Design challenges of Highperformance. MPI over InfiniBand. Presented by Karthik

Size: px
Start display at page:

Download "Design challenges of Highperformance. MPI over InfiniBand. Presented by Karthik"

Transcription

1 Design challenges of Highperformance and Scalable MPI over InfiniBand Presented by Karthik

2 Presentation Overview In depth analysis of High-Performance and scalable MPI with Reduced Memory Usage Zero Copy protocol using Unreliable Datagram MVAPICH-Aptus : A scalable High performance Multi-Transport MPI over InfiniBand

3 High Performance and Scalable MPI with Reduced Memory usage Motivation Does aggressively reducing communication buffer memory lead to degradation of end application performance? How much memory can we expect the MPI library to consume during execution of a typical application, while still proving the best available performance?

4 High Performance and Scalable MPI with Reduced Memory usage IB provides several types of transport services Reliable Connection (RC) - Used as the primary transport for MVAPICH and other MPIs over InfiniBand. - Most feature-rich -- supports RDMA and provides reliable service. - Dedicated QP must be created for each communicating peer. Reliable Datagram (RD) - Most of the same features as RC, however, a dedicated QP is not required. - Not implemented with current hardware. Unreliable Connection (UC) - Provides RDMA capability. - No guarantees on ordering or reliability. - Dedicated QP must be created for each communicating peer. Unreliable Datagram (UD) - Connection-less. Single QP can communicate with any other peer QP. - Limited message size. - No guarantees on ordering or reliability.

5 High Performance and Scalable MPI with Reduced Memory usage Upper level software service Shared Receive Queue - This allows multiple QPs to be attached to one receive queue (even for connection oriented transport) - This approach is memory efficient

6 High Performance and Scalable MPI with Reduced Memory usage Remote Direct Memory Access (RDMA) - Application can directly access the memory of the remove process. - RDMA has very low latency.

7 High Performance and Scalable MPI with Reduced Memory usage MVAPICH Design Overview MVAPICH uses two major protocols 1. Eager Protocol - It is used to transfer small messages. - The messages are buffered inside the MPI library. - pre-allocated communication buffers are required on the sender and receiver side 2. Rendezvous Protocol - It is used to transfer large messages. - The message are sent directly to receiver s user memory.

8 High Performance and Scalable MPI with Reduced Memory usage 1. Adaptive RDMA with Send/Receive - In order to avoid a memory-scalability problem when the number of nodes increase, this channel is adaptive. - Limited buffers are allocated initially. - Once a threshold number of messages are exchanged, next messages are transferred using RDMA.

9 High Performance and Scalable MPI with Reduced Memory usage 2. Adaptive RDMA with SQR Channel - Idea is based on ARDMA-SR. Only Difference is the Shared Queue Receiver is used. - Drawback : Sender doesn t know the receiver buffer availability. - Solution : Setting a low-watermark for the SQR.

10 High Performance and Scalable MPI with Reduced Memory usage 3. Shared Receive Queue - This channel exclusively utilizes the SRQ feature. - This follows the same low-watermark technique as the ARDMA-SRQ. - Even though RDMA has low latency, they consume more memory.

11 High Performance and Scalable MPI with Reduced Memory usage NAS Benchmark

12 High Performance and Scalable MPI with Reduced Memory usage High Performance Linpack - Benchmark for solving linear equations. - It is used as the primary measure for ranking biannual Top 500 list of the world s fastest supercomputers

13 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram

14 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Motivation 1. Performance Scalability - Memory copies are detrimental to the overall performance of the application. - HCA cache can only hold a limited number of QPs 2. Resource Scalability - With a connection oriented transport the memory requirements increase linearly with the number of connected processes.

15 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Traditional Zero-Copy 1. Matched Queues Interface - The receiver deciphers the message tag from the sent message and matches it with the posted receive operations. 2. Rendezvous Protocol using RDMA - Initially a handshake protocol is used, followed by RDMA.

16 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram UD vs RC memory usage For 16k connections UD = 40 MB / process RC = 240 MB / process

17 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Challenges for true zero copy design Limited MTU Size - UD transport has a Maximum Transfer Unit(MTU) limit of 2KB. - Segmentation required. Lack of dedicated Receive Buffers - Difficult to post receive buffers for a particular peer as they are all shared. - If no buffer is posted to a QP, message sent is silently dropped. Lack of Reliability - There is no guarantee that a message will arrive at the receiver Lack of ordering - Message may not arrive in the same order they are sent. Lack of RDMA - RDMA only works for connection oriented transport.

18 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Proposed Design - Design is based on serialized communication since RDMA is not specified for UD transport - Serialized implies that the order of transfer is agreed beforehand, and only sender transmit to a QP at a single time.

19 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Solutions to design challenges 1. Efficient Segmentation - The design chooses to get completion signal only for the last packet. - The underlying reliability layer would mark packets as missing at the receiver s end and the sender is notified. 2. Zero Copy Pool - A pool of QPs are maintained. - When a message transfer is initiated, a QP is taken from the pool and the application receive buffer is posted to it. 3. Optimized Reliability and Ordering for Large Messages - One approach is the perform a checksum for the entire receive buffer. - Each operation can specify a 32-bit immediate field that will be available to the receiver as part of the completion entry.

20 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Experimental Evaluation Ping Pong Latency

21 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Uni-Directional Bandwidth

22 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Bi-Directional Bandwidth

23 MVAPICH-Aptus : Scalable High-Performance Multi-Transport MPI over InfiniBand

24 MVAPICH-Aptus : Scalable High-Performance Multi-Transport MPI over InfiniBand Motivation This paper seeks to address two mains questions - 1. What are the different protocols developed for MPI over IB? How well do they perform at scale? 2. Given this knowledge, can the MPI Library be designed to dynamically select protocols to optimized for performance and scalability?

