DON T CRY OVER SPILLED RECORDS Memory elasticity of data-parallel applications and its application to cluster scheduling
|
|
|
- Elwin Parrish
- 6 years ago
- Views:
Transcription
1 DON T CRY OVER SPILLED RECORDS Memory elasticity of data-parallel applications and its application to cluster scheduling Călin Iorgulescu (EPFL), Florin Dinu (EPFL), Aunn Raza (NUST Pakistan), Wajih Ul Hassan (UIUC), Willy Zwaenepoel (EPFL)
2 Cluster operators care about resource utilization Best bang for your buck! Maximize performance of data-parallel applications Idea: Efficient resource utilization through under-provisioning 7/12/17 École Polytechnique Fédérale de Lausanne 2 / 33
3 Cluster memory is under-utilized! Avg. mem. utilization: 78% 7/12/17 École Polytechnique Fédérale de Lausanne 3 / 33
4 Cluster memory is under-utilized! Leverage this idle memory! Avg. mem. utilization: 78% 7/12/17 École Polytechnique Fédérale de Lausanne 3 / 33
5 Impact of memory constraining applications Conventional wisdom: do not touch memory! Risks: crashes severe performance degradation (e.g., thrashing) Can we safely, deterministically, and with modest impact constrain memory? 7/12/17 École Polytechnique Fédérale de Lausanne 4 / 33
6 Context: batch jobs and their memory usage 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
7 Context: batch jobs and their memory usage Input 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
8 Context: batch jobs and their memory usage Input Ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
9 Context: batch jobs and their memory usage Input Ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
10 Context: batch jobs and their memory usage Input Ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
11 Context: batch jobs and their memory usage Input Disk I/O CPU Ideal memory Read Compute Write 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
12 Context: batch jobs and their memory usage Input Disk I/O CPU Ideal memory Ideal duration Read Compute Write 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
13 Context: batch jobs and their memory usage Input Disk I/O CPU Ideal memory Ideal duration Read Compute Write Non-ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
14 Context: batch jobs and their memory usage Input Disk I/O CPU Ideal memory Ideal duration Read Compute Write Non-ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
15 Context: batch jobs and their memory usage Input Disk I/O CPU Ideal memory Ideal duration Read Compute Write Non-ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
16 Context: batch jobs and their memory usage Input Disk I/O CPU Ideal memory Ideal duration Read Compute Write Non-ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
17 Context: batch jobs and their memory usage Input Disk I/O CPU Ideal memory Ideal duration Read Compute Write Non-ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
18 Context: batch jobs and their memory usage Input Disk I/O CPU Ideal memory Ideal duration Read Compute Write Non-ideal memory Read Spill Read Spill Read Compute Merge Write 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
19 Context: batch jobs and their memory usage Input Disk I/O CPU Ideal memory Ideal duration Penalty Read Compute Write Non-ideal memory Read Spill Read Spill Read Compute Merge Write 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
20 Context: batch jobs and their memory usage Input Disk I/O CPU Ideal memory Ideal duration Penalty Read Compute Write Batch jobs handle memory under-provisioning intermediate results spilled to disk Non-ideal memory Read Spill Read Spill Read Compute Merge Write 7/12/17 École Polytechnique Fédérale de Lausanne 5 / 33
