, Inspur has developed a solution to accommodate different compute nodes in the rack and manage them from a single stand point. Wilson will introduce the new feature of the management solution and talk about the usage model to the audiences.

Speakers
avatar for Wilson Guo

Wilson Guo

Technology Director, Inspur


Friday September 27, 2019 14:00 - 14:25
G104

14:00

OpenEDGE Product plans from ASUS
Speakers
NW

Nelson Wang

General Manger, Enterprise Solutions, ASUS
On behalf of the Enterprise Solutions team at Asus, it is a great honor for us to join the Open Compute Project Foundation as a Platinum Member. Asus looks forward to participate, learn, and contribute to the OCP community.


Friday September 27, 2019 14:00 - 14:25
G105

14:00

Open Rack V3 Frame Spec Update
This will be an overview of the work the group is doing on the Open Rack V3 Specification.

Speakers
avatar for Steve Mills

Steve Mills

Technical Lead, Facebook
Open Rack Project Co-Lead


Friday September 27, 2019 14:00 - 14:25
Emerald Room

14:30

Datacenter-ready Secure Control Module for OAI (OAI-SCM)
The Modular Building Block Architecture (MBA) is a Catalyst for interoperable Innovation!  Clearly defined physical dimensions, power, cooling, management, and input/output ports enables Module interoperability in various OAI Systems.  OAI-SCM facilitates MBA.  As a datacenter-ready secure control module for OAI, OAI-SCM streamlines the design and deployment of OAI Modules for Hyperscale Datacenters by driving a common and secure monitoring, control, and remote debug for various OAI Modules and systems (firmware, diagnostic tools, manufacturing tools, at-scale remote debug.

Speakers
avatar for Siamak Tavallaei

Siamak Tavallaei

Principal Architect, Microsoft
Siamak Tavallaei is a Principal Architect at Microsoft’s Azure division. Collaborating with industry partners, he drives a number of initiatives in research, design, and deployment of hardware for Microsoft’s cloud-scale services such as Azure, Machine Learning, Bing, Office 365... Read More →


Friday September 27, 2019 14:30 - 14:50
G103

14:30

Datacenter Synergy Ecosystem
This session covers the way in which datacentres can be transformed in energy production facilities and how this impacts the operating and business models around how we are used to work with datacentres.

Speakers
avatar for Wim Buters

Wim Buters

Principal Commercial Officer, Asperitas
Seasoned professional on business-to-business delivery. Accountable for further expansion of Asperitas' commercial position on the market. Asperitas being developer, supplier and producer of high-end solutions for the data centre industry based on immersed compute technology.


Friday September 27, 2019 14:30 - 14:55
G106

14:30

Enabling OCP GPON Optical Line Terminals (OLT) using DmOS
The process of standardization of networking platforms is moving beyond datacenter frontiers, reaching access and transport networks. EdgeCore Networks announced almost one year ago the contribution of the first PON optical line terminal (OLT), unveiling a line of OCP-related GPON OLTs. Almost at the same time, Deutsche Telekom published a design of another GPON OLT alternative. In parallel, ONF has been promoting the disaggregation of the software side through the VOLTHA Project. However, while VOLTHA does not reach the necessary maturity and completeness, all those design contributions are useless since there is no other software alternative in the market, open source or commercial. DmOS - Datacom NOS - was developed as a disaggregated NOS by design and it was initially introduced in the market in Datacom’s first GPON OLT. When DmOS started being ported to different OCP switches, a natural next step was to extend its reach to GPON devices. We would like to demonstrate how a modular architecture can extend a switch-oriented NOS by adding support to optical devices and being prepared to take advantage of the abstraction layers that are being developed in the context of the VOLTHA project.

Speakers
avatar for Ricardo Pianta

Ricardo Pianta

Engineering Manager, Datacom - Teracom Telematica


Friday September 27, 2019 14:30 - 14:55
G102

14:30

Platform Hardware Error Handling Requirements for Large Scale Data Center
To date, there's no clear defined requirements on how hardware error should be handled in OCP platforms. For a large scale data center, lack of solid error handling will make operation inefficient and increase operation cost while slow down remediation time. A generally defined common hardware error handling goal with a unified, scalable and platform performance aware hardware error handling requirement for various OCP platforms is proposed here, to help provide general reference and guidance on how OCP platform providers should implement the hardware error handling solutions.

Speakers
AA

Anil Agrawal

RAS Technical Lead, Intel Corp.
Anil has been working as electrical engineer at Intel since 2000. He is passionate about solving customer issues related to system architecture, design, and validation. His focus has been to develop system fault handling solutions. Please refer to Anil's linked-in profile for more... Read More →


Friday September 27, 2019 14:30 - 14:55
G104

14:30

openEDGE RMC
Nokia has contributed the openEDGE chassis as OCP accepted product. As part of chassis solution Nokia has developed a rack management controller (RMC). Nokia’s plan is to contribute the RMC design as OCP Accepted product. This presentation highlights the main functionalities of RMC as part of the openEDGE chassis.

Speakers
TM

Tomi Männikkö

Product Architect, Nokia


Friday September 27, 2019 14:30 - 14:55
G105

14:30

Open Rack V3 Power System
Overview of Open Rack V3 Power and Battery system.

Speakers
avatar for Shaun Harris

Shaun Harris

Director of Engineering, Windows Cloud Server, Microsoft
avatar for Hamid Keyhani

Hamid Keyhani

Technical Lead, Facebook


Friday September 27, 2019 14:30 - 14:55
Emerald Room

14:55

Cooling solution on OAM Reference systems
Open Accelerator Infrastructure (OAI) is a platform built around the Open Accelerator Module (OAM). Earlier this year, we contributed the OAM specification with a reference air-cooled heatsink solution that could support up to 450W (the module itself rated up to 700W). Advanced cooling solutions become necessary to support future OAM products with higher power. The OAI group is working together to standardize these cooling solutions and develop guidelines to help enablement of the OAM products and OAM-based platforms. In this talk, we will share the liquid cooling solution used in Baidu's X-MAN4.0, and the current status of the OAI community's standardized cooling solution at the chassis level. In addition, we will share Baidu’s liquid cooling solutions at the rack and data center facility level

Speakers
avatar for Tiffany Jin

Tiffany Jin

Mechanical Engineer, Facebook
avatar for Richard Ding

Richard Ding

AI System Architect, Baidu
Richard Ding is AI System Architect for heterogeneous computing in Technical Group of Baidu. He leads architecture design of Baidu's AI computing platform X-MAN, the high-performance parallel file system FAST-F, and the large-scale training cluster KongMing. His research focuses... Read More →


Friday September 27, 2019 14:55 - 15:15
G103

15:00

Re-imagining the Multi-tenant DC Gateway Router using Open Whitebox Switching
Even as open whitebox switching takes over data center networks, it has barely touched the DC gateway router. In a shared, multi-tenant infrastructure, routing northbound traffic to different providers while maintaining the segmentation among tenants is traditionally achieved by using multiple VRFs. This presentation discusses a new approach, building on innovations in whitebox internal hardware architecture and network operating systems. This approach brings in an additional level of network slicing, dedicating distinct virtual routers to the different tenants, while still being able to partition the network once more within each layer 3 space by using the legacy VRFs. For the service provider, this approach translates to high multi-tenant scalability, and for the tenants, it translates into extra layers of security, flexibility, control over shared resource usage and even mobility, considering that these virtual routers could be migrated from one physical box to another.

Speakers
avatar for Andreas Falkner

Andreas Falkner

Technical Sales Director EMEA, Pluribus Networks


Friday September 27, 2019 15:00 - 15:25
G102

15:00

Server Base Manageability Guide for SBSA Compliant Arm (AARCH64) Servers.
This presentation guides the Arm Server System designers towards the common foundation for Server Management. Common manageability capabilities are standardized while allowing scope for value add customization. Server Base Manageability Guide for SBSA Compliant Arm (AARCH64) Servers leverages industry standard system management specifications including but not limited to Redfish, Platform Level Data Model (PLDM), Management Component Transport Protocol (MCTP) as defined by the Distributed Management Task Force (DMTF), Hardware Management Specifications and designs as defined by Open Compute Group (OCP). This presentation will further introduce the audience to manageability capabilities in SBSA Compliant Arm (AARCH64) servers including but not limited to Remote Debug, Reliability, Availability, Serviceability features (RAS), platform monitoring, control and provide status update on Arm Server Base Manageability Guide specification.

Speakers
DW

Dong Wei

Fellow, Arm


Friday September 27, 2019 15:00 - 15:25
G104

15:00

openEDGE Server
Nokia has contributed the openEDGE chassis as OCP accepted product. As part of openEDGE solution Nokia has also contributed 1U and 2U server sleds as OCP inspired products. This presentation highlights the main functionalities and use cases of openEDGE servers.

Speakers
TM

Tomi Männikkö

Product Architect, Nokia


Friday September 27, 2019 15:00 - 15:25
G105

15:00

Multi vendor Interop Specification Review
In the past sessions I have presented team's work on challenges and solutions for Interoperability covering both electrical and Mechanical aspects of the ORV2 design. In this session I plan to present a methodology as to how the team takes those learnings to develop interoperability specifications.

Speakers
avatar for Harry Soin

Harry Soin

Senior Director Technical Marketing, Artesyn Embedded Technologies


Friday September 27, 2019 15:00 - 15:25
Emerald Room

15:20

OAM Panel discussion
Speakers
avatar for Ola Tørudbakken

Ola Tørudbakken

SVP/GM Rackscale, Graphcore
Ola is SVP at Graphcorewhere he leads the Rackscale team building IPU-POD scaleout systems designed to support next-generation deep learning and machine intelligence workloads. He´s a co-founder/CTO of Skala Technologies AS. Previously he was Chief Architect of Networking at Oracle... Read More →
avatar for Siamak Tavallaei

Siamak Tavallaei

Principal Architect, Microsoft
Siamak Tavallaei is a Principal Architect at Microsoft’s Azure division. Collaborating with industry partners, he drives a number of initiatives in research, design, and deployment of hardware for Microsoft’s cloud-scale services such as Azure, Machine Learning, Bing, Office 365... Read More →
avatar for Whitney Zhao

Whitney Zhao

Hardware Engineer, Facebook Inc.
Whitney is a seasoned hardware engineer leading AI/ML system design in Facebook. Whitney has led multiple hardware generations ranging from general purpose 2S system such as Tioga Pass to ML JBOG Big Basin systems, all of which have been contributed to OCP. She has been driving multiple... Read More →
avatar for Robert Ober

Robert Ober

Chief Platform Architect, NVIDIA
Rob is NVIDIA’s data center Chief Platform Architect, working with Hyperscales to build GPU clusters for AI and Deep Learning, develop systems and platform architecture, and influence the HW and SW GPU roadmaps at NVIDIA. His interest in AI and DL was driven by its impact on computer... Read More →
avatar for Frank Helms

Frank Helms

Data Center GPU Portfolio Architect, AMD
Frank Helms is an AMD Senior Fellow and Data Center GPU Portfolio Architect. 30+ years computer industry experience ranging from Notebook PCs and power management breakthroughs to Server CPUs and GPUs in the Data Center. In 2018, Frank was part of the team that defined the OAM... Read More →
AR

AK Roy

Intel


Friday September 27, 2019 15:20 - 16:00
G103

15:30

White Boxes for BNG Use Case
DT's program Access 4.0 develops and engineers DT's next generation SDN-based access with initial focus on fixed line business FTTH/B/C. Key components are whitebox switches built on merchant silicon and open / modular software. In this talk, DT and Delta Networks will present to the community their current view on a switch specification supporting BNG function.

Speakers
CM

Carsten Michel

Senior Expert NGN, Deutsche Telekom
Carsten Michel is a Senior Expert NGN at Deutsche Telekom Technik GmbH working in DT’s Access 4.0 project team. He is responsible for defining hardware and software standards to enable disaggregation of service edge and data center functionality. Before joining DT, he worked as... Read More →
WL

Winnie Lin

Senior R&D Director, Advanced Switching Platforms, Delta Electronics Inc.
Winnie is leading an R&D team developing advanced switching platforms within Delta’s Networking business group. She is responsible for delivering open networking switches based on merchant silicon, which targeting at the carrier and hyper-scale datacenter markets. With more than... Read More →


Friday September 27, 2019 15:30 - 15:55
G102

15:30

44OU OCP Rack
The 44OU was developed to read the demands of the US telco customer. The upper busbars are flexible, that the networking switches can be installed. This work was done in cooperation with Penguin

Speakers
avatar for Steven Moore

Steven Moore

Project Engineer, RITTAL
Contact Information: Steven Moore Project Engineer Phone: +44 01752 207 654 Email: smoore@rittal-csm.co.ukI have worked for Rittal over the past year and a quarter, starting out as a Design Engineer and transitioning to a Project Engineer. I am now the Project Engineer of one of Rittal’s... Read More →


Friday September 27, 2019 15:30 - 15:55
Emerald Room