25 MVAPICH-Aptus : Scalable High-Performance Multi-Transport MPI over InfiniBand IB provides several types of transport services Reliable Connection (RC) - Used as the primary transport for MVAPICH and other MPIs over InfiniBand. - Most feature-rich -- supports RDMA and provides reliable service. - Dedicated QP must be created for each communicating peer. Reliable Datagram (RD) - Most of the same features as RC, however, a dedicated QP is not required. - Not implemented with current hardware. Unreliable Connection (UC) - Provides RDMA capability. - No guarantees on ordering or reliability. - Dedicated QP must be created for each communicating peer. Unreliable Datagram (UD) - Connection-less. Single QP can communicate with any other peer QP. - Limited message size. - No guarantees on ordering or reliability.

26 MVAPICH-Aptus : Scalable High-Performance Multi-Transport MPI over InfiniBand Eager Protocol Channel Message Channel

27 MVAPICH-Aptus : Scalable High-Performance Multi-Transport MPI over InfiniBand Rendezvous Protocol Channel Message Channel

28 MVAPICH-Aptus : Scalable High-Performance Multi-Transport MPI over InfiniBand Performance : Eager Latency Channel Evaluation

29 MVAPICH-Aptus : Scalable High-Performance Multi-Transport MPI over InfiniBand Channel Evaluation Performance : Uni-Directional Bandwidth

30 MVAPICH-Aptus : Scalable High-Performance Multi-Transport MPI over InfiniBand Scalability Test : Memory Usage Channel Evaluation

31 MVAPICH-Aptus : Scalable High-Performance Multi-Transport MPI over InfiniBand Scalability Test : Latency Channel Evaluation

32 MVAPICH-Aptus : Scalable High-Performance Multi-Transport MPI over InfiniBand Channel Characteristics Summary

33 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Overview of Design As seen from the experimental results, using only one channel is not sufficient to achieve performance and scalability. The solution is to use a combination of message channels and transports to optimize for performance as well as scalability. Design Challenges 1. When should a channel be created? 2. When should a channel be used?

34 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Channel Allocation

35 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Channel Usage From the experimental results we can see the channels behave differently for different message size A flexible form is defined when sending a message Using this flexible framework, send rules can be changed on a per-system or job level to meet application needs without changing the code within MPI library.

36 Zero-Copy Protocol for MPI using InfiniBand Unreliable Datagram Performance Evaluation

37 QUESTIONS?

MVAPICH-Aptus: Scalable High-Performance Multi-Transport MPI over InfiniBand

MVAPICH-Aptus: Scalable High-Performance Multi-Transport MPI over InfiniBand MVAPICH-Aptus: Scalable High-Performance Multi-Transport MPI over InfiniBand Matthew Koop 1,2 Terry Jones 2 D. K. Panda 1 {koop, panda}@cse.ohio-state.edu trj@llnl.gov 1 Network-Based Computing Lab, The

More information

High Performance MPI on IBM 12x InfiniBand Architecture

High Performance MPI on IBM 12x InfiniBand Architecture High Performance MPI on IBM 12x InfiniBand Architecture Abhinav Vishnu, Brad Benton 1 and Dhabaleswar K. Panda {vishnu, panda} @ cse.ohio-state.edu {brad.benton}@us.ibm.com 1 1 Presentation Road-Map Introduction

More information

Implementing Efficient and Scalable Flow Control Schemes in MPI over InfiniBand

Implementing Efficient and Scalable Flow Control Schemes in MPI over InfiniBand Implementing Efficient and Scalable Flow Control Schemes in MPI over InfiniBand Jiuxing Liu and Dhabaleswar K. Panda Computer Science and Engineering The Ohio State University Presentation Outline Introduction

More information

Shared Receive Queue based Scalable MPI Design for InfiniBand Clusters

Shared Receive Queue based Scalable MPI Design for InfiniBand Clusters Shared Receive Queue based Scalable MPI Design for InfiniBand Clusters Sayantan Sur Lei Chai Hyun-Wook Jin Dhabaleswar K. Panda Network-Based Computing Laboratory Department of Computer Science and Engineering

More information

Improving Application Performance and Predictability using Multiple Virtual Lanes in Modern Multi-Core InfiniBand Clusters

Improving Application Performance and Predictability using Multiple Virtual Lanes in Modern Multi-Core InfiniBand Clusters Improving Application Performance and Predictability using Multiple Virtual Lanes in Modern Multi-Core InfiniBand Clusters Hari Subramoni, Ping Lai, Sayantan Sur and Dhabhaleswar. K. Panda Department of

More information

Performance Analysis and Evaluation of Mellanox ConnectX InfiniBand Architecture with Multi-Core Platforms

Performance Analysis and Evaluation of Mellanox ConnectX InfiniBand Architecture with Multi-Core Platforms Performance Analysis and Evaluation of Mellanox ConnectX InfiniBand Architecture with Multi-Core Platforms Sayantan Sur, Matt Koop, Lei Chai Dhabaleswar K. Panda Network Based Computing Lab, The Ohio State

More information

Welcome to the IBTA Fall Webinar Series

Welcome to the IBTA Fall Webinar Series Welcome to the IBTA Fall Webinar Series A four-part webinar series devoted to making I/O work for you Presented by the InfiniBand Trade Association The webinar will begin shortly. 1 September 23 October

More information

MVAPICH-Aptus: Scalable High-Performance Multi-Transport MPI over InfiniBand

MVAPICH-Aptus: Scalable High-Performance Multi-Transport MPI over InfiniBand MVAPICH-Aptus: Scalable High-Performance Multi-Transport MPI over InfiniBand Matthew J. Koop Terry Jones Dhabaleswar K. Panda Network-Based Computing Laboratory The Ohio State University Columbus, OH 4321

More information

Unifying UPC and MPI Runtimes: Experience with MVAPICH

Unifying UPC and MPI Runtimes: Experience with MVAPICH Unifying UPC and MPI Runtimes: Experience with MVAPICH Jithin Jose Miao Luo Sayantan Sur D. K. Panda Network-Based Computing Laboratory Department of Computer Science and Engineering The Ohio State University,

More information

Can Memory-Less Network Adapters Benefit Next-Generation InfiniBand Systems?

Can Memory-Less Network Adapters Benefit Next-Generation InfiniBand Systems? Can Memory-Less Network Adapters Benefit Next-Generation InfiniBand Systems? Sayantan Sur, Abhinav Vishnu, Hyun-Wook Jin, Wei Huang and D. K. Panda {surs, vishnu, jinhy, huanwei, panda}@cse.ohio-state.edu

More information

Screencast: OMPI OpenFabrics Protocols (v1.2 series)

Screencast: OMPI OpenFabrics Protocols (v1.2 series) Screencast: OMPI OpenFabrics Protocols (v1.2 series) Jeff Squyres May 2008 May 2008 Screencast: OMPI OpenFabrics Protocols (v1.2 series) 1 Short Messages For short messages memcpy() into / out of pre-registered

More information

Reducing Network Contention with Mixed Workloads on Modern Multicore Clusters

Reducing Network Contention with Mixed Workloads on Modern Multicore Clusters Reducing Network Contention with Mixed Workloads on Modern Multicore Clusters Matthew Koop 1 Miao Luo D. K. Panda matthew.koop@nasa.gov {luom, panda}@cse.ohio-state.edu 1 NASA Center for Computational

More information

Low latency, high bandwidth communication. Infiniband and RDMA programming. Bandwidth vs latency. Knut Omang Ifi/Oracle 2 Nov, 2015

Low latency, high bandwidth communication. Infiniband and RDMA programming. Bandwidth vs latency. Knut Omang Ifi/Oracle 2 Nov, 2015 Low latency, high bandwidth communication. Infiniband and RDMA programming Knut Omang Ifi/Oracle 2 Nov, 2015 1 Bandwidth vs latency There is an old network saying: Bandwidth problems can be cured with

More information

Performance Evaluation of InfiniBand with PCI Express

Performance Evaluation of InfiniBand with PCI Express Performance Evaluation of InfiniBand with PCI Express Jiuxing Liu Amith Mamidala Abhinav Vishnu Dhabaleswar K Panda Department of Computer and Science and Engineering The Ohio State University Columbus,

More information

Op#miza#on and Tuning of Hybrid, Mul#rail, 3D Torus Support and QoS in MVAPICH2

Op#miza#on and Tuning of Hybrid, Mul#rail, 3D Torus Support and QoS in MVAPICH2 Op#miza#on and Tuning of Hybrid, Mul#rail, 3D Torus Support and QoS in MVAPICH2 MVAPICH2 User Group (MUG) Mee#ng by Hari Subramoni The Ohio State University E- mail: subramon@cse.ohio- state.edu h

More information

10-Gigabit iwarp Ethernet: Comparative Performance Analysis with InfiniBand and Myrinet-10G

10-Gigabit iwarp Ethernet: Comparative Performance Analysis with InfiniBand and Myrinet-10G 10-Gigabit iwarp Ethernet: Comparative Performance Analysis with InfiniBand and Myrinet-10G Mohammad J. Rashti and Ahmad Afsahi Queen s University Kingston, ON, Canada 2007 Workshop on Communication Architectures

More information

The Exascale Architecture

The Exascale Architecture The Exascale Architecture Richard Graham HPC Advisory Council China 2013 Overview Programming-model challenges for Exascale Challenges for scaling MPI to Exascale InfiniBand enhancements Dynamically Connected

More information

What communication library can do with a little hint from programmers? Takeshi Nanri (Kyushu Univ. and JST CREST, Japan)

What communication library can do with a little hint from programmers? Takeshi Nanri (Kyushu Univ. and JST CREST, Japan) 1 What communication library can do with a little hint from programmers? Takeshi Nanri (Kyushu Univ. and JST CREST, Japan) 2 Background Various tuning opportunities in communication libraries:. Protocol?

More information

Designing High-Performance and Resilient Message Passing on InfiniBand

Designing High-Performance and Resilient Message Passing on InfiniBand Designing High-Performance and Resilient Message Passing on InfiniBand Matthew J. Koop 1 Pavel Shamis 2 Ishai Rabinovitz 2 Dhabaleswar K. (DK) Panda 3 1 High Performance Technologies, Inc (HPTi), mkoop@hpti.com

More information

Intra-MIC MPI Communication using MVAPICH2: Early Experience

Intra-MIC MPI Communication using MVAPICH2: Early Experience Intra-MIC MPI Communication using MVAPICH: Early Experience Sreeram Potluri, Karen Tomko, Devendar Bureddy, and Dhabaleswar K. Panda Department of Computer Science and Engineering Ohio State University

More information

RDMA Read Based Rendezvous Protocol for MPI over InfiniBand: Design Alternatives and Benefits

RDMA Read Based Rendezvous Protocol for MPI over InfiniBand: Design Alternatives and Benefits RDMA Read Based Rendezvous Protocol for MPI over InfiniBand: Design Alternatives and Benefits Sayantan Sur Hyun-Wook Jin Lei Chai D. K. Panda Network Based Computing Lab, The Ohio State University Presentation

More information

Scalable High Performance Message Passing over InfiniBand for Open MPI

Scalable High Performance Message Passing over InfiniBand for Open MPI Scalable High Performance Message Passing over InfiniBand for Open MPI AndrewFriedley 123 TorstenHoefler 1 MatthewL.Leininger 23 AndrewLumsdaine 1 1 OpenSystemsLaboratory,IndianaUniversity,BloomingtonIN47405,USA

More information

Evaluating the Impact of RDMA on Storage I/O over InfiniBand

Evaluating the Impact of RDMA on Storage I/O over InfiniBand Evaluating the Impact of RDMA on Storage I/O over InfiniBand J Liu, DK Panda and M Banikazemi Computer and Information Science IBM T J Watson Research Center The Ohio State University Presentation Outline

More information

Performance Evaluation of InfiniBand with PCI Express

Performance Evaluation of InfiniBand with PCI Express Performance Evaluation of InfiniBand with PCI Express Jiuxing Liu Server Technology Group IBM T. J. Watson Research Center Yorktown Heights, NY 1598 jl@us.ibm.com Amith Mamidala, Abhinav Vishnu, and Dhabaleswar

More information

Designing High Performance Communication Middleware with Emerging Multi-core Architectures

Designing High Performance Communication Middleware with Emerging Multi-core Architectures Designing High Performance Communication Middleware with Emerging Multi-core Architectures Dhabaleswar K. (DK) Panda Department of Computer Science and Engg. The Ohio State University E-mail: panda@cse.ohio-state.edu

More information

OpenFabrics Interface WG A brief introduction. Paul Grun co chair OFI WG Cray, Inc.