21 MEMORY ELASTICITY 7/12/17 École Polytechnique Fédérale de Lausanne 6 / 33
22 What is Memory Elasticity? Non-ideal memory Disk I/O CPU Task Heap Task Heap 7/12/17 École Polytechnique Fédérale de Lausanne 7 / 33
23 What is Memory Elasticity? Non-ideal memory Disk I/O CPU Safely constrain memory Task Heap Task Heap 7/12/17 École Polytechnique Fédérale de Lausanne 7 / 33
24 What is Memory Elasticity? Non-ideal memory Disk I/O CPU Safely constrain memory Moderate penalties Task Heap Task Heap 7/12/17 École Polytechnique Fédérale de Lausanne 7 / 33
25 What is Memory Elasticity? Non-ideal memory Disk I/O CPU Safely constrain memory Moderate penalties Highly predictable Task Heap Task Heap 7/12/17 École Polytechnique Fédérale de Lausanne 7 / 33
26 What is Memory Elasticity? Non-ideal memory Disk I/O CPU Safely constrain memory Moderate penalties Highly predictable Task Heap Ubiquitous for most data-parallel apps Task Heap 7/12/17 École Polytechnique Fédérale de Lausanne 7 / 33
27 An empirical study of Memory Elasticity Analysis of 18 jobs across 8 different applications Constrain tasks memory measure penalty Bypass disk buffer cache (to not mask impact of spilling to disk) 7/12/17 École Polytechnique Fédérale de Lausanne 8 / 33
28 Questions about Memory Elasticity 7/12/17 École Polytechnique Fédérale de Lausanne 9 / 33
29 Questions about Memory Elasticity Are the penalties large? 7/12/17 École Polytechnique Fédérale de Lausanne 9 / 33
30 Questions about Memory Elasticity Are the penalties large? Do penalties vary considerably w.r.t given memory? 7/12/17 École Polytechnique Fédérale de Lausanne 9 / 33
31 Questions about Memory Elasticity Are the penalties large? Do penalties vary considerably w.r.t given memory? Does the additional I/O cause disk contention? 7/12/17 École Polytechnique Fédérale de Lausanne 9 / 33
32 Questions about Memory Elasticity Are the penalties large? Do penalties vary considerably w.r.t given memory? Does the additional I/O cause disk contention? NOT SO MUCH! 7/12/17 École Polytechnique Fédérale de Lausanne 9 / 33
33 Normalized task execution time Elasticity of Hadoop workloads: Reducers 10% 50% 90% Nutch Indexing Pagerank 1 Pagerank 2 TPC-DS Q40 TPC-DS Q7 Conn. Comp. 1 Word Count Terasort Recomm. 1 Recomm. 2 7/12/17 École Polytechnique Fédérale de Lausanne 10 / 33
34 Normalized task execution time Elasticity of Hadoop workloads: Reducers 10% 50% 90% Nutch Indexing Pagerank 1 Surprise! The median penalty is <1.6x! Pagerank 2 TPC-DS Q40 TPC-DS Q7 Conn. Comp. 1 Word Count Terasort Recomm x Recomm. 2 7/12/17 École Polytechnique Fédérale de Lausanne 10 / 33
35 Why are the penalties so modest? Data buffer most of app. memory Sequential disk access Logarithmic external merge algorithms Memory pressure absorbed by data buffer Spilling records to disk is faster than OS paging Merge steps required << disk spills 7/12/17 École Polytechnique Fédérale de Lausanne 11 / 33
36 Normalized task execution time Elasticity of Hadoop workloads: Reducers 10% 50% 90% Nutch Indexing Pagerank 1 Pagerank 2 TPC-DS Q40 TPC-DS Q7 Conn. Comp. 1 Word Count Terasort Recomm x Recomm. 2 7/12/17 École Polytechnique Fédérale de Lausanne 12 / 33
37 Normalized task execution time Elasticity of Hadoop workloads: Reducers Nutch Indexing Pagerank 1 Pagerank 2 TPC-DS Q40 10% 50% 90% Surprise! For 10%, 50%, and 90% memory, penalties vary by at most 0.25x! TPC-DS Q7 Conn. Comp. 1 Word Count Terasort Recomm x Recomm. 2 7/12/17 École Polytechnique Fédérale de Lausanne 12 / 33
38 Why do penalties vary so little w/ memory? Static spilling threshold comparable data spilling for 90% and 10% of memory 7/12/17 École Polytechnique Fédérale de Lausanne 13 / 33
39 Why do penalties vary so little w/ memory? Static spilling threshold comparable data spilling for 90% and 10% of memory Input 2.1 GB Input 2.1 GB 7/12/17 École Polytechnique Fédérale de Lausanne 13 / 33
40 Why do penalties vary so little w/ memory? Static spilling threshold comparable data spilling for 90% and 10% of memory Input 2.1 GB Input 2.1 GB Buffer 2 GB 7/12/17 École Polytechnique Fédérale de Lausanne 13 / 33
41 Why do penalties vary so little w/ memory? Static spilling threshold comparable data spilling for 90% and 10% of memory Input 2.1 GB Input 2.1 GB Buffer Spills 2 GB 1 x 2 GB 7/12/17 École Polytechnique Fédérale de Lausanne 13 / 33
42 Why do penalties vary so little w/ memory? Static spilling threshold comparable data spilling for 90% and 10% of memory Input 2.1 GB Input 2.1 GB Buffer 2 GB Buffer 200 MB Spills 1 x 2 GB 7/12/17 École Polytechnique Fédérale de Lausanne 13 / 33
43 Why do penalties vary so little w/ memory? Static spilling threshold comparable data spilling for 90% and 10% of memory Input 2.1 GB Input 2.1 GB Buffer 2 GB Buffer 200 MB Spills 1 x 2 GB Spills 10 x 200 MB 7/12/17 École Polytechnique Fédérale de Lausanne 13 / 33
44 Why do penalties vary so little w/ memory? Static spilling threshold comparable data spilling for 90% and 10% of memory Input 2.1 GB Input 2.1 GB Buffer 2 GB Buffer 200 MB Spills 1 x 2 GB Spills 10 x 200 MB Total data spilled in both cases: 2GB 7/12/17 École Polytechnique Fédérale de Lausanne 13 / 33
45 Normalized task execution time Elasticity of Hadoop workloads: Mappers WordCount w/ combiner Conn. Comp. 1 10% 50% 90% Pagerank 2 Pagerank 1 WordCount 7/12/17 École Polytechnique Fédérale de Lausanne 14 / 33
46 Normalized task execution time Elasticity of Hadoop workloads: Mappers WordCount w/ combiner Conn. Comp. 1 10% 50% 90% Penalties are even lower! Median penalty is ~1.2x! Pagerank 2 Pagerank 1 WordCount 1.38x 7/12/17 École Polytechnique Fédérale de Lausanne 14 / 33
47 Normalized task execution time Elasticity of Hadoop workloads: Mappers WordCount w/ combiner Conn. Comp. 1 10% 50% 90% Penalties are even lower! Median penalty is ~1.2x! Penalties for 10%, 50%, and 90% memory vary by at most 0.05x! Pagerank 2 Pagerank 1 WordCount 1.38x 7/12/17 École Polytechnique Fédérale de Lausanne 14 / 33
48 Normalized task execution time Elasticity of Spark and Tez workloads % 50% 90% Spark Terasort Spark WordCount Tez SortMergeJoin Tez WordCount 7/12/17 École Polytechnique Fédérale de Lausanne 15 / 33
49 Normalized task execution time Elasticity of Spark and Tez workloads % 50% 90% Median penalty is ~1.75x! 2.4x 1 Spark Terasort Spark WordCount Tez SortMergeJoin Tez WordCount 7/12/17 École Polytechnique Fédérale de Lausanne 15 / 33
50 Normalized task execution time Elasticity of Spark and Tez workloads % 50% 90% Median penalty is ~1.75x! 2.4x Penalties for 10%, 50%, and 90% memory vary by at most 0.3x! Spark Terasort Spark WordCount Tez SortMergeJoin Tez WordCount 7/12/17 École Polytechnique Fédérale de Lausanne 15 / 33
51 Does the additional I/O cause disk contention? Measure slowdown of elastic tasks on same machine spilling to the same disk Number of concurrent reducers 7/12/17 École Polytechnique Fédérale de Lausanne 16 / 33
52 Does the additional I/O cause disk contention? Measure slowdown of elastic tasks on same machine spilling to the same disk Cluster operators usually provision 1 disk / 2 cores* <10% slowdown! * Facebook (2010) and Nutanix Number of concurrent reducers 7/12/17 École Polytechnique Fédérale de Lausanne 16 / 33
53 Does the additional I/O cause disk contention? Measure slowdown of elastic tasks on same machine spilling to the same disk Cluster operators usually provision 1 disk / 2 cores* <10% slowdown! Degradation <25% for up to 1 disk / 5 cores! * Facebook (2010) and Nutanix Number of concurrent reducers 7/12/17 École Polytechnique Fédérale de Lausanne 16 / 33
54 Summary: Memory Elasticity of real workloads Modest performance penalties (<1.6x median) Similar penalties for 10% and 90% of ideal memory Disk contention negligible for existing clusters setup (<10%) 7/12/17 École Polytechnique Fédérale de Lausanne 17 / 33
55 MODELING MEMORY ELASTICITY 7/12/17 École Polytechnique Fédérale de Lausanne 18 / 33
56 Normalized task execution time Modeling Memory Elasticity How does penalty vary for a task? % 20% 30% 40% 50% 60% 70% 80% 90% 100% 110% Percentage of ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 19 / 33
57 Normalized task execution time Modeling Memory Elasticity How does penalty vary for a task? Penalty 2 = f ( disk speed, input size, framework configuration ) % 20% 30% 40% 50% 60% 70% 80% 90% 100% 110% Percentage of ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 19 / 33
58 Normalized task execution time Modeling Memory Elasticity Penalties vary little between percentages step model % 20% 30% 40% 50% 60% 70% 80% 90% 100% 110% Percentage of ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 20 / 33
59 Normalized task execution time Modeling Memory Elasticity Penalties vary little between percentages step model Disk I/ O overhead % 20% 30% 40% 50% 60% 70% 80% 90% 100% 110% Percentage of ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 20 / 33
60 Normalized task execution time Modeling Memory Elasticity Requires 2 profiling runs infers all other points % 20% 30% 40% 50% 60% 70% 80% 90% 100% 110% Percentage of ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 21 / 33
61 Normalized task execution time Modeling Memory Elasticity Requires 2 profiling runs infers all other points Non-ideal memory % 20% 30% 40% 50% 60% 70% 80% 90% 100% 110% Percentage of ideal memory Ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 21 / 33
62 Normalized task execution time Modeling Memory Elasticity Requires 2 profiling runs infers all other points Non-ideal memory The step model represents our baseline. More complex models are possible. 0 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 110% Percentage of ideal memory Ideal memory 7/12/17 École Polytechnique Fédérale de Lausanne 21 / 33
63 Real duration normalized to model Accuracy of our reducer model 1.15 Avg. offset Pagerank 1 Pagerank 2 TPC-DS Q7 TPC-DS Q40 Conn. Comp. 1 Terasort Recomm. 1 Recomm. 2 Word Count Nutch Indexing Spark Terasort Tez SMJ Tez Word Count 7/12/17 École Polytechnique Fédérale de Lausanne 22 / 33
64 Real duration normalized to model Accuracy of our reducer model 1.15 Avg. offset 1.1 Most reducers are off at most by +/- 5%! Pagerank 1 Pagerank 2 TPC-DS Q7 TPC-DS Q40 Conn. Comp. 1 Terasort Recomm. 1 Recomm. 2 Word Count Nutch Indexing Spark Terasort Tez SMJ Tez Word Count 7/12/17 École Polytechnique Fédérale de Lausanne 22 / 33
65 Summary: Modeling memory elasticity Step model is adequate (more complex models available) Only 2 profiling points full model Models are very robust (+/- 5% error for most reducers) 7/12/17 École Polytechnique Fédérale de Lausanne 23 / 33
66 LEVERAGING MEMORY ELASTICITY IN CLUSTER SCHEDULING 7/12/17 École Polytechnique Fédérale de Lausanne 24 / 33
67 memory memory How can a scheduler reason about Memory Elasticity? Trade-off between task queueing time task execution time Elastic allocation time time 7/12/17 École Polytechnique Fédérale de Lausanne 25 / 33
68 memory memory memory YARN-ME: Decision process Make an elastic allocation iff it does not exceed the expected job completion time YARN (no ME) time YARN-ME time 2 YARN-ME 1 3 time 7/12/17 École Polytechnique Fédérale de Lausanne 26 / 33
69 YARN-ME: Design and components Resource Manager YARN-ME Timeline generator Components Timeline generator computes expected JCTs Profiler generates the model metadata Ideal duration Ideal memory Penalty Job ME metadata Profiler 7/12/17 École Polytechnique Fédérale de Lausanne 27 / 33
70 Memory utilization analysis for YARN-ME 50 node cluster Homogeneous trace: 5x Pagerank jobs 7/12/17 École Polytechnique Fédérale de Lausanne 28 / 33
71 Memory utilization analysis for YARN-ME 50 node cluster Homogeneous trace: 5x Pagerank jobs Memory utilization increased to 95% 7/12/17 École Polytechnique Fédérale de Lausanne 28 / 33
72 YARN-ME improvement over YARN (%) What gains can YARN-ME achieve for heterogeneous workloads? JCT Makespan MAP REDUCE node cluster Mixed trace 14 jobs 3x PageRank 3x Recommender 8x Wordcount 10 0 Total Pagerank Recommender WordCount 7/12/17 École Polytechnique Fédérale de Lausanne 29 / 33
73 YARN-ME improvement over YARN (%) What gains can YARN-ME achieve for heterogeneous workloads? JCT Makespan MAP REDUCE Up to 65% improvement for JCT and makespan node cluster Mixed trace 14 jobs 3x PageRank 3x Recommender 8x Wordcount 10 0 Total Pagerank Recommender WordCount 7/12/17 École Polytechnique Fédérale de Lausanne 29 / 33
74 YARN-ME improvement over YARN (%) What gains can YARN-ME achieve for homogeneous workloads? JCT Makespan MAP REDUCE node cluster Pagerank concurrent runs /12/17 École Polytechnique Fédérale de Lausanne 30 / 33
75 YARN-ME improvement over YARN (%) What gains can YARN-ME achieve for homogeneous workloads? JCT Makespan MAP REDUCE Up to 40% improvement for JCT and makespan node cluster Pagerank concurrent runs /12/17 École Polytechnique Fédérale de Lausanne 30 / 33
76 Trace-driven simulation of YARN-ME We built DSS (the Discrete Scheduler Simulator) and it is open-source! Trace parameter sweep > 8,000 traces up to 3,000 nodes results comparable to real workloads Robustness analysis > 20,000 traces YARN-ME is robust to model mis-estimations 7/12/17 École Polytechnique Fédérale de Lausanne 31 / 33
77 Related work Efficient packing better resource utilization Tetris [SIGCOMM 14], GRAPHENE [OSDI 16] Collocate batch-jobs with latency-critical services Heracles [ISCA 15] Resource over-committing Apollo [OSDI 14], Borg [EuroSys 15] Suspend tasks under memory pressure ITask [SOSP 15] 7/12/17 École Polytechnique Fédérale de Lausanne 32 / 33
78 Conclusion: Don t cry over spilled records! Memory Elasticity highly predictable, low penalty Memory Elasticity in scheduling trade task queueing-time for running-time YARN-ME up to 60% improvement in average JCT DSS code available: 7/12/17 École Polytechnique Fédérale de Lausanne 33 / 33
Don t cry over spilled records: Memory elasticity of data-parallel applications and its application to cluster scheduling
Don t cry over spilled records: Memory elasticity of data-parallel applications and its application to cluster scheduling Călin Iorgulescu and Florin Dinu, EPFL; Aunn Raza, NUST Pakistan; Wajih Ul Hassan,
Don t cry over spilled records: Memory elasticity of data-parallel applications and its application to cluster scheduling
Don t cry over spilled records: Memory elasticity of data-parallel applications and its application to cluster scheduling Calin Iorgulescu *, Florin Dinu *, Aunn Raza, Wajih Ul Hassan, and Willy Zwaenepoel
Preemptive, Low Latