Adaptive Power Profiling for Many-Core HPC Architectures

Size: px
Start display at page:

Download "Adaptive Power Profiling for Many-Core HPC Architectures"

Transcription

1 Adaptive Power Profiling for Many-Core HPC Architectures J A I M I E K E L L E Y, C H R I S TO P H E R S T E WA R T T H E O H I O S TAT E U N I V E R S I T Y D E V E S H T I WA R I, S A U R A B H G U P TA O A K R I D G E N AT I O N A L L A B O R AT O RY S U M M A R I Z E D B Y: D A R W I N M A C H F O R : CS 788 A U TO N O M I C C O M P U T I N G FA L L G E O R G E M A S O N U N I V E R S I T Y

2 Overview Background & Problem Statement Experimental Design Observations Observations on Power Consumption Predicting Peak Power Using Reference Workloads Analyzing Power Consumption Profile of Scientific Applications Adaptive Power Profiling 2

3 Background Amount of cores available for HPC (high performance computing) continues to increase Aggregation of computing power, supercomputers such as Jaguar, Titan, etc Many workloads don t use every core effectively PARSEC benchmark gets 90% of its speed from just 35 out of 442 cores NAS workload on Intel Phi gets 85% of its speed from 32 of the 61 cores Core scaling: restrict workload to subset of cores Side effect: increases peak power use for workload cores Increasing demand to set power caps 3

4 Problem Current HPC schedulers use static workload profiles to allocate resources & adjust provisioning while workload runs But resource contention is dynamic & lots of things can affect it Actual power usage is complex and varies Static models only determine what s possible, not what actually happens How can we accurately predict peak power dynamically with as little time as possible? Peak power because that s going to determine minimum power cap 4

5 Experimental Design Power Measurement Architectures Platforms Workloads 5

6 Experimental Design: Power Measurement Use Intel Running Average Power Limit (RAPL) Stores measurements per CPU socket in Machine State Registers (MSRs) Measure energy and convert to power by associating with timestamp Power = Work / time Measure every 100 ms For Xeon Phi, use micsmc Because it s a coprocessor on a PCIe card (at least the one they used is) 6

7 Experimental Design Architecture I7-2600K (I7) On-Demand CPU Governor Xeon E (Sandy Bridge, SB) P-states (per core) Xeon Phi 5110P (Phi) Many Integrated Core (MIC) P-states and C3 (whole CPU) All 3 CPUs are P-state and C3 capable. Not sure why they didn t keep this constant. Or did they? Also, the I7 and Xeon E5 are both based on SB architecture. Better aliases would have been I7 and E5 7

8 Intel Xeon Phi Source: 8

9 Intel Xeon Phi (Internal) Source: 9

10 Observations 3 Sections: Observations on Power Consumption Predicting Peak Power Using Reference Workloads Analyzing Power Consumption Profile of Scientific Applications An attempt to characterize the peak power of workloads and corresponding HPC components (L1, L2, L3 cache, memory, etc) 10

11 Observation 1 Different architectures have different increases in power Floor Benchmarks that target only the CPU registers (no caches or memory) Ceiling Benchmarks that target everything I7 and SB have more dramatic increases because other resources (cache, interconnects, etc) scale up with increasing cores Phi has a ring bus that is fully powered if even 1 core is in use 11

12 Observation 2 Figure 2A Relative peak power increases are different between architectures They are also different depending on the chosen workload More cores, more variation 12

13 Observation 3 Using 1 reference workload to predict peak power of all others isn t accurate at all 13

14 Observation 3 (continued) (A) Pairs of workloads that are similar on one architecture can be different on another (B) Different parallelization platforms (MPI vs OMP, same workload) can be similar on one architecture and different on another 14

15 Observation 4 Different workloads reach their peak power usage at different times That same workload may reach peak power at a different time on a different architecture Same workloads have different power profiles from architecture to architecture (see next slide) 15

16 Observation 4 (continued) 16

17 Observation 5 Power profiles are similar with different number of active cores on the same architecture Regardless of parallelization platform (MPI vs OMP) 17

18 Observation 5 (continued) Seeing peak power spikes early in execution After 40% execution of workload, predicted peak power error below 5% 18

19 Adaptive Power Profiling (APP) k% Sampling Authors Approach Evaluation Corner Cases 19

20 APP: k% sampling Widely used approach (k% sampling) 1. Choose % of workload to run (k%) 2. Run the workload for k% time 3. Collect power usage For multiple cores (# cores) * k% 5 cores 5k% of workload needs to be run (e.g. 1, 2, 3, 4, 5) Rationale: consistent with observation #5 (figure 9) After a certain % of workload is run, error is minimum More cores need more of the workload to run to be accurate 20

21 APP: Authors Approach 1. Profile k% using maximum core count 2. Construct estimation error curve 3. Find the normalized run time where error is below user specified maximum error (k%) 4. Profile remaining core scaling settings using the new k% 21

22 APP: Author s Approach (continued) Run a k% for max cores and collect power trace PP(i) = Power at time i Peak power so far = PPmax(i) Calculate expected error curve, PPEC(i) For user specified accuracy (a%), find i where PPEC(i) < a 22

23 APP: Author s Approach (continued) 23

24 Evaluation: Time Used vs Requested Shows a reduction actual time used vs k% method, which simply uses requested time Not sure of the criteria used to determine workload percentiles 24

25 Evaluation: APP vs k% Shows a difference in prediction from the authors method as vs k% method used as a baseline 25

26 Evaluation: Relaxing Accuracy Relaxing accuracy requirement (a%) doesn t necessarily mean it will be inaccurate by that same amount Not sure of the criteria used to determine workload percentiles 26