OpenFabrics Interface WG A brief introduction. Paul Grun co chair OFI WG Cray, Inc. OpenFabrics Interface WG A brief introduction Paul Grun co chair OFI WG Cray, Inc. OFI WG a brief overview and status report 1. Keep everybody on the same page, and 2. An example of a possible model for

More information

Unified Runtime for PGAS and MPI over OFED

Unified Runtime for PGAS and MPI over OFED Unified Runtime for PGAS and MPI over OFED D. K. Panda and Sayantan Sur Network-Based Computing Laboratory Department of Computer Science and Engineering The Ohio State University, USA Outline Introduction

More information

Application-Transparent Checkpoint/Restart for MPI Programs over InfiniBand

Application-Transparent Checkpoint/Restart for MPI Programs over InfiniBand Application-Transparent Checkpoint/Restart for MPI Programs over InfiniBand Qi Gao, Weikuan Yu, Wei Huang, Dhabaleswar K. Panda Network-Based Computing Laboratory Department of Computer Science & Engineering

More information

Paving the Road to Exascale

Paving the Road to Exascale Paving the Road to Exascale Gilad Shainer August 2015, MVAPICH User Group (MUG) Meeting The Ever Growing Demand for Performance Performance Terascale Petascale Exascale 1 st Roadrunner 2000 2005 2010 2015

More information

Design and Evaluation of Efficient Collective Communications on Modern Interconnects and Multi-core Clusters

Design and Evaluation of Efficient Collective Communications on Modern Interconnects and Multi-core Clusters Design and Evaluation of Efficient Collective Communications on Modern Interconnects and Multi-core Clusters by Ying Qian A thesis submitted to the Department of Electrical and Computer Engineering in

More information

Implementing Efficient and Scalable Flow Control Schemes in MPI over InfiniBand

Implementing Efficient and Scalable Flow Control Schemes in MPI over InfiniBand Implementing Efficient and Scalable Flow Control Schemes in MPI over InfiniBand Jiuxing Liu Dhabaleswar K. Panda Computer and Information Science The Ohio State University Columbus, OH 43210 liuj, panda

More information

Optimizing non-blocking Collective Operations for InfiniBand

Optimizing non-blocking Collective Operations for InfiniBand Optimizing non-blocking Collective Operations for InfiniBand Open Systems Lab Indiana University Bloomington, USA IPDPS 08 - CAC 08 Workshop Miami, FL, USA April, 14th 2008 Introduction Non-blocking collective

More information

Impact of HPC Cloud Networking Technologies on Accelerating Hadoop RPC and HBase

Impact of HPC Cloud Networking Technologies on Accelerating Hadoop RPC and HBase 2 IEEE 8th International Conference on Cloud Computing Technology and Science Impact of HPC Cloud Networking Technologies on Accelerating Hadoop RPC and HBase Xiaoyi Lu, Dipti Shankar, Shashank Gugnani,

More information

RDMA programming concepts

RDMA programming concepts RDMA programming concepts Robert D. Russell InterOperability Laboratory & Computer Science Department University of New Hampshire Durham, New Hampshire 03824, USA 2013 Open Fabrics Alliance,

More information

Memory Management Strategies for Data Serving with RDMA

Memory Management Strategies for Data Serving with RDMA Memory Management Strategies for Data Serving with RDMA Dennis Dalessandro and Pete Wyckoff (presenting) Ohio Supercomputer Center {dennis,pw}@osc.edu HotI'07 23 August 2007 Motivation Increasing demands

More information

UDP, TCP, IP multicast

UDP, TCP, IP multicast UDP, TCP, IP multicast Dan Williams In this lecture UDP (user datagram protocol) Unreliable, packet-based TCP (transmission control protocol) Reliable, connection oriented, stream-based IP multicast Process-to-Process

More information

RDMA enabled NIC (RNIC) Verbs Overview. Renato Recio

RDMA enabled NIC (RNIC) Verbs Overview. Renato Recio RDMA enabled NIC () Verbs Overview Renato Recio Verbs!The RDMA Protocol Verbs Specification describes the behavior of hardware, firmware, and software as viewed by the host, "not the host software itself,

More information

In-Network Computing. Sebastian Kalcher, Senior System Engineer HPC. May 2017

In-Network Computing. Sebastian Kalcher, Senior System Engineer HPC. May 2017 In-Network Computing Sebastian Kalcher, Senior System Engineer HPC May 2017 Exponential Data Growth The Need for Intelligent and Faster Interconnect CPU-Centric (Onload) Data-Centric (Offload) Must Wait

More information

Adaptive Connection Management for Scalable MPI over InfiniBand

Adaptive Connection Management for Scalable MPI over InfiniBand Adaptive Connection Management for Scalable MPI over InfiniBand Weikuan Yu Qi Gao Dhabaleswar K. Panda Network-Based Computing Lab Dept. of Computer Sci. & Engineering The Ohio State University {yuw,gaoq,panda}@cse.ohio-state.edu

More information

MPI Alltoall Personalized Exchange on GPGPU Clusters: Design Alternatives and Benefits

MPI Alltoall Personalized Exchange on GPGPU Clusters: Design Alternatives and Benefits MPI Alltoall Personalized Exchange on GPGPU Clusters: Design Alternatives and Benefits Ashish Kumar Singh, Sreeram Potluri, Hao Wang, Krishna Kandalla, Sayantan Sur, and Dhabaleswar K. Panda Network-Based

More information

Request for Comments: 4755 Category: Standards Track December 2006

Request for Comments: 4755 Category: Standards Track December 2006 Network Working Group V. Kashyap Request for Comments: 4755 IBM Category: Standards Track December 2006 Status of This Memo IP over InfiniBand: Connected Mode This document specifies an Internet standards

More information

Hybrid MPI - A Case Study on the Xeon Phi Platform

Hybrid MPI - A Case Study on the Xeon Phi Platform Hybrid MPI - A Case Study on the Xeon Phi Platform Udayanga Wickramasinghe Center for Research on Extreme Scale Technologies (CREST) Indiana University Greg Bronevetsky Lawrence Livermore National Laboratory

More information

UNDERSTANDING THE IMPACT OF MULTI-CORE ARCHITECTURE IN CLUSTER COMPUTING: A CASE STUDY WITH INTEL DUAL-CORE SYSTEM

UNDERSTANDING THE IMPACT OF MULTI-CORE ARCHITECTURE IN CLUSTER COMPUTING: A CASE STUDY WITH INTEL DUAL-CORE SYSTEM UNDERSTANDING THE IMPACT OF MULTI-CORE ARCHITECTURE IN CLUSTER COMPUTING: A CASE STUDY WITH INTEL DUAL-CORE SYSTEM Sweety Sen, Sonali Samanta B.Tech, Information Technology, Dronacharya College of Engineering,

More information

Hot-Spot Avoidance With Multi-Pathing Over InfiniBand: An MPI Perspective

Hot-Spot Avoidance With Multi-Pathing Over InfiniBand: An MPI Perspective Hot-Spot Avoidance With Multi-Pathing Over InfiniBand: An MPI Perspective A. Vishnu M. Koop A. Moody A. R. Mamidala S. Narravula D. K. Panda Computer Science and Engineering The Ohio State University {vishnu,

More information

Future Routing Schemes in Petascale clusters

Future Routing Schemes in Petascale clusters Future Routing Schemes in Petascale clusters Gilad Shainer, Mellanox, USA Ola Torudbakken, Sun Microsystems, Norway Richard Graham, Oak Ridge National Laboratory, USA Birds of a Feather Presentation Abstract

More information

INAM 2 : InfiniBand Network Analysis and Monitoring with MPI

INAM 2 : InfiniBand Network Analysis and Monitoring with MPI INAM 2 : InfiniBand Network Analysis and Monitoring with MPI H. Subramoni, A. A. Mathews, M. Arnold, J. Perkins, X. Lu, K. Hamidouche, and D. K. Panda Department of Computer Science and Engineering The

More information

IBRMP: a Reliable Multicast Protocol for InfiniBand

IBRMP: a Reliable Multicast Protocol for InfiniBand 2014 IEEE 22nd Annual Symposium on High-Performance Interconnects IBRMP: a Reliable Multicast Protocol for InfiniBand Qian Liu, Robert D. Russell Department of Computer Science University of New Hampshire

More information

MELLANOX EDR UPDATE & GPUDIRECT MELLANOX SR. SE 정연구

MELLANOX EDR UPDATE & GPUDIRECT MELLANOX SR. SE 정연구 MELLANOX EDR UPDATE & GPUDIRECT MELLANOX SR. SE 정연구 Leading Supplier of End-to-End Interconnect Solutions Analyze Enabling the Use of Data Store ICs Comprehensive End-to-End InfiniBand and Ethernet Portfolio

More information

Assessing the Ability of Computation/Communication Overlap and Communication Progress in Modern Interconnects

Assessing the Ability of Computation/Communication Overlap and Communication Progress in Modern Interconnects Assessing the Ability of Computation/Communication Overlap and Communication Progress in Modern Interconnects Mohammad J. Rashti Ahmad Afsahi Department of Electrical and Computer Engineering Queen s University,

More information

2008 International ANSYS Conference

2008 International ANSYS Conference 2008 International ANSYS Conference Maximizing Productivity With InfiniBand-Based Clusters Gilad Shainer Director of Technical Marketing Mellanox Technologies 2008 ANSYS, Inc. All rights reserved. 1 ANSYS,

More information

LiMIC: Support for High-Performance MPI Intra-Node Communication on Linux Cluster

LiMIC: Support for High-Performance MPI Intra-Node Communication on Linux Cluster LiMIC: Support for High-Performance MPI Intra-Node Communication on Linux Cluster H. W. Jin, S. Sur, L. Chai, and D. K. Panda Network-Based Computing Laboratory Department of Computer Science and Engineering

More information

FaSST: Fast, Scalable, and Simple Distributed Transactions with Two-Sided (RDMA) Datagram RPCs

FaSST: Fast, Scalable, and Simple Distributed Transactions with Two-Sided (RDMA) Datagram RPCs FaSST: Fast, Scalable, and Simple Distributed Transactions with Two-Sided (RDMA) Datagram RPCs Anuj Kalia (CMU), Michael Kaminsky (Intel Labs), David Andersen (CMU) RDMA RDMA is a network feature that

More information

Understanding MPI on Cray XC30

Understanding MPI on Cray XC30 Understanding MPI on Cray XC30 MPICH3 and Cray MPT Cray MPI uses MPICH3 distribution from Argonne Provides a good, robust and feature rich MPI Cray provides enhancements on top of this: low level communication

More information

CSE 461 Module 10. Introduction to the Transport Layer

CSE 461 Module 10. Introduction to the Transport Layer CSE 461 Module 10 Introduction to the Transport Layer Last Time We finished up the Network layer Internetworks (IP) Routing (DV/RIP, LS/OSPF, BGP) It was all about routing: how to provide end-to-end delivery

More information

Sayantan Sur, Intel. Presenting work done by Arun Ilango, Dmitry Gladkov, Dmitry Durnov and Sean Hefty and others in the OFIWG community

