The Open Fabrics Verbs Working Group Pavel Shamis and Liran Liss.

Slides:



Advertisements
Similar presentations
Middleware Support for RDMA-based Data Transfer in Cloud Computing Yufei Ren, Tan Li, Dantong Yu, Shudong Jin, Thomas Robertazzi Department of Electrical.
Advertisements

The Development of Mellanox - NVIDIA GPUDirect over InfiniBand A New Model for GPU to GPU Communications Gilad Shainer.
Chapter 17 Networking Patricia Roy Manatee Community College, Venice, FL ©2008, Prentice Hall Operating Systems: Internals and Design Principles, 6/E William.
OFED TCP Port Mapper Proposal June 15, Overview Current NE020 Linux OFED driver uses host TCP/IP stack MAC and IP address for RDMA connections Hardware.
IWARP Update #OFADevWorkshop.
Cracow Grid Workshop, November 5-6, 2001 Towards the CrossGrid Architecture Marian Bubak, Marek Garbacz, Maciej Malawski, and Katarzyna Zając.
Keith Wiles DPACC vNF Overview and Proposed methods Keith Wiles – v0.5.
Chapter 17: Client/Server Computing Business Data Communications, 4e.
VIA and Its Extension To TCP/IP Network Yingping Lu Based on Paper “Queue Pair IP, …” by Philip Buonadonna.
© 2007 Pearson Education Inc., Upper Saddle River, NJ. All rights reserved.1 Computer Networks and Internets with Internet Applications, 4e By Douglas.
OKI Focus Groups at Educause, October 2002 Page 1 Open Knowledge Initiative Educause Focus Group Geoff Collier and Robby Robson, Eduworks Educause 2002,
Chapter 7: Client/Server Computing Business Data Communications, 5e.
Grids and Grid Technologies for Wide-Area Distributed Computing Mark Baker, Rajkumar Buyya and Domenico Laforenza.
Polaris Financial Technologies Welcomes the members of Hyderabad chapter for the 2nd event on 4 th July 14 held by PACE (The Testing Practice)
Illinois Campus Cluster Program User Forum October 24, 2012 Illini Union Room 210 2:00PM – 3:30PM.
Interoperability is Key to Accelerating SDN Adoption Neela Jacques Executive Director OpenDaylight Projectt.
Chapter 1: Hierarchical Network Design
New Direction Proposal: An OpenFabrics Framework for high-performance I/O apps OFA TAC, Key drivers: Sean Hefty, Paul Grun.
Architecture and Usages of Accelio 2014 OFA Developer Workshop Sunday, March 30 - Wednesday, April 2, 2014 Monterey CA Eyal Salomon Mellanox Technologies.
Discussing an I/O Framework SC13 - Denver. #OFADevWorkshop 2 The OpenFabrics Alliance has recently undertaken an effort to review the dominant paradigm.
Roland Dreier Technical Lead – Cisco Systems, Inc. OpenIB Maintainer Sean Hefty Software Engineer – Intel Corporation OpenIB Maintainer Yaron Haviv CTO.
Effective User Services for High Performance Computing A White Paper by the TeraGrid Science Advisory Board May 2009.
Version 4.0. Objectives Describe how networks impact our daily lives. Describe the role of data networking in the human network. Identify the key components.
Layered Protocol. 2 Types of Networks by Logical Connectivity Peer to Peer and Client-Server Peer-to-peer Networks  Every computer can communicate directly.
. Traffic Flow Management System Benefits Flexibility for Future Growth: TFMS provides a modern software architecture to meet future growth and support.
REXAPP Bilal Saqib. REXAPP  Radio EXperimentation And Prototyping Platform Based on NOC  REXAPP Compiler.
© 2012 MELLANOX TECHNOLOGIES 1 The Exascale Interconnect Technology Rich Graham – Sr. Solutions Architect.
InfiniSwitch Company Confidential. 2 InfiniSwitch Agenda InfiniBand Overview Company Overview Product Strategy Q&A.
Scalable name and address resolution infrastructure -- Ira Weiny/John Fleck #OFADevWorkshop.
Management Scalability Author: Todd Rimmer Date: April 2014.
Taking the Complexity out of Cluster Computing Vendor Update HPC User Forum Arend Dittmer Director Product Management HPC April,
Northwest Interlibrary Loan and Resource Sharing Conference Communication in the ILL Community Joe Anderson, WJ Content Manager September 20, 2007.
Remote Direct Memory Access (RDMA) over IP PFLDNet 2003, Geneva Stephen Bailey, Sandburst Corp., Allyn Romanow, Cisco Systems,
Chapter 17 Internetworking: Concepts, Architecture, and Protocols
© 2012 MELLANOX TECHNOLOGIES 1 Disruptive Technologies in HPC Interconnect HPC User Forum April 16, 2012.
Chapter 17: Client/Server Computing Business Data Communications, 4e.
RDMA Bonding Liran Liss Mellanox Technologies. Agenda Introduction Transport-level bonding RDMA bonding design Recovering from failure Implementation.
1 Public DAFS Storage for High Performance Computing using MPI-I/O: Design and Experience Arkady Kanevsky & Peter Corbett Network Appliance Vijay Velusamy.
Next Generation Operating Systems Zeljko Susnjar, Cisco CTG June 2015.
Intel Research & Development ETA: Experience with an IA processor as a Packet Processing Engine HP Labs Computer Systems Colloquium August 2003 Greg Regnier.
SDN and Openflow. Motivation Since the invention of the Internet, we find many innovative ways to use the Internet – Google, Facebook, Cloud computing,
Mellanox Connectivity Solutions for Scalable HPC Highest Performing, Most Efficient End-to-End Connectivity for Servers and Storage April 2010.
iSER update 2014 OFA Developer Workshop Eyal Salomon
OpenFabrics Interface WG A brief introduction Paul Grun – co chair OFI WG Cray, Inc.
Ex Libris Developers Network Develop. Experiment. Collaborate.
Open MPI OpenFabrics Update April 2008 Jeff Squyres.
REST By: Vishwanath Vineet.
Mellanox Connectivity Solutions for Scalable HPC Highest Performing, Most Efficient End-to-End Connectivity for Servers and Storage September 2010 Brandon.
ELECTRONIC SERVICES & TOOLS Strategic Plan
Sockets Direct Protocol for Hybrid Network Stacks: A Case Study with iWARP over 10G Ethernet P. Balaji, S. Bhagvat, R. Thakur and D. K. Panda, Mathematics.
CI R1 LCO Review Panel Preliminary Report. General Comments –Provide clear definition of the goals of the phase (e.g. inception), the scope, etc. in order.
Open Fabrics Interfaces Software Sean Hefty - Intel Corporation.
© 2008 Cisco Systems, Inc. All rights reserved.Cisco ConfidentialPresentation_ID 1 Chapter 1: Hierarchical Network Design Connecting Networks.
Stan Smith Intel SSG/DPD June, 2015 Kernel Fabric Interface Kfabric Framework.
Progress in Standardization of RDMA technology Arkady Kanevsky, Ph.D Chair of DAT Collaborative.
An open source user space fast path TCP/IP stack and more…
SC’13 BoF Discussion Sean Hefty Intel Corporation.
Introduction to Operating Systems Concepts
Multi-layer software defined networking in GÉANT
CASE STUDY 1: Linux and Android
LSO Hackathon Summary Charles Eckel, Cisco DevNet.
Grid Computing.
Chapter 18 MobileApp Design
Advancing open fabrics interfaces
RoCEE in OFED Update Liran Liss, Mellanox Technologies March 15, 2010
OFED 1.2 Status and Contents
Top #1 in China Top #3 in the world
Application taxonomy & characterization
Factors Driving Enterprise NVMeTM Growth
Presentation transcript:

The Open Fabrics Verbs Working Group Pavel Shamis and Liran Liss

Introduction Verbs is a high-performance mature and robust interface –Widely adopted –Stable –Scalable Open-source development –Peer-reviewed code patches –Peer tested Development rate and novelty constantly increasing –New applications and features –New systems and architectures The Open Fabrics Verbs Working Group (OFV-WG) –Enhancing the verbs interface the meet the needs of future systems and applications 2

OFV Mission Enhance the Verbs interface to meet the needs of future systems and applications –HPC, storage, cloud applications and more –CPUs, GPUs, and other compute elements –RDMA and other interconnect accelerated capabilities Guide the development of the Verbs eco-system –Forum to discuss approach for new features –Raise community awareness –Partner with collaborators for introducing new features 3

Middle-ware and ULPs The Verbs Eco-System Extended Verbs –Enable new features Infrastructure –Common services –Common abstractions Middleware and ULPs –Application-facing APIs 4 Extended Verbs RDMA Sockets RDMA Sockets RDMA RPC DPDK PMD Vendor Extensions Vendor Extensions Application Verbs Infrastructure RDMACM Scalable SA services

The Verbs Eco-System: Extended Verbs HPC optimizations –Accelerated Verbs Raw Ethernet support –Rx flow steering –Tx flow anti-spoofing –Stateless offloads –VLAN stripping –RSS, TSS –Tunneling (VXLAN, NVGRE) Memory –Indirect Memory Regions Signature operations Extended atomics Name space operations –QPNs, CQNs, MRs, etc. QP control –Suspend/resume Time stamp operations –Read HCA clock –Time stamp completions 5

The Verbs Eco-System: Verbs Infrastructure RDMACM improvements –E.g., APM support Bulk address resolution for HPC Kernel-managed user- space QPs –Connection management –fork() support SoftRoCE Multi-rail bonding abstraction On-Demand-Paging Container support SELinux support 6

The Verbs Eco-System: Verbs Middleware Group communication Transaction-based IO –Efficient RPC Active messages RDMA sockets User-space TCP/IP Packet processing Java bindings RDMA accelerators –E.g., HPC, Hadoop, Ceph 7

Discuss New Approaches Solicit feedback from a larger community and users –Not all of our contributors track kernel mailing lists –Storage, Big-Data, HPC, etc. Discuss concepts before writing code or detailed [RFC] patches Converge faster by interactive feedback –Faster acceptance Tackle hard-to-crack concepts by focused discussions –Weekly increments Examples: Virtualization APIs, Container support 8

Community Awareness Raise the need for important features –Introduce new use-cases –Describe feature requests –Aid in prioritizing the focus of the development community Point out pain points and urgent issues –E.g., connection rate, “debugability” 9

Collaboration Raise issues that you would like to work on See if other community members are currently tackling the same issues Establish collaboration to speed up development and increase efficiency Cross-community collaboration –Discuss challenges with broader community 10

Summary The pace of Verbs development is increasing The user base of the Verbs Eco-system is expanding The OFV WG will facilitate –Agreement on approach for new fronts Faster acceptance and convergence –Prioritize and focus OFA development efforts –Encourage collaboration –Broaden the use of the interface –Increase adoption of RDMA technology 11

Thank You