Disk-Locality in Datacenter Computing Considered Irrelevant Ganesh Ananthanarayanan, Ali Ghodsi, Scott Shenker, Ion Stoica 1.

Slides:



Advertisements
Similar presentations
DataGarage: Warehousing Massive Performance Data on Commodity Servers
Advertisements

Aggressive Cloning of Jobs for Effective Straggler Mitigation Ganesh Ananthanarayanan, Ali Ghodsi, Scott Shenker, Ion Stoica.
Effective Straggler Mitigation: Attack of the Clones Ganesh Ananthanarayanan, Ali Ghodsi, Srikanth Kandula, Scott Shenker, Ion Stoica.
The Datacenter Needs an Operating System Matei Zaharia, Benjamin Hindman, Andy Konwinski, Ali Ghodsi, Anthony Joseph, Randy Katz, Scott Shenker, Ion Stoica.
Locality-Aware Dynamic VM Reconfiguration on MapReduce Clouds Jongse Park, Daewoo Lee, Bokyeong Kim, Jaehyuk Huh, Seungryoul Maeng.
1 Degraded-First Scheduling for MapReduce in Erasure-Coded Storage Clusters Runhui Li, Patrick P. C. Lee, Yuchong Hu The Chinese University of Hong Kong.
THE DATACENTER NEEDS AN OPERATING SYSTEM MATEI ZAHARIA, BENJAMIN HINDMAN, ANDY KONWINSKI, ALI GHODSI, ANTHONY JOSEPH, RANDY KATZ, SCOTT SHENKER, ION STOICA.
Making Sense of Performance in Data Analytics Frameworks Kay Ousterhout, Ryan Rasti, Sylvia Ratnasamy, Scott Shenker, Byung-Gon Chun.
Making Sense of Spark Performance
CPSC 2031 What is a computer? A machine that processes information.
Mesos A Platform for Fine-Grained Resource Sharing in Data Centers Benjamin Hindman, Andy Konwinski, Matei Zaharia, Ali Ghodsi, Anthony D. Joseph, Randy.
Matei Zaharia, Dhruba Borthakur *, Joydeep Sen Sarma *, Khaled Elmeleegy +, Scott Shenker, Ion Stoica UC Berkeley, * Facebook Inc, + Yahoo! Research Delay.
Improving Proxy Cache Performance: Analysis of Three Replacement Policies Dilley, J.; Arlitt, M. A journal paper of IEEE Internet Computing, Volume: 3.
Making Sense of Performance in Data Analytics Frameworks Kay Ousterhout Joint work with Ryan Rasti, Sylvia Ratnasamy, Scott Shenker, Byung-Gon Chun UC.
Jennifer Rexford Princeton University MW 11:00am-12:20pm Data-Center Traffic Management COS 597E: Software Defined Networking.
The Power of Choice in Data-Aware Cluster Scheduling
The Case for Tiny Tasks in Compute Clusters Kay Ousterhout *, Aurojit Panda *, Joshua Rosen *, Shivaram Venkataraman *, Reynold Xin *, Sylvia Ratnasamy.
Outline | Motivation| Design | Results| Status| Future
1 CS : Technology Trends Ion Stoica ( September 12, 2011.
SYSTEMS SUPPORT FOR GRAPHICAL LEARNING Ken Birman 1 CS6410 Fall /18/2014.
Advanced Topics: MapReduce ECE 454 Computer Systems Programming Topics: Reductions Implemented in Distributed Frameworks Distributed Key-Value Stores Hadoop.
U.S. Department of the Interior U.S. Geological Survey David V. Hill, Information Dynamics, Contractor to USGS/EROS 12/08/2011 Satellite Image Processing.
Lecture 11: DMBS Internals
A Dynamic MapReduce Scheduler for Heterogeneous Workloads Chao Tian, Haojie Zhou, Yongqiang He,Li Zha 簡報人:碩資工一甲 董耀文.
SOFTWARE SYSTEMS DEVELOPMENT MAP-REDUCE, Hadoop, HBase.
Infrastructure for Better Quality Internet Access & Web Publishing without Increasing Bandwidth Prof. Chi Chi Hung School of Computing, National University.
Department of Computer Science at Florida State LFTI: A Performance Metric for Assessing Interconnect topology and routing design Background ‒ Innovations.
Hadoop Hardware Infrastructure considerations ©2013 OpalSoft Big Data.
Tachyon: memory-speed data sharing Haoyuan (HY) Li, Ali Ghodsi, Matei Zaharia, Scott Shenker, Ion Stoica Good morning everyone. My name is Haoyuan,
Bleeding edge technology to transform Data into Knowledge HADOOP In pioneer days they used oxen for heavy pulling, and when one ox couldn’t budge a log,
So far we have covered … Basic visualization algorithms Parallel polygon rendering Occlusion culling They all indirectly or directly help understanding.
Mesos A Platform for Fine-Grained Resource Sharing in the Data Center Benjamin Hindman, Andy Konwinski, Matei Zaharia, Ali Ghodsi, Anthony Joseph, Randy.
Dominant Resource Fairness: Fair Allocation of Multiple Resource Types Ali Ghodsi, Matei Zaharia, Benjamin Hindman, Andy Konwinski, Scott Shenker, Ion.
Reining in the Outliers in Map-Reduce Clusters using Mantri Ganesh Ananthanarayanan, Srikanth Kandula, Albert Greenberg, Ion Stoica, Yi Lu, Bikas Saha,
L7: Performance Frans Kaashoek Spring 2013.
1 Making MapReduce Scheduling Effective in Erasure-Coded Storage Clusters Runhui Li and Patrick P. C. Lee The Chinese University of Hong Kong LANMAN’15.
1 Virtual Memory Main memory can act as a cache for the secondary storage (disk) Advantages: –illusion of having more physical memory –program relocation.
DMBS Internals I. What Should a DBMS Do? Store large amounts of data Process queries efficiently Allow multiple users to access the database concurrently.
MROrder: Flexible Job Ordering Optimization for Online MapReduce Workloads School of Computer Engineering Nanyang Technological University 30 th Aug 2013.
A Platform for Fine-Grained Resource Sharing in the Data Center
Storage Systems CSE 598d, Spring 2007 Lecture ?: Rules of thumb in data engineering Paper by Jim Gray and Prashant Shenoy Feb 15, 2007.
DMBS Internals I. What Should a DBMS Do? Store large amounts of data Process queries efficiently Allow multiple users to access the database concurrently.
Presented by Qifan Pu With many slides from Ali’s NSDI talk Ali Ghodsi, Matei Zaharia, Benjamin Hindman, Andy Konwinski, Scott Shenker, Ion Stoica.
Dominant Resource Fairness: Fair Allocation of Multiple Resource Types Ali Ghodsi, Matei Zaharia, Benjamin Hindman, Andy Konwinski, Scott Shenker, Ion.
Data Centers and Cloud Computing 1. 2 Data Centers 3.
6.888 Lecture 8: Networking for Data Analytics Mohammad Alizadeh Spring  Many thanks to Mosharaf Chowdhury (Michigan) and Kay Ousterhout (Berkeley)
1 Lecture 16: Data Storage Wednesday, November 6, 2006.
1 Student Date Time Wei Li Nov 30, 2015 Monday 9:00-9:25am Shubbhi Taneja Nov 30, 2015 Monday9:25-9:50am Rodrigo Sanandan Dec 2, 2015 Wednesday9:00-9:25am.
PACMan: Coordinated Memory Caching for Parallel Jobs Ganesh Ananthanarayanan, Ali Ghodsi, Andrew Wang, Dhruba Borthakur, Srikanth Kandula, Scott Shenker,
1 Paolo Bianco Storage Architect Sun Microsystems An overview on Hybrid Storage Technologies.
Canadian Bioinformatics Workshops
Supercomputing versus Big Data processing — What's the difference?
Big Data Analytics with Parallel Jobs
Big Data is a Big Deal!.
BD-Cache: Big Data Caching for Datacenters
Chris Cai, Shayan Saeed, Indranil Gupta, Roy Campbell, Franck Le
BD-CACHE Big Data Caching for Datacenters
Improving Datacenter Performance and Robustness with Multipath TCP
Auburn University COMP7330/7336 Advanced Parallel and Distributed Computing MapReduce - Introduction Dr. Xiao Qin Auburn.
Trends: Technology Doubling Periods – storage: 12 mos, bandwidth: 9 mos, and (what law is this?) cpu compute capacity: 18 mos Then and Now Bandwidth 1985:
So far we have covered … Basic visualization algorithms
PA an Coordinated Memory Caching for Parallel Jobs
PL 7.2 The Times They are A-Changin’
Datacenter As a Computer
The University of Adelaide, School of Computer Science
湖南大学-信息科学与工程学院-计算机与科学系
COS 518: Advanced Computer Systems Lecture 14 Michael Freedman
Reining in the Outliers in MapReduce Jobs using Mantri
Fast, Interactive, Language-Integrated Cluster Computing
CSE 552 preparation for reading
Presentation transcript:

Disk-Locality in Datacenter Computing Considered Irrelevant Ganesh Ananthanarayanan, Ali Ghodsi, Scott Shenker, Ion Stoica 1

Data Intensive Computing Basis of analytics in modern Internet services ◦ Infrastructure of O(10,000) machines ◦ Peta-bytes of storage ◦ E.g., Google MapReduce [OSDI’04], Hadoop [Open Source], Dryad [EuroSys’07]… Job  {Phase}  {Task} 2

Disk Locality Tasks are I/O intensive Disk bandwidth >> Network bandwidth Co-locate tasks with their input 3 Solutions focus on disk-locality: Improve it [EuroSys’10, EuroSys’11] Fairness considerations [SOSP’09] Evaluation metric [NSDI’11] …

Datacenter trends indicate… 4

Fast Networks [1] Three-layer hierarchy, traditionally ◦ Access, Aggregate, Core switches Link rates are improving… ◦ Rack-local ~ Disk-local [Google, Facebook] 5 Hadoop logs from Facebook (Rack-local/Disk- local) In 85% of jobs, rack-local tasks are as fast as disk-local tasks Rate = (Data)/(Time)

Fast Networks [2] Over-subscription is fast reducing… Full bisection bandwidth topologies [SIGCOMM-’08, ‘09] Commodity switches  cost saving ($$$)  Adoption in today’s datacenters (Google?) 6

Storage Crunch [1] Data mining algorithms perform better when fed with more data ◦ Recommendations, advertisements etc. Storage is no longer plentiful [Facebook] ◦ Limits to growing the datacenter, non-linear if to move to a new datacenter ◦ Data is stored compressed 7

Storage Crunch [2] Data compression  less data to read 8 Hadoop logs from Facebook (Rack-local/Off- rack) Off-rack tasks only 1.4x slower (over- subscription of 10x) Rate = (Data)/(Time)

Disk Locality will be irrelevant! 1. Networks are getting faster, disks aren’t  Disks are the bottleneck 2. Storage is becoming a precious commodity  Data compression (  reads don’t dominate) 9

Run any task anywhere? Not so fast… Memory reads are two magnitudes faster Machines have memory of a few GB  Memory-locality is relevant! 10

Let’s build a memory cache Capacity is three orders less than disk ◦ 96% of jobs fit their data in memory 75% of blocks are singly-accessed ◦ But only 11% of jobs 11

Cache Hit Rates Memory-locality of 52% ◦ Aggregated memory model doesn’t buy much ◦ LFU is better than LRU 12 64% jobs have all their tasks memory-local

What next? Pre-fetching Blocks ◦ Out-of-band mechanisms Cache Eviction ◦ Preserve “whole” job inputs Effect of workload ◦ What if there aren’t so many small jobs? 13

Summary Disk-locality is not required anymore ◦ Networks are getting faster than disks ◦ Storage crunch  Data compression  Reduces read component Memory-locality should be the focus ◦ Data fits into memory for 96% jobs ◦ Encouraging early results 14

SSDs will not save us Unlikely to replace disks – Economics don’t work out ◦ Costs need to drop by ~3 orders, but are dropping by only 50% per year Ever-increasing storage demands will not be met by deploying SSDs 15