1 The Case for Versatile Storage System NetSysLab The University of British Columbia Samer Al-Kiswany, Abdullah Gharaibeh, Matei Ripeanu.

Slides:



Advertisements
Similar presentations
Towards Automating the Configuration of a Distributed Storage System Lauro B. Costa Matei Ripeanu {lauroc, NetSysLab University of British.
Advertisements

Starfish: A Self-tuning System for Big Data Analytics.
1 StoreGPU Exploiting Graphics Processing Units to Accelerate Distributed Storage Systems NetSysLab The University of British Columbia Samer Al-Kiswany.
CoMPI: Enhancing MPI based applications performance and scalability using run-time compression. Rosa Filgueira, David E.Singh, Alejandro Calderón and Jesús.
1 A GPU Accelerated Storage System NetSysLab The University of British Columbia Abdullah Gharaibeh with: Samer Al-Kiswany Sathish Gopalakrishnan Matei.
ParaMEDIC: Parallel Metadata Environment for Distributed I/O and Computing P. Balaji, Argonne National Laboratory W. Feng and J. Archuleta, Virginia Tech.
Size Matters : Space/Time Tradeoffs to Improve GPGPU Application Performance Abdullah Gharaibeh Matei Ripeanu NetSysLab The University of British Columbia.
The Energy Case for Graph Processing on Hybrid Platforms Abdullah Gharaibeh, Lauro Beltrão Costa, Elizeu Santos-Neto and Matei Ripeanu NetSysLab The University.
1 Harvesting the Opportunity of GPU-Based Acceleration for Data-Intensive Applications Matei Ripeanu Networked Systems Laboratory (NetSysLab) University.
Are P2P Data-Dissemination Techniques Viable in Today's Data- Intensive Scientific Collaborations? Samer Al-Kiswany – University of British Columbia joint.
Multithreading and Dataflow Architectures CPSC 321 Andreas Klappenecker.
VMFlock: VM Co-Migration Appliance for the Cloud Samer Al-Kiswany With: Dinesh Subhraveti Prasenjit Sarkar Matei Ripeanu.
Cloud Computing for Chemical Property Prediction Paul Watson School of Computing Science Newcastle University, UK Microsoft Cloud.
Enabling Cross-Layer Optimizations in Storage Systems with Custom Metadata Elizeu Santos-Neto Samer Al-Kiswany Nazareno Andrade Sathish Gopalakrishnan.
EEC-681/781 Distributed Computing Systems Lecture 3 Wenbing Zhao Department of Electrical and Computer Engineering Cleveland State University
Figure 1.1 Interaction between applications and the operating system.
Where to go from here? Get real experience building systems! Opportunities: 496 projects –More projects:
eGovernance Under guidance of Dr. P.V. Kamesam IBM Research Lab New Delhi Ashish Gupta 3 rd Year B.Tech, Computer Science and Engg. IIT Delhi.
1 stdchk : A Checkpoint Storage System for Desktop Grid Computing Matei Ripeanu – UBC Sudharshan S. Vazhkudai – ORNL Abdullah Gharaibeh – UBC The University.
IBM RS/6000 SP POWER3 SMP Jari Jokinen Pekka Laurila.
Beyond Music Sharing: An Evaluation of Peer-to-Peer Data Dissemination Techniques in Large Scientific Collaborations Thesis defense: Samer Al-Kiswany.
1 Exploring Data Reliability Tradeoffs in Replicated Storage Systems NetSysLab The University of British Columbia Abdullah Gharaibeh Matei Ripeanu.
Exploring the Tradeoffs of Configurability and Heterogeneity in Multicore Embedded Systems + Also Affiliated with NSF Center for High- Performance Reconfigurable.
Computer System Lifecycle Chapter 1. Introduction Computer System users, administrators, and designers are all interested in performance evaluation. Whether.
1 Exploring Data Reliability Tradeoffs in Replicated Storage Systems NetSysLab The University of British Columbia Abdullah Gharaibeh Advisor: Professor.
A Workflow-Aware Storage System Emalayan Vairavanathan 1 Samer Al-Kiswany, Lauro Beltrão Costa, Zhao Zhang, Daniel S. Katz, Michael Wilde, Matei Ripeanu.
Computer System Architectures Computer System Software
ECE 526 – Network Processing Systems Design Network Processor Architecture and Scalability Chapter 13,14: D. E. Comer.
Oracle on Windows Server Introduction to Oracle10g on Microsoft Windows Server.
11 If you were plowing a field, which would you rather use? Two oxen, or 1024 chickens? (Attributed to S. Cray) Abdullah Gharaibeh, Lauro Costa, Elizeu.
Emalayan Vairavanathan
Principles of Scalable HPC System Design March 6, 2012 Sue Kelly Sandia National Laboratories Abstract: Sandia National.
DISTRIBUTED COMPUTING
The Center for Autonomic Computing is supported by the National Science Foundation under Grant No NSF CAC Seminannual Meeting, October 5 & 6,
1. 2 Corollary 3 System Overview Second Key Idea: Specialization Think GoogleFS.
Experience with Using a Performance Predictor During Development a Distributed Storage System Tale Lauro Beltrão Costa *, João Brunet +, Lile Hattori #,
1 Configurable Security for Scavenged Storage Systems NetSysLab The University of British Columbia Abdullah Gharaibeh with: Samer Al-Kiswany, Matei Ripeanu.
DotSlash An Automated Web Hotspot Rescue System Jonathan Bulava CSC8530 – Distributed Systems Dr. Paul Schragger.
An Autonomic Framework in Cloud Environment Jiedan Zhu Advisor: Prof. Gagan Agrawal.
Introduction to Apache OODT Yang Li Mar 9, What is OODT Object Oriented Data Technology Science data management Archiving Systems that span scientific.
Energy Prediction for I/O Intensive Workflow Applications 1 MASc Exam Hao Yang NetSysLab The Electrical and Computer Engineering Department The University.
Building a Parallel File System Simulator E Molina-Estolano, C Maltzahn, etc. UCSC Lab, UC Santa Cruz. Published in Journal of Physics, 2009.
SciDAC All Hands Meeting, March 2-3, 2005 Northwestern University PIs:Alok Choudhary, Wei-keng Liao Graduate Students:Avery Ching, Kenin Coloma, Jianwei.
Amy Apon, Pawel Wolinski, Dennis Reed Greg Amerson, Prathima Gorjala University of Arkansas Commercial Applications of High Performance Computing Massive.
Copyright © 2009 Pearson Education, Inc. Publishing as Pearson Addison-Wesley Principles of Parallel Programming First Edition by Calvin Lin Lawrence Snyder.
DOE PI Meeting at BNL 1 Lightweight High-performance I/O for Data-intensive Computing Jun Wang Computer Architecture and Storage System Laboratory (CASS)
1 MosaStore -A Versatile Storage System Lauro Costa, Abdullah Gharaibeh, Samer Al-Kiswany, Matei Ripeanu, Emalayan Vairavanathan, (and many others from.
Towards Exascale File I/O Yutaka Ishikawa University of Tokyo, Japan 2009/05/21.
CCGrid 2014 Improving I/O Throughput of Scientific Applications using Transparent Parallel Compression Tekin Bicer, Jian Yin and Gagan Agrawal Ohio State.
Architecture Models. Readings r Coulouris, Dollimore and Kindberg Distributed Systems: Concepts and Design Edn. 3 m Note: All figures from this book.
Core-Selectability in Chip-Multiprocessors Hashem H. Najaf-abadi Niket K. Choudhary Eric Rotenberg.
Improving Disk Throughput in Data-Intensive Servers Enrique V. Carrera and Ricardo Bianchini Department of Computer Science Rutgers University.
Computer Architecture 2 nd year (computer and Information Sc.)
Department of Computer Science MapReduce for the Cell B. E. Architecture Marc de Kruijf University of Wisconsin−Madison Advised by Professor Sankaralingam.
1 If you were plowing a field, which would you rather use? Two oxen, or 1024 chickens? (Attributed to S. Cray)
Vector and symbolic processors
1 If you were plowing a field, which would you rather use? Two oxen, or 1024 chickens? (Attributed to S. Cray)
DOE Network PI Meeting 2005 Runtime Data Management for Data-Intensive Scientific Applications Xiaosong Ma NC State University Joint Faculty: Oak Ridge.
Lawrence Livermore National Laboratory 1 Science & Technology Principal Directorate - Computation Directorate Scalable Fault Tolerance for Petascale Systems.
Background Computer System Architectures Computer System Software.
LIOProf: Exposing Lustre File System Behavior for I/O Middleware
Breaking the frontiers of the Grid R. Graciani EGI TF 2012.
Presented by: Nick Kirchem Feb 13, 2004
University of Maryland College Park
SOFTWARE DESIGN AND ARCHITECTURE
Liang Chen Advisor: Gagan Agrawal Computer Science & Engineering
Flow Path Model of Superscalars
Algorithm Design.
A Software-Defined Storage for Workflow Applications
On the Role of Burst Buffers in Leadership-Class Storage Systems
Presentation transcript:

1 The Case for Versatile Storage System NetSysLab The University of British Columbia Samer Al-Kiswany, Abdullah Gharaibeh, Matei Ripeanu

2 Introduction HotStorage ‘09 Versatile Storage System for large-scale platforms: Underutilized resources Application specialization The Deployment Approach: Configured at deployment time Coupled with the target application Potential: Higher performance and scalability

3 Platform Example – Argonne Blue Gene/P 160K cores 10 Gb/s Switch Complex GPFS 24 servers IO rate : 8GBps= 51KBps / core !! HotStorage ‘09 2.5K IO Nodes Torus Network 2.5 GBps per node 3D Torus 850 MBps per 64 nodes Tree Under utilized resources.

4 Workload Characteristics HotStorage ‘09 Workflows – Execution stages communicating through intermediate temporary files Source [Zhao et. al. SIGMOD record ‘05] Input file Output file Compute

5 Workload Characteristics HotStorage ‘09 Workflows – Execution stages communicating through intermediate temporary files Tibi Stef-Praun, et. al. [ e-Social Science ‘07 ]

6 Workload Characteristics Workflows – Execution stages communicating through intermediate temporary files HotStorage ‘09 AxesOptimizations Data life time (temporary  ) Application informed caching Read (Seq.  )Read-ahead Write (Seq.  ) Asynch. write Consistency (no  )Relaxed Consistency Workflows 

7 Workload Characteristics Data Analysis – Analyze/search large data sets (e.g. BLAST) HotStorage ‘09 BLAST Match new sequences with a data set of known sequences (linear search) AxesOptimizations Data life time (temporary  ) Application informed caching Read (Seq.   )Read-ahead Write (Seq.  ) Asynch. write Consistency (no   ) Relaxed Consistency Locality  Caching Workflows  – Data Analysis 

8 Workload Characteristics Checkpointing HotStorage ‘09 AxesOptimizations Data life time (temporary   ) Application informed caching Read (Seq.   )Read-ahead Write (Seq.   ) Asynch. write Consistency (no    ) Relaxed Consistency Locality  Caching Compressibility  Similarity detection Workflows  Data Analysis  Checkpointing 

9 Workload Characteristics HotStorage ‘09 Workflows  Data Analysis  Checkpointing  AxesOptimizations Data life time (temporary   ) Application informed caching Read (Seq.   )Read-ahead Write (Seq.   ) Asynch. write Consistency (no    ) Relaxed Consistency Locality  Caching Compressibility  Similarity detection Security Tunable sec. levels

10 Opportunities  Specialization: Application specialized storage  Under utilized resources  Compute node storage space  Interconnect bandwidth HotStorage ‘09

11 Our Solution Versatile Storage System: Application specialized The Deployment Approach: Configured at deployment time Life time coupled with the target application Potential : Higher performance and scalability HotStorage ‘09

12 Versatile Storage System Architecture Manager (Metadata management) HotStorage ‘09 Access Module Storage Node Compute Node

13 Configurable / Extensible IO Pipeline HotStorage ‘09 Application IO Queue Dispatcher Buffer Manag. … Consistency Metadata Operations Content Addressability Data Security Communication Agent Application IO Queue Dispatcher Buffer Manag. Metadata Operations Access Module Storage Node

14 Configurable / Extensible IO Pipeline HotStorage ‘09 Application IO Queue Dispatcher Buffer Manag. … Consistency Metadata Operations Content Addressability Data Security Communication Agent Dispatcher … Consistency Content Addressability Data Security Communication Agent Access Module Storage Node

15 Configurable / Extensible Support HotStorage ‘09 Metadata Service API Dispatcher Request New Module Support … Application IO Queue Dispatcher Buffer Manag. … Metadata Operations NM Communication Agent Access Module Storage Node Manager Access Module Header Request data

16 Preliminary Evaluation – Real Application HotStorage ‘09 DOCK6 workflow: Overall: 1.52x Stages Read input, compute, and write temporary results Summarize, sort, and select Archive Versatile Storage Optimizations Cache the input data Cache temporary files Asynch. flush results to GPFS Results (8K processors) 1.06x 11.76x 1.51x

17 Summary HotStorage ‘09 Versatile Storage System Underutilized resources Application specialization The Deployment Approach: Configured at deployment time Coupled with the target application Potential: Higher performance and scalability

18 Not addressed – Future work HotStorage ‘09  Configurability / extensibility evaluation  Complete prototype  Evaluation with a diverse set of applications  Configuration  Application profiling  File system automated configuration

19 Thank you netsyslab.ece.ubc.ca