Sayantan Sur, Intel. Presenting work done by Arun Ilango, Dmitry Gladkov, Dmitry Durnov and Sean Hefty and others in the OFIWG community Sayantan Sur, Intel Presenting work done by Arun Ilango, Dmitry Gladkov, Dmitry Durnov and Sean Hefty and others in the OFIWG community 6 th Annual MVAPICH User Group (MUG) 2018 Legal Disclaimer & Optimization

More information

Unified Communication X (UCX)

Unified Communication X (UCX) Unified Communication X (UCX) Pavel Shamis / Pasha ARM Research SC 18 UCF Consortium Mission: Collaboration between industry, laboratories, and academia to create production grade communication frameworks

More information

Lecture 2 Communication services The Trasport Layer. Antonio Cianfrani DIET Department Networking Group netlab.uniroma1.it

Lecture 2 Communication services The Trasport Layer. Antonio Cianfrani DIET Department Networking Group netlab.uniroma1.it Lecture 2 Communication services The Trasport Layer Antonio Cianfrani DIET Department Networking Group netlab.uniroma1.it The structure edge: applications and hosts core: routers of s access s, media:

More information

GPUnet: Networking Abstractions for GPU Programs. Author: Andrzej Jackowski

GPUnet: Networking Abstractions for GPU Programs. Author: Andrzej Jackowski Author: Andrzej Jackowski 1 Author: Andrzej Jackowski 2 GPU programming problem 3 GPU distributed application flow 1. recv req Network 4. send repl 2. exec on GPU CPU & Memory 3. get results GPU & Memory

More information

IMPROVING MESSAGE-PASSING PERFORMANCE AND SCALABILITY IN HIGH-PERFORMANCE CLUSTERS

IMPROVING MESSAGE-PASSING PERFORMANCE AND SCALABILITY IN HIGH-PERFORMANCE CLUSTERS IMPROVING MESSAGE-PASSING PERFORMANCE AND SCALABILITY IN HIGH-PERFORMANCE CLUSTERS by Mohammad Javad Rashti A thesis submitted to the Department of Electrical and Computer Engineering In conformity with

More information

High Performance MPI-2 One-Sided Communication over InfiniBand

High Performance MPI-2 One-Sided Communication over InfiniBand High Performance MPI-2 One-Sided Communication over InfiniBand Weihang Jiang Jiuxing Liu Hyun-Wook Jin Dhabaleswar K. Panda William Gropp Rajeev Thakur Computer and Information Science The Ohio State University

More information

Interconnect Your Future

Interconnect Your Future Interconnect Your Future Smart Interconnect for Next Generation HPC Platforms Gilad Shainer, August 2016, 4th Annual MVAPICH User Group (MUG) Meeting Mellanox Connects the World s Fastest Supercomputer

More information

Birds of a Feather Presentation

Birds of a Feather Presentation Mellanox InfiniBand QDR 4Gb/s The Fabric of Choice for High Performance Computing Gilad Shainer, shainer@mellanox.com June 28 Birds of a Feather Presentation InfiniBand Technology Leadership Industry Standard

More information

High Performance MPI-2 One-Sided Communication over InfiniBand

High Performance MPI-2 One-Sided Communication over InfiniBand High Performance MPI-2 One-Sided Communication over InfiniBand Weihang Jiang Jiuxing Liu Hyun-Wook Jin Dhabaleswar K. Panda William Gropp Rajeev Thakur Computer and Information Science The Ohio State University

More information

Study. Dhabaleswar. K. Panda. The Ohio State University HPIDC '09

Study. Dhabaleswar. K. Panda. The Ohio State University HPIDC '09 RDMA over Ethernet - A Preliminary Study Hari Subramoni, Miao Luo, Ping Lai and Dhabaleswar. K. Panda Computer Science & Engineering Department The Ohio State University Introduction Problem Statement

More information

Optimizing MPI Communication on Multi-GPU Systems using CUDA Inter-Process Communication

Optimizing MPI Communication on Multi-GPU Systems using CUDA Inter-Process Communication Optimizing MPI Communication on Multi-GPU Systems using CUDA Inter-Process Communication Sreeram Potluri* Hao Wang* Devendar Bureddy* Ashish Kumar Singh* Carlos Rosales + Dhabaleswar K. Panda* *Network-Based

More information

Designing Power-Aware Collective Communication Algorithms for InfiniBand Clusters

Designing Power-Aware Collective Communication Algorithms for InfiniBand Clusters Designing Power-Aware Collective Communication Algorithms for InfiniBand Clusters Krishna Kandalla, Emilio P. Mancini, Sayantan Sur, and Dhabaleswar. K. Panda Department of Computer Science & Engineering,

More information

Performance of RDMA-capable Storage Protocols on Wide-Area Network

Performance of RDMA-capable Storage Protocols on Wide-Area Network Performance of RDMA-capable Storage Protocols on Wide-Area Network Weikuan Yu Nageswara S.V. Rao Pete Wyckoff Jeffrey S. Vetter Oak Ridge National Laboratory Ohio Supercomputer Center Computer Science

More information

Designing Shared Address Space MPI libraries in the Many-core Era

Designing Shared Address Space MPI libraries in the Many-core Era Designing Shared Address Space MPI libraries in the Many-core Era Jahanzeb Hashmi hashmi.29@osu.edu (NBCL) The Ohio State University Outline Introduction and Motivation Background Shared-memory Communication

More information

Network bandwidth is a performance bottleneck for cluster computing. Especially for clusters built with SMP machines.

Network bandwidth is a performance bottleneck for cluster computing. Especially for clusters built with SMP machines. Mingzhe Li Motivation Network bandwidth is a performance bottleneck for cluster computing. Especially for clusters built with SMP machines. Multirail network is an efficient way to alleviate this problem

More information

Infiniband Scalability in Open MPI

Infiniband Scalability in Open MPI Infiniband Scalability in Open MPI G. M. Shipman 1,2, T. S. Woodall 1, R. L. Graham 1, A. B. Maccabe 2 1 Advanced Computing Laboratory Los Alamos National Laboratory 2 Scalable Systems Laboratory Computer

More information

CERN openlab Summer 2006: Networking Overview

CERN openlab Summer 2006: Networking Overview CERN openlab Summer 2006: Networking Overview Martin Swany, Ph.D. Assistant Professor, Computer and Information Sciences, U. Delaware, USA Visiting Helsinki Institute of Physics (HIP) at CERN swany@cis.udel.edu,

More information

Designing and Enhancing the Sockets Direct Protocol (SDP) over iwarp and InfiniBand

Designing and Enhancing the Sockets Direct Protocol (SDP) over iwarp and InfiniBand Designing and Enhancing the Sockets Direct Protocol (SDP) over iwarp and InfiniBand A Thesis Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School

More information

Efficient and Truly Passive MPI-3 RMA Synchronization Using InfiniBand Atomics

Efficient and Truly Passive MPI-3 RMA Synchronization Using InfiniBand Atomics 1 Efficient and Truly Passive MPI-3 RMA Synchronization Using InfiniBand Atomics Mingzhe Li Sreeram Potluri Khaled Hamidouche Jithin Jose Dhabaleswar K. Panda Network-Based Computing Laboratory Department

More information

Memory Scalability Evaluation of the Next-Generation Intel Bensley Platform with InfiniBand

Memory Scalability Evaluation of the Next-Generation Intel Bensley Platform with InfiniBand Memory Scalability Evaluation of the Next-Generation Intel Bensley Platform with InfiniBand Matthew Koop, Wei Huang, Ahbinav Vishnu, Dhabaleswar K. Panda Network-Based Computing Laboratory Department of

More information

LiMIC: Support for High-Performance MPI Intra-Node Communication on Linux Cluster

LiMIC: Support for High-Performance MPI Intra-Node Communication on Linux Cluster : Support for High-Performance MPI Intra-Node Communication on Linux Cluster Hyun-Wook Jin Sayantan Sur Lei Chai Dhabaleswar K. Panda Department of Computer Science and Engineering The Ohio State University

More information

A New Design of RDMA-based Small Message Channels for InfiniBand Clusters

A New Design of RDMA-based Small Message Channels for InfiniBand Clusters A New Design of RDMA-based Small Message Channels for InfiniBand Clusters Matthew Small Xin Yuan Department of Computer Science, Florida State University, Tallahassee, FL 32306 {small,xyuan}@cs.fsu.edu

More information

Noise Injection Techniques to Expose Subtle and Unintended Message Races

Noise Injection Techniques to Expose Subtle and Unintended Message Races Noise Injection Techniques to Expose Subtle and Unintended Message Races PPoPP2017 February 6th, 2017 Kento Sato, Dong H. Ahn, Ignacio Laguna, Gregory L. Lee, Martin Schulz and Christopher M. Chambreau

More information

Accelerating MPI Message Matching and Reduction Collectives For Multi-/Many-core Architectures Mohammadreza Bayatpour, Hari Subramoni, D. K.

Accelerating MPI Message Matching and Reduction Collectives For Multi-/Many-core Architectures Mohammadreza Bayatpour, Hari Subramoni, D. K. Accelerating MPI Message Matching and Reduction Collectives For Multi-/Many-core Architectures Mohammadreza Bayatpour, Hari Subramoni, D. K. Panda Department of Computer Science and Engineering The Ohio

More information

The Role of InfiniBand Technologies in High Performance Computing. 1 Managed by UT-Battelle for the Department of Energy

The Role of InfiniBand Technologies in High Performance Computing. 1 Managed by UT-Battelle for the Department of Energy The Role of InfiniBand Technologies in High Performance Computing 1 Managed by UT-Battelle Contributors Gil Bloch Noam Bloch Hillel Chapman Manjunath Gorentla- Venkata Richard Graham Michael Kagan Vasily

More information

10-Gigabit iwarp Ethernet: Comparative Performance Analysis with InfiniBand and Myrinet-10G

10-Gigabit iwarp Ethernet: Comparative Performance Analysis with InfiniBand and Myrinet-10G -Gigabit iwarp Ethernet: Comparative Performance Analysis with InfiniBand and Myrinet-G Mohammad J. Rashti Ahmad Afsahi Department of Electrical and Computer Engineering Queen s University, Kingston, ON,

More information

Runtime Algorithm Selection of Collective Communication with RMA-based Monitoring Mechanism

Runtime Algorithm Selection of Collective Communication with RMA-based Monitoring Mechanism 1 Runtime Algorithm Selection of Collective Communication with RMA-based Monitoring Mechanism Takeshi Nanri (Kyushu Univ. and JST CREST, Japan) 16 Aug, 2016 4th Annual MVAPICH Users Group Meeting 2 Background

More information

Transport Layer Protocols TCP

Transport Layer Protocols TCP Transport Layer Protocols TCP Gail Hopkins Introduction Features of TCP Packet loss and retransmission Adaptive retransmission Flow control Three way handshake Congestion control 1 Common Networking Issues

More information

Question 1 (6 points) Compare circuit-switching and packet-switching networks based on the following criteria:

Question 1 (6 points) Compare circuit-switching and packet-switching networks based on the following criteria: Question 1 (6 points) Compare circuit-switching and packet-switching networks based on the following criteria: (a) Reserving network resources ahead of data being sent: (2pts) In circuit-switching networks,

More information

Advancing RDMA. A proposal for RDMA on Enhanced Ethernet. Paul Grun SystemFabricWorks

Advancing RDMA. A proposal for RDMA on Enhanced Ethernet.  Paul Grun SystemFabricWorks Advancing RDMA A proposal for RDMA on Enhanced Ethernet Paul Grun SystemFabricWorks pgrun@systemfabricworks.com Objective: Accelerate the adoption of RDMA technology Why bother? I mean, who cares about

More information

Micro-Benchmark Level Performance Comparison of High-Speed Cluster Interconnects

Micro-Benchmark Level Performance Comparison of High-Speed Cluster Interconnects Micro-Benchmark Level Performance Comparison of High-Speed Cluster Interconnects Jiuxing Liu Balasubramanian Chandrasekaran Weikuan Yu Jiesheng Wu Darius Buntinas Sushmitha Kini Peter Wyckoff Dhabaleswar

More information

UNIT IV -- TRANSPORT LAYER

UNIT IV -- TRANSPORT LAYER UNIT IV -- TRANSPORT LAYER TABLE OF CONTENTS 4.1. Transport layer. 02 4.2. Reliable delivery service. 03 4.3. Congestion control. 05 4.4. Connection establishment.. 07 4.5. Flow control 09 4.6. Transmission

More information

HPC Customer Requirements for OpenFabrics Software

HPC Customer Requirements for OpenFabrics Software HPC Customer Requirements for OpenFabrics Software Matt Leininger, Ph.D. Sandia National Laboratories Scalable Computing R&D Livermore, CA 16 November 2006 I'll focus on software requirements (well maybe)

More information

Designing Multi-Leader-Based Allgather Algorithms for Multi-Core Clusters *

Designing Multi-Leader-Based Allgather Algorithms for Multi-Core Clusters * Designing Multi-Leader-Based Allgather Algorithms for Multi-Core Clusters * Krishna Kandalla, Hari Subramoni, Gopal Santhanaraman, Matthew Koop and Dhabaleswar K. Panda Department of Computer Science and

More information

Accelerating MPI Message Matching and Reduction Collectives For Multi-/Many-core Architectures

Accelerating MPI Message Matching and Reduction Collectives For Multi-/Many-core Architectures Accelerating MPI Message Matching and Reduction Collectives For Multi-/Many-core Architectures M. Bayatpour, S. Chakraborty, H. Subramoni, X. Lu, and D. K. Panda Department of Computer Science and Engineering

More information

Design Alternatives for Implementing Fence Synchronization in MPI-2 One-Sided Communication for InfiniBand Clusters

Design Alternatives for Implementing Fence Synchronization in MPI-2 One-Sided Communication for InfiniBand Clusters Design Alternatives for Implementing Fence Synchronization in MPI-2 One-Sided Communication for InfiniBand Clusters G.Santhanaraman, T. Gangadharappa, S.Narravula, A.Mamidala and D.K.Panda Presented by:

More information

Experimental Analysis of InfiniBand Transport Services on WAN

Experimental Analysis of InfiniBand Transport Services on WAN International Conference on Networking, Architecture, and Storage Experimental Analysis of InfiniBand Transport Services on WAN Abstract InfiniBand Architecture (IBA) has emerged as a standard system-area

More information

Enhancing Checkpoint Performance with Staging IO & SSD

Enhancing Checkpoint Performance with Staging IO & SSD Enhancing Checkpoint Performance with Staging IO & SSD Xiangyong Ouyang Sonya Marcarelli Dhabaleswar K. Panda Department of Computer Science & Engineering The Ohio State University Outline Motivation and

More information

Progress Report on Transparent Checkpointing for Supercomputing

Progress Report on Transparent Checkpointing for Supercomputing Progress Report on Transparent Checkpointing for Supercomputing Jiajun Cao, Rohan Garg College of Computer and Information Science, Northeastern University {jiajun,rohgarg}@ccs.neu.edu August 21, 2015

More information

Performance of HPC Middleware over InfiniBand WAN

Performance of HPC Middleware over InfiniBand WAN Performance of HPC Middleware over InfiniBand WAN SUNDEEP NARRAVULA, HARI SUBRAMONI, PING LAI, BHARGAVI RAJARAMAN, RANJIT NORONHA, DHABALESWAR K. PANDA Technical Report OSU-CISRC-12/7-TR77. Performance

More information

CRFS: A Lightweight User-Level Filesystem for Generic Checkpoint/Restart

CRFS: A Lightweight User-Level Filesystem for Generic Checkpoint/Restart CRFS: A Lightweight User-Level Filesystem for Generic Checkpoint/Restart Xiangyong Ouyang, Raghunath Rajachandrasekar, Xavier Besseron, Hao Wang, Jian Huang, Dhabaleswar K. Panda Department of Computer

More information

Performance Optimizations via Connect-IB and Dynamically Connected Transport Service for Maximum Performance on LS-DYNA

Performance Optimizations via Connect-IB and Dynamically Connected Transport Service for Maximum Performance on LS-DYNA Performance Optimizations via Connect-IB and Dynamically Connected Transport Service for Maximum Performance on LS-DYNA Pak Lui, Gilad Shainer, Brian Klaff Mellanox Technologies Abstract From concept to

More information

Application Acceleration Beyond Flash Storage

Application Acceleration Beyond Flash Storage Application Acceleration Beyond Flash Storage Session 303C Mellanox Technologies Flash Memory Summit July 2014 Accelerating Applications, Step-by-Step First Steps Make compute fast Moore s Law Make storage

More information

SCTP versus TCP for MPI

SCTP versus TCP for MPI SCTP versus TCP for MPI Humaira Kamal Department of Computer Science University of British Columbia Vancouver, BC kamal@cs.ubc.ca Brad Penoff Department of Computer Science University of British Columbia

More information

MM5 Modeling System Performance Research and Profiling. March 2009

MM5 Modeling System Performance Research and Profiling. March 2009 MM5 Modeling System Performance Research and Profiling March 2009 Note The following research was performed under the HPC Advisory Council activities AMD, Dell, Mellanox HPC Advisory Council Cluster Center

More information

Leveraging Burst Buffer Coordination to Prevent I/O Interference

Leveraging Burst Buffer Coordination to Prevent I/O Interference Leveraging Burst Buffer Coordination to Prevent I/O Interference Anthony Kougkas akougkas@hawk.iit.edu Matthieu Dorier, Rob Latham, Rob Ross, Xian-He Sun Wednesday, October 26th Baltimore, USA Outline

More information