27 Evaluation: Accuracy vs Profiling Time Small changes to relax accuracy greatly reduces time to profile (up to 5%) Characterizes accuracy vs profiling time tradeoff Not sure of the criteria used to determine workload percentiles 27

28 Corner Cases Comparison of finding k% for APP with min, median, and max core counts Starting with max core count to find k% (like previously described) is optimal Would be helpful to state how many cores for each architecture was used Likely Min = 2 for all of them Median for SB = 4 What about I7? Phi? Doesn t fit 2 n constraint for benchmarks they imposed in experimental design 28

ECLIPSE 2012 Performance Benchmark and Profiling. August 2012

ECLIPSE 2012 Performance Benchmark and Profiling. August 2012 ECLIPSE 2012 Performance Benchmark and Profiling August 2012 Note The following research was performed under the HPC Advisory Council activities Participating vendors: Intel, Dell, Mellanox Compute resource

More information

A FRAMEWORK FOR CAPACITY ANALYSIS D E B B I E S H E E T Z P R I N C I P A L C O N S U L T A N T M B I S O L U T I O N S

A FRAMEWORK FOR CAPACITY ANALYSIS D E B B I E S H E E T Z P R I N C I P A L C O N S U L T A N T M B I S O L U T I O N S A FRAMEWORK FOR CAPACITY ANALYSIS D E B B I E S H E E T Z P R I N C I P A L C O N S U L T A N T M B I S O L U T I O N S Presented at St. Louis CMG Regional Conference, 4 October 2016 (c) MBI Solutions

More information

Optimize code for Intel Xeon Phi. Discovering bottlenecks without pain

Optimize code for Intel Xeon Phi. Discovering bottlenecks without pain Optimize code for Intel Xeon Phi Discovering bottlenecks without pain Agenda Introduction Allinea MAP and optimizing for Xeon Phi Conclusion Allinea Unified environment A modern integrated environment

More information

Heterogeneity and Dynamicity of Clouds at Scale: Google Trace Analysis

Heterogeneity and Dynamicity of Clouds at Scale: Google Trace Analysis Heterogeneity and Dynamicity of Clouds at Scale: Google Trace Analysis Charles Reiss *, Alexey Tumanov, Gregory R. Ganger, Randy H. Katz *, Michael A. Kozuch * UC Berkeley CMU Intel Labs http://www.istc-cc.cmu.edu/

More information

HPC Power Measurements Procurement Guidelines

HPC Power Measurements Procurement Guidelines Center for Information Services and High Performance Computing (ZIH) HPC Power Measurements Procurement Guidelines Building Energy Efficient High Performance Computing: 4th Annual EE HPC WG Workshop Daniel

More information

NVIDIA QUADRO VIRTUAL DATA CENTER WORKSTATION APPLICATION SIZING GUIDE FOR SIEMENS NX APPLICATION GUIDE. Ver 1.0

NVIDIA QUADRO VIRTUAL DATA CENTER WORKSTATION APPLICATION SIZING GUIDE FOR SIEMENS NX APPLICATION GUIDE. Ver 1.0 NVIDIA QUADRO VIRTUAL DATA CENTER WORKSTATION APPLICATION SIZING GUIDE FOR SIEMENS NX APPLICATION GUIDE Ver 1.0 EXECUTIVE SUMMARY This document provides insights into how to deploy NVIDIA Quadro Virtual

More information

HPC USAGE ANALYTICS. Supercomputer Education & Research Centre Akhila Prabhakaran

HPC USAGE ANALYTICS. Supercomputer Education & Research Centre Akhila Prabhakaran HPC USAGE ANALYTICS Supercomputer Education & Research Centre Akhila Prabhakaran OVERVIEW: BATCH COMPUTE SERVERS Dell Cluster : Batch cluster consists of 3 Nodes of Two Intel Quad Core X5570 Xeon CPUs.

More information

NaviCloud Sphere. NaviCloud Pricing and Billing: By Resource Utilization, at the 95th Percentile. A Time Warner Cable Company.

NaviCloud Sphere. NaviCloud Pricing and Billing: By Resource Utilization, at the 95th Percentile. A Time Warner Cable Company. NaviCloud Sphere NaviCloud Pricing and Billing: By Resource Utilization, at the 95th Percentile June 29, 2011 A Time Warner Cable Company NaviCloud Sphere Pricing, Billing: By Resource Utilization, at

More information

LS-DYNA Performance With MPI Collectives Acceleration. April 2011

LS-DYNA Performance With MPI Collectives Acceleration. April 2011 LS-DYNA Performance With MPI Collectives Acceleration April 2011 Note The following research was performed under the HPC Advisory Council HPC works working group activities Participating vendors: HP, Intel,

More information

High-Performance Computing (HPC) Up-close

High-Performance Computing (HPC) Up-close High-Performance Computing (HPC) Up-close What It Can Do For You In this InfoBrief, we examine what High-Performance Computing is, how industry is benefiting, why it equips business for the future, what

More information

Multi-core Management A new Approach

Multi-core Management A new Approach Multi-core Management A new Approach Dr Marc GATTI, Thales Avionics Marc-j.gatti@fr.thalesgroup.com MAKS IMA Conference 20 th July, Moscow www.thalesgroup.com Abstract Multi-core Management A new Approach

More information

CS 147: Computer Systems Performance Analysis

CS 147: Computer Systems Performance Analysis CS 147: Computer Systems Performance Analysis Approaching Performance Projects CS 147: Computer Systems Performance Analysis Approaching Performance Projects 1 / 35 Overview Overview Overview Planning

More information

Accelerate HPC Development with Allinea Performance Tools. Olly Perks & Florent Lebeau

Accelerate HPC Development with Allinea Performance Tools. Olly Perks & Florent Lebeau Accelerate HPC Development with Allinea Performance Tools Olly Perks & Florent Lebeau Olly.Perks@arm.com Florent.Lebeau@arm.com Agenda 09:00 09:15 09:15 09:30 09:30 09:45 09:45 10:15 10:15 10:30 Introduction

More information

DB12 Benchmark + LHCb Benchmarking. Andrew McNab University of Manchester LHCb and GridPP

DB12 Benchmark + LHCb Benchmarking. Andrew McNab University of Manchester LHCb and GridPP DB12 Benchmark + LHCb Benchmarking Andrew McNab University of Manchester LHCb and GridPP DIRAC 2012 fast benchmark ( DB12 ) What is benchmarking about? DB12 origins and current status: within DIRAC; within

More information

Stay Tuned Computational Science NeSI. Jordi Blasco

Stay Tuned Computational Science NeSI. Jordi Blasco Computational Science Team @ NeSI Jordi Blasco (jordi.blasco@nesi.org.nz) Outline 1 About NeSI CS Team Who we are? 2 Identify the Bottlenecks Identify the Most Popular Apps Profile and Debug 3 Tuning Increase

More information

Addressing the I/O bottleneck of HPC workloads. Professor Mark Parsons NEXTGenIO Project Chairman Director, EPCC

Addressing the I/O bottleneck of HPC workloads. Professor Mark Parsons NEXTGenIO Project Chairman Director, EPCC Addressing the I/O bottleneck of HPC workloads Professor Mark Parsons NEXTGenIO Project Chairman Director, EPCC I/O is key Exascale challenge Parallelism beyond 100 million threads demands a new approach

More information

Cray and Allinea. Maximizing Developer Productivity and HPC Resource Utilization. SHPCP HPC Theater SEG 2016 C O M P U T E S T O R E A N A L Y Z E 1

Cray and Allinea. Maximizing Developer Productivity and HPC Resource Utilization. SHPCP HPC Theater SEG 2016 C O M P U T E S T O R E A N A L Y Z E 1 Cray and Allinea Maximizing Developer Productivity and HPC Resource Utilization SHPCP HPC Theater SEG 2016 10/27/2016 C O M P U T E S T O R E A N A L Y Z E 1 History Cluster Computing to Super Computing

More information

July, 10 th From exotics to vanillas with GPU Murex 2014

July, 10 th From exotics to vanillas with GPU Murex 2014 July, 10 th 2014 From exotics to vanillas with GPU Murex 2014 COMPANY Selected Industry Recognition and Rankings 2013-2014 OVERALL #1 TOP TECHNOLOGY VENDOR #1 Trading Systems #1 Pricing & Risk Analytics

More information

AMR (Adaptive Mesh Refinement) Performance Benchmark and Profiling

AMR (Adaptive Mesh Refinement) Performance Benchmark and Profiling AMR (Adaptive Mesh Refinement) Performance Benchmark and Profiling July 2011 Acknowledgment: - The DoD High Performance Computing Modernization Program - John Bell from Lawrence Berkeley Laboratory Note

More information

STAR-CCM+ Performance Benchmark. August 2010

STAR-CCM+ Performance Benchmark. August 2010 STAR-CCM+ Performance Benchmark August 2010 Note The following research was performed under the HPC Advisory Council activities Participating members: CD-adapco, Dell, Intel, Mellanox Compute resource

More information

Towards Seamless Integration of Data Analytics into Existing HPC Infrastructures

Towards Seamless Integration of Data Analytics into Existing HPC Infrastructures Towards Seamless Integration of Data Analytics into Existing HPC Infrastructures Michael Gienger High Performance Computing Center Stuttgart (HLRS), Germany Redmond May 11, 2017 :: 1 Outline Introduction

More information

Cellulosomes: One of Life s Strongest Biomolecular Bonds Discovered with Use of Supercomputers

Cellulosomes: One of Life s Strongest Biomolecular Bonds Discovered with Use of Supercomputers Published on Scientific Computing (http://www.scientificcomputing.com) Home > Cellulosomes: One of Life s Strongest Biomolecular Bonds Discovered with Use of Supercomputers Cellulosomes: One of Life s

More information

Avoid Paying The Virtualization Tax: Deploying Virtualized BI 4.0 The Right Way

Avoid Paying The Virtualization Tax: Deploying Virtualized BI 4.0 The Right Way Avoid Paying The Virtualization Tax: Deploying Virtualized BI 4.0 The Right Way Material by Ashish C. Morzaria, SAP. @AshishMorzaria Presented by Matthew Shaw, SAP. @MattShaw_on_BI LEARNING POINTS Understanding

More information

Understanding the Behavior of In-Memory Computing Workloads. Rui Hou Institute of Computing Technology,CAS July 10, 2014

Understanding the Behavior of In-Memory Computing Workloads. Rui Hou Institute of Computing Technology,CAS July 10, 2014 Understanding the Behavior of In-Memory Computing Workloads Rui Hou Institute of Computing Technology,CAS July 10, 2014 Outline Background Methodology Results and Analysis Summary 2 Background The era

More information

Joe Butler, Sharon Ruane Intel Labs Europe. May 11, 2018.

Joe Butler, Sharon Ruane Intel Labs Europe. May 11, 2018. Joe Butler, Sharon Ruane Intel Labs Europe. May 11, 2018. Orchestrating apps (content) and network. Application And Content Complexity & demand for network performance. Immersive Media, V2X, IoT. Streaming,

More information

STAND: New Tool for Performance Estimation of the Block Data Processing Algorithms in High-load Systems

STAND: New Tool for Performance Estimation of the Block Data Processing Algorithms in High-load Systems STAND: New Tool for Performance Estimation of the Block Data Processing Algorithms in High-load Systems Victor Minchenkov, Vladimir Bashun St-Petersburg State University of Aerospace Instrumentation {victor,

More information

Capacity Management - Telling the story

Capacity Management - Telling the story Capacity Management - Telling the story What is a Story? It is either: a. an account of incidents or events b. a statement regarding the facts pertinent to a situation in question Data is nothing more

More information

Pre-announcement of Upcoming Procurement, AC2018, at National Supercomputing Centre at Linköping University

Pre-announcement of Upcoming Procurement, AC2018, at National Supercomputing Centre at Linköping University Pre-announcement of Upcoming Procurement, AC2018, at National Supercomputing Centre at Linköping University 2017-06-14 Abstract Linköpings universitet hereby announces the opportunity to participate in

More information

Cori: A Cray XC Pre-Exascale System for NERSC

Cori: A Cray XC Pre-Exascale System for NERSC Cori: A Cray XC Pre-Exascale System for NERSC Katie Antypas, KAntypas@lbl.gov Nicholas Wright, NJWright@lbl.gov Nicholas P. Cardo, NPCardo@lbl.gov Allison Andrews, MNAndrews@lbl.gov Matthew Cordery, MJCordery@lbl.gov

More information

Application-Aware Power Management. Karthick Rajamani, Heather Hanson*, Juan Rubio, Soraya Ghiasi, Freeman Rawson

Application-Aware Power Management. Karthick Rajamani, Heather Hanson*, Juan Rubio, Soraya Ghiasi, Freeman Rawson Application-Aware Power Management Karthick Rajamani, Heather Hanson*, Juan Rubio, Soraya Ghiasi, Freeman Rawson Power-Aware Systems, IBM Austin Research Lab *University of Texas at Austin Outline Motivation

More information

Priority-enabled Scheduling for Resizable Parallel Applications

Priority-enabled Scheduling for Resizable Parallel Applications Priority-enabled Scheduling for Resizable Parallel Applications Rajesh Sudarsan, Student Member, IEEE, Calvin J. Ribbens,and Srinidhi Varadarajan, Member, IEEE Abstract In this paper, we illustrate the

More information

Windows Server Capacity Management 101

Windows Server Capacity Management 101 Windows Server Capacity Management 101 What is Capacity Management? ITIL definition of Capacity Management is: Capacity Management is responsible for ensuring that adequate capacity is available at all

More information

The Impact of Agile. Quantified.

The Impact of Agile. Quantified. The Impact of Agile. Quantified. Agile and lean are built on a foundation of continuous improvement: You need to inspect, learn from and adapt your performance to keep improving. Enhancing performance

More information

Application Migration Patterns for the Service Oriented Cloud

Application Migration Patterns for the Service Oriented Cloud Topic: Cloud Computing Date: July 2011 Author: Lawrence Wilkes Application Migration Patterns for the Service Oriented Cloud Abstract: As well as deploying new applications to the cloud, many organizations

More information

Scalability and High Performance with MicroStrategy 10

Scalability and High Performance with MicroStrategy 10 Scalability and High Performance with MicroStrategy 10 Enterprise Analytics and Mobility at Scale. Copyright Information All Contents Copyright 2017 MicroStrategy Incorporated. All Rights Reserved. Trademark

More information

PULLING BACK THE CURTAIN: VIEWABILITY & DIRECT RESPONSE

PULLING BACK THE CURTAIN: VIEWABILITY & DIRECT RESPONSE PULLING BACK THE CURTAIN: VIEWABILITY & DIRECT RESPONSE WHAT WE KNOW WITH BRANDING CAMPAIGNS VIEWABILITY + BRANDING IMPACT = STRONG RELATIONSHIP BUT, SOME IMPRESSIONS BELOW THE STANDARD CAN HAVE AN IMPACT

More information

IBM xseries 430. Versatile, scalable workload management. Provides unmatched flexibility with an Intel architecture and open systems foundation

IBM xseries 430. Versatile, scalable workload management. Provides unmatched flexibility with an Intel architecture and open systems foundation Versatile, scalable workload management IBM xseries 430 With Intel technology at its core and support for multiple applications across multiple operating systems, the xseries 430 enables customers to run

More information

DELL EMC POWEREDGE 14G SERVER PORTFOLIO

DELL EMC POWEREDGE 14G SERVER PORTFOLIO DELL EMC POWEREDGE 14G SERVER PORTFOLIO Transformation without compromise Seize your share of a huge market opportunity and accelerate your business by combining sales of the exciting new Dell EMC PowerEdge

More information

IBM Power Systems. Bringing Choice and Differentiation to Linux Infrastructure

IBM Power Systems. Bringing Choice and Differentiation to Linux Infrastructure IBM Power Systems Bringing Choice and Differentiation to Linux Infrastructure Stefanie Chiras, Ph.D. Vice President IBM Power Systems Offering Management Client initiatives Cognitive Cloud Economic value

More information

Infor LN Minimum hardware requirements. Sizing Documentation

Infor LN Minimum hardware requirements. Sizing Documentation Infor LN Minimum hardware requirements Sizing Documentation Copyright 2014 Infor Important Notices The material contained in this publication (including any supplementary information) constitutes and contains

More information

The MLC Cost Reduction Cookbook. Scott Chapman Enterprise Performance Strategies

The MLC Cost Reduction Cookbook. Scott Chapman Enterprise Performance Strategies The MLC Cost Reduction Cookbook Scott Chapman Enterprise Performance Strategies Scott.Chapman@epstrategies.com Contact, Copyright, and Trademark Notices Questions? Send email to Scott at scott.chapman@epstrategies.com,

More information

Get the Most Bang for Your Buck #EC2 #Winning

Get the Most Bang for Your Buck #EC2 #Winning Get the Most Bang for Your Buck #EC2 #Winning Joshua Burgin General Manager, EC2 Spot Amazon Web Services June 28, 2017 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon EC2

More information

Understanding and Controlling Processor Affinity PRESENTED BY: Kent Milfeld. Slides at: tinyurl.com/chpc-2017-affinity. National Conference

Understanding and Controlling Processor Affinity PRESENTED BY: Kent Milfeld. Slides at: tinyurl.com/chpc-2017-affinity. National Conference Understanding and Controlling Processor Affinity PRESENTED BY: National Conference Kent Milfeld Slides at: tinyurl.com/chpc-2017-affinity Outline Motivation Affinity -- what is it OpenMP Affinity Ways

More information

OpenSHMEM Birds of a Feather. November 15, 2017

OpenSHMEM Birds of a Feather. November 15, 2017 Open Birds of a Feather November 15, 2017 Legal Notices and Disclaimers Intel technologies features and benefits depend on system configuration and may require enabled hardware, software or service activation.

More information

OpenMX Performance Benchmark and Profiling. May 2011

OpenMX Performance Benchmark and Profiling. May 2011 OpenMX Performance Benchmark and Profiling May 2011 Note The following research was performed under the HPC Advisory Council HPC works working group activities Participating vendors: HP, Intel, Mellanox

More information

HPC Trends for Addison Snell

HPC Trends for Addison Snell HPC Trends for 2015 Addison Snell addison@intersect360.com A New Era in HPC Supercomputing for U.S. Industry Report by U.S. Council on Competitiveness Justifications for new levels of supercomputing for

More information

Programming Models for Heterogeneous Computing Systems

Programming Models for Heterogeneous Computing Systems Programming Models for Heterogeneous Computing Systems Paul Chow University of Toronto Department of Electrical and Computer Engineering Workshop on Many-Core Embedded Systems October 3, 2013 A Heterogeneous

More information

The Modular Supercomputer Architecture and its application in HPC and HPDA

The Modular Supercomputer Architecture and its application in HPC and HPDA The Modular Supercomputer Architecture and its application in HPC and HPDA Damian Alvarez Jülich Supercomputing Centre, JSC (Germany) The DEEP Projects Research & innovation projects co-funded by the European

More information

IBM Virtualization Manager Xen Summit, April 2007

IBM Virtualization Manager Xen Summit, April 2007 IBM Virtualization Manager Xen Summit, April 2007 Senthil Bakthavachalam 2006 IBM Corporation The Promise of Virtualization System Administrator Easily deploy new applications and adjust priorities Easily

More information

Performance Study: STAR-CD v4 on PanFS

Performance Study: STAR-CD v4 on PanFS Performance Study: STAR-CD v4 on PanFS Stan Posey Industry and Applications Market Development Panasas, Fremont, CA, USA Bill Loewe Technical Staff Member, Applications Engineering Panasas, Fremont, CA,

More information

Leistungsanalyse von Rechnersystemen

Leistungsanalyse von Rechnersystemen Center for Information Services and High Performance Computing (ZIH) Leistungsanalyse von Rechnersystemen Capacity Planning Zellescher Weg 12 Raum WIL A113 Tel. +49 351-463 - 39835 Matthias Müller (matthias.mueller@tu-dresden.de)

More information

Graph Optimization Algorithms for Sun Grid Engine. Lev Markov

Graph Optimization Algorithms for Sun Grid Engine. Lev Markov Graph Optimization Algorithms for Sun Grid Engine Lev Markov Sun Grid Engine SGE management software that optimizes utilization of software and hardware resources in heterogeneous networked environment.

More information

ALCF SITE UPDATE IXPUG 2018 MIDDLE EAST MEETING. DAVID E. MARTIN Manager, Industry Partnerships and Outreach. MARK FAHEY Director of Operations

ALCF SITE UPDATE IXPUG 2018 MIDDLE EAST MEETING. DAVID E. MARTIN Manager, Industry Partnerships and Outreach. MARK FAHEY Director of Operations erhtjhtyhy IXPUG 2018 MIDDLE EAST MEETING ALCF SITE UPDATE DAVID E. MARTIN Manager, Industry Partnerships and Outreach Argonne Leadership Computing Facility MARK FAHEY Director of Operations Argonne Leadership

More information

Multi-tenant GPU Clusters for Deep Learning Workloads: Analysis and Implications

Multi-tenant GPU Clusters for Deep Learning Workloads: Analysis and Implications Multi-tenant GPU Clusters for Deep Learning Workloads: Analysis and Implications Myeongjae Jeon, Shivaram Venkataraman, Amar Phanishayee, Junjie Qian, Wencong Xiao, Fan Yang Microsoft Research Abstract

More information

Bias Scheduling in Heterogeneous Multicore Architectures. David Koufaty Dheeraj Reddy Scott Hahn

Bias Scheduling in Heterogeneous Multicore Architectures. David Koufaty Dheeraj Reddy Scott Hahn Bias Scheduling in Heterogeneous Multicore Architectures David Koufaty Dheeraj Reddy Scott Hahn Motivation Mainstream multicore processors consist of identical cores Complexity dictated by product goals,

More information

Optimizing Fine-grained Communication in a Biomolecular Simulation Application on Cray XK6

Optimizing Fine-grained Communication in a Biomolecular Simulation Application on Cray XK6 Optimizing Fine-grained Communication in a Biomolecular Simulation Application on Cray XK6 Yanhua Sun 1 Gengbin Zheng 1 Chao Mei 1 Eric J. Bohm 1 James C. Phillips 1 Terry Jones 2 Laxmikant(Sanjay) V.

More information

HPC Trends for 2017 HPCAC Lugano. Michael Feldman, Managing Editor, TOP500 News

HPC Trends for 2017 HPCAC Lugano. Michael Feldman, Managing Editor, TOP500 News HPC Trends for 2017 HPCAC Lugano Michael Feldman, Managing Editor, TOP500 News Intersect360 Research in 2017 10 year history of HPC analyst business Covering high performance data center markets, including

More information

Power measurement at the exascale

Power measurement at the exascale Power measurement at the exascale Nick Johnson, James Perry & Michèle Weiland Nick Johnson Adept Project, EPCC nick.johnson@ed.ac.uk Motivation The current exascale targets are: One exaflop at a power

More information

Rapid ICT prototyping in Ireland with ICHEC

Rapid ICT prototyping in Ireland with ICHEC Rapid ICT prototyping in Ireland with ICHEC Overview Jean- Christophe JC Desplat 11 th February 2015 Agenda Centre overview Technology walkthrough Training & educaion programme Data analyics Business engagement

More information

Dell EMC Ready Solutions for HPC Lustre Storage. Forrest Ling HPC Enterprise Technolgist at Dell EMC Greater China

Dell EMC Ready Solutions for HPC Lustre Storage. Forrest Ling HPC Enterprise Technolgist at Dell EMC Greater China Dell EMC Ready Solutions for HPC Lustre Storage Forrest Ling HPC Enterprise Technolgist at Dell EMC Greater China 2018.10.23 Dell EMC Supports HPC Open Source Software Support Open Source Software projects

More information

comp 180 Lecture 04 Outline of Lecture 1. The Role of Computer Performance 2. Measuring Performance

comp 180 Lecture 04 Outline of Lecture 1. The Role of Computer Performance 2. Measuring Performance Outline of Lecture 1. The Role of Computer Performance 2. Measuring Performance Summary The CPU time can be decomposed as follows: CPU time = Instructions --------------------------------- Program Clock

More information

The Myths Behind Software Metrics. Myths and Superstitions

The Myths Behind Software Metrics. Myths and Superstitions The Myths Behind Software Metrics Pacific Northwest Software Quality Conference October 14, 2013 Douglas Hoffman, BACS, MBA, MSEE, ASQ-CSQE, ASQ-CMQ/OE, ASQ Fellow Software Quality Methods, LLC. (SQM)

More information

Goya Inference Platform & Performance Benchmarks. Rev January 2019

Goya Inference Platform & Performance Benchmarks. Rev January 2019 Goya Inference Platform & Performance Benchmarks Rev. 1.6.1 January 2019 Habana Goya Inference Platform Table of Contents 1. Introduction 2. Deep Learning Workflows Training and Inference 3. Goya Deep

More information

IBM. Processor Migration Capacity Analysis in a Production Environment. Kathy Walsh. White Paper January 19, Version 1.2

IBM. Processor Migration Capacity Analysis in a Production Environment. Kathy Walsh. White Paper January 19, Version 1.2 IBM Processor Migration Capacity Analysis in a Production Environment Kathy Walsh White Paper January 19, 2015 Version 1.2 IBM Corporation, 2015 Introduction Installations migrating to new processors often

More information

InfoSphere DataStage Grid Solution

InfoSphere DataStage Grid Solution InfoSphere DataStage Grid Solution Julius Lerm IBM Information Management 1 2011 IBM Corporation What is Grid Computing? Grid Computing doesn t mean the same thing to all people. GRID Definitions include:

More information

SE350: Operating Systems. Lecture 6: Scheduling

SE350: Operating Systems. Lecture 6: Scheduling SE350: Operating Systems Lecture 6: Scheduling Main Points Definitions Response time, throughput, scheduling policy, Uniprocessor policies FIFO, SJF, Round Robin, Multiprocessor policies Scheduling sequential

More information

High Performance Computing(HPC) & Software Stack

High Performance Computing(HPC) & Software Stack IBM HPC Developer Education @ TIFR, Mumbai High Performance Computing(HPC) & Software Stack January 30-31, 2012 Pidad D'Souza (pidsouza@in.ibm.com) IBM, System & Technology Group 2002 IBM Corporation Agenda

More information

Fast Tracking Product Design can you afford the luxury of more time?

Fast Tracking Product Design can you afford the luxury of more time? Wes Shimanek, Workstation Segment Manager, Datacenter Group, Technical Computing Group, Intel Corporation Fast Tracking Product Design can you afford the luxury of more time? Solid Edge University 2014

More information

A Profile Guided Approach to Scheduling in Cluster and Multi-cluster Systems

A Profile Guided Approach to Scheduling in Cluster and Multi-cluster Systems A Profile Guided Approach to Scheduling in Cluster and Multi-cluster Systems Arvind Sridhar and Dan Stanzione Ph. D. {ahsridha, dstanzi}@asu.edu Fulton High Performance Computing, Arizona State University

More information

David Martin IXPUG President

David Martin IXPUG President David Martin IXPUG President dem@alcf.anl.gov Worldwide organization Optimization of scientific applications on Intel based HPC systems Now Intel extreme Performance Users Group Originally Intel Xeon Phi

More information

Innovation Without Limits. Your Guide to High Performance Computing in the Cloud

Innovation Without Limits. Your Guide to High Performance Computing in the Cloud Innovation Without Limits Your Guide to High Performance Computing in the Cloud 4 5 6 7 8 10 12 What Could You Accomplish with a Million Cores? Access Resources Quickly Leverage Latest Technology Collaborate

More information

Improve Your Productivity with Simplified Cluster Management. Copyright 2010 Platform Computing Corporation. All Rights Reserved.

Improve Your Productivity with Simplified Cluster Management. Copyright 2010 Platform Computing Corporation. All Rights Reserved. Improve Your Productivity with Simplified Cluster Management TORONTO 12/2/2010 Agenda Overview Platform Computing & Fujitsu Partnership PCM Fujitsu Edition Overview o Basic Package o Enterprise Package

More information

Micro-Virtualization. Maximize processing power use and improve system/energy efficiency

Micro-Virtualization. Maximize processing power use and improve system/energy efficiency Micro-Virtualization Maximize processing power use and improve system/energy efficiency Disclaimers We don t know everything But we know there is a problem and we re solving (at least part of) it And we

More information

Ensure Your Servers Can Support All the Benefits of Virtualization and Private Cloud The State of Server Virtualization... 8

Ensure Your Servers Can Support All the Benefits of Virtualization and Private Cloud The State of Server Virtualization... 8 ... 4 The State of Server Virtualization... 8 Virtualization Comfort Level SQL Server... 12 Case in Point SAP... 14 Virtualization The Server Platform Really Matters... 18 The New Family of Intel-based

More information

Business Insight at the Speed of Thought

Business Insight at the Speed of Thought BUSINESS BRIEF Business Insight at the Speed of Thought A paradigm shift in data processing that will change your business Advanced analytics and the efficiencies of Hybrid Cloud computing models are radically

More information

NWChem Performance Benchmark and Profiling. October 2010

NWChem Performance Benchmark and Profiling. October 2010 NWChem Performance Benchmark and Profiling October 2010 Note The following research was performed under the HPC Advisory Council activities Participating vendors: Intel, Dell, Mellanox Compute resource

More information

REQUEST FOR PROPOSAL FOR

REQUEST FOR PROPOSAL FOR REQUEST FOR PROPOSAL FOR HIGH PERFORMANCE COMPUTING (HPC) SOLUTION Ref. No. PHY/ALK/43 (27/11/2012) by DEPARTMENT OF PHYSICS UNIVERSITY OF PUNE PUNE - 411 007 INDIA NOVEMBER 27, 2012 1 Purpose of this

More information

Job Scheduling with Lookahead Group Matchmaking for Time/Space Sharing on Multi-core Parallel Machines

Job Scheduling with Lookahead Group Matchmaking for Time/Space Sharing on Multi-core Parallel Machines Job Scheduling with Lookahead Group Matchmaking for Time/Space Sharing on Multi-core Parallel Machines Xijie Zeng and Angela Sodan University of Windsor, Windsor ON N9B 3P4, Canada zengx@uwindsor.ca,acsodan@uwindsor.ca

More information

Workflow Analysis An Approach to Characterize Application and System Needs

Workflow Analysis An Approach to Characterize Application and System Needs Slide 1 Workflow Analysis An Approach to Characterize Application and System Needs MSST 2016 Dave Montoya May 3, 2016 Slide 2 Why are we discussing workflow? Exascale is driving tighter integration! Economics

More information

8 Key Steps to Getting TV Attribution Right

8 Key Steps to Getting TV Attribution Right 8 Key Steps to Getting TV Attribution Right BY JUAN PABLO PEREIRA HEAD OF BUSINESS INNOVATION, VP, MARKETING SERVICES NEUSTAR TV advertising is the giant megaphone that drives customers to a next action,

More information

MEASUREMENT DIVIDE INSIGHTS

MEASUREMENT DIVIDE INSIGHTS T H E FA C E B O O K MEASUREMENT DIVIDE R A K U T E N M A R K E T I N G INSIGHTS MEASUREMENT DISCREPANCIES Attributable revenue from ad campaigns might be higher than advertisers' web analytics lead them

More information

Oracle PaaS and IaaS Universal Credits Service Descriptions

Oracle PaaS and IaaS Universal Credits Service Descriptions Oracle PaaS and IaaS Universal Credits Service Descriptions December 1, 2017 Oracle PaaS_IaaS_Universal_CreditsV120117 1 Metrics... 3 Oracle PaaS and IaaS Universal Credit... 8 Oracle PaaS and IaaS Universal

More information

Ask the right question, regardless of scale

Ask the right question, regardless of scale Ask the right question, regardless of scale Customers use 100s to 1,000s Of cores to answer business-critical Questions they couldn t have done before. Trivial to support different use cases Different

More information

UNITE Conference. May 13-16, 2012 St Louis, MO. Copyright 2012 TeamQuest Corporation. All Rights Reserved.

UNITE Conference. May 13-16, 2012 St Louis, MO. Copyright 2012 TeamQuest Corporation. All Rights Reserved. UNITE Conference May 13-16, 2012 St Louis, MO The Renaissance of Performance & Capacity Management in the 21st Century Steve Cullen TeamQuest and the TeamQuest logo are registered trademarks in the US,

More information

Dennis Bradford, Sundaram Chinthamani, Jesus Corbal, Adhiraj Hassan, Ken Janik, Nawab Ali

Dennis Bradford, Sundaram Chinthamani, Jesus Corbal, Adhiraj Hassan, Ken Janik, Nawab Ali Dennis Bradford, Sundaram Chinthamani, Jesus Corbal, Adhiraj Hassan, Ken Janik, Nawab Ali 1 Legal Disclaimers INFORMATION IN THIS DOCUMENT IS PROVIDED IN CONNECTION WITH INTEL PRODUCTS. NO LICENSE, EXPRESS

More information

HETEROGENEOUS SYSTEM ARCHITECTURE: FROM THE HPC USAGE PERSPECTIVE

HETEROGENEOUS SYSTEM ARCHITECTURE: FROM THE HPC USAGE PERSPECTIVE HETEROGENEOUS SYSTEM ARCHITECTURE: FROM THE HPC USAGE PERSPECTIVE Haibo Xie, Ph.D. Chief HSA Evangelist AMD China AGENDA: GPGPU in HPC, what are the challenges Introducing Heterogeneous System Architecture

More information

ORACLE DATABASE PERFORMANCE: VMWARE CLOUD ON AWS PERFORMANCE STUDY JULY 2018

ORACLE DATABASE PERFORMANCE: VMWARE CLOUD ON AWS PERFORMANCE STUDY JULY 2018 ORACLE DATABASE PERFORMANCE: VMWARE CLOUD ON AWS PERFORMANCE STUDY JULY 2018 Table of Contents Executive Summary...3 Introduction...3 Test Environment... 4 Test Workload... 6 Virtual Machine Configuration...

More information

MANAGEMENT CLOUD. Leveraging Your E-Business Suite

MANAGEMENT CLOUD. Leveraging Your E-Business Suite MANAGEMENT CLOUD Leveraging Your E-Business Suite Leverage Oracle E-Business Suite with Oracle Management Cloud. Oracle E-Business Suite is the industry s most comprehensive suite of business applications

More information

White paper A Reference Model for High Performance Data Analytics(HPDA) using an HPC infrastructure

White paper A Reference Model for High Performance Data Analytics(HPDA) using an HPC infrastructure White paper A Reference Model for High Performance Data Analytics(HPDA) using an HPC infrastructure Discover how to reshape an existing HPC infrastructure to run High Performance Data Analytics (HPDA)

More information

What's new in Allinea's tools From easy batch script integration & remote access to energy profiling

What's new in Allinea's tools From easy batch script integration & remote access to energy profiling What's new in Allinea's tools From easy batch script integration & remote access to energy profiling Introduction Agenda Overview of HPC current and future needs What s new in Allinea s tools Transitioning

More information

Novel HPC technologies for Rapid Analysis in Bioinformatics Presenter: Paul Walsh, nsilico Life Science Ltd, Ireland

Novel HPC technologies for Rapid Analysis in Bioinformatics Presenter: Paul Walsh, nsilico Life Science Ltd, Ireland Presenter: Paul Walsh, nsilico Life Science Ltd, Ireland Tristan Cabel, Gabriel Hautreux, Eric Boyer: CINES, France Simon Wong: ICHEC, Ireland Nicolas Mignerey: GENCI, France Synthetic Biotech Revolution!

More information

Performance Engineering for High-Tech Systems: Crossing Boundaries

Performance Engineering for High-Tech Systems: Crossing Boundaries Twan Basten Eindhoven University of Technology & TNO Embedded Systems Innovation P A G E 1 Joint work with many others Funding: Artemis EMC2, Almarvi STW Robust CPS program Min. of Economic Affairs, Océ

More information

IBM HPC DIRECTIONS. Dr Don Grice. ECMWF Workshop October 31, IBM Corporation

IBM HPC DIRECTIONS. Dr Don Grice. ECMWF Workshop October 31, IBM Corporation IBM HPC DIRECTIONS Dr Don Grice ECMWF Workshop October 31, 2006 2006 IBM Corporation IBM HPC Directions What s Changing? The Rate of Frequency Improvement is Slowing Moore s Law (Frequency improvement)

More information

Econ 300: Intermediate Microeconomics, Spring 2014 Final Exam Study Guide 1

Econ 300: Intermediate Microeconomics, Spring 2014 Final Exam Study Guide 1 Econ 300: Intermediate Microeconomics, Spring 2014 Final Exam Study Guide 1 Chronological order of topics covered in class (to the best of my memory). Introduction to Microeconomics (Chapter 1) What is

More information

Cisco Unified Workforce Optimization for Cisco Unified Contact Center Express 9.0

Cisco Unified Workforce Optimization for Cisco Unified Contact Center Express 9.0 Data Sheet Cisco Unified Workforce Optimization for Cisco Unified Contact Center Express 9.0 Cisco Unified Communications Solutions unify voice, video, data, and mobile applications on fixed and mobile

More information

ANSYS 16.0 高性能计算技术 - for CFD & FEA Applications

ANSYS 16.0 高性能计算技术 - for CFD & FEA Applications ANSYS 16.0 高性能计算技术 - for CFD & FEA Applications 李占营 / 高级应用工程师 2015 ANSYS, Inc. 1 Outline 概述 High-Performance Computing Brief Intro 简介 R16.0 Update on HPC Features & Capabilities 新功能 Optimized for latest

More information

Get a Second Opinion: Enterprise GIS Health Checks. Matt Marino, Esri Sam Libby, Esri

Get a Second Opinion: Enterprise GIS Health Checks. Matt Marino, Esri Sam Libby, Esri Get a Second Opinion: Enterprise GIS Health Checks Matt Marino, Esri Sam Libby, Esri What is an Enterprise GIS Health Check An Onsite Engagement Focusing On: - Proactively reviewing and assessing current

More information

NAS-Wide Performance: Impact of Select Uncertainty Factors and Implications for Experimental Design

NAS-Wide Performance: Impact of Select Uncertainty Factors and Implications for Experimental Design Approved for Public Release: 12-0370. Distribution Unlimited. NAS-Wide Performance: Impact of Select Uncertainty Factors and Implications for Experimental Design Gareth O. Coville, MITRE Corporation Billy

More information

RAPIDS GPU POWERED MACHINE LEARNING

RAPIDS GPU POWERED MACHINE LEARNING RAPIDS GPU POWERED MACHINE LEARNING RISE OF GPU COMPUTING APPLICATIONS 10 7 10 6 GPU-Computing perf 1.5X per year 1000X by 2025 ALGORITHMS 10 5 1.1X per year SYSTEMS 10 4 CUDA ARCHITECTURE 10 3 1.5X per

More information

Accelerating Computing - Enhance Big Data & in Memory Analytics. Michael Viray Product Manager, Power Systems

Accelerating Computing - Enhance Big Data & in Memory Analytics. Michael Viray Product Manager, Power Systems Accelerating Computing - Enhance Big Data & in Memory Analytics Michael Viray Product Manager, Power Systems viraymv@ph.ibm.com ASEAN Real-time Cognitive Analytics -- a Game Changer Deliver analytics in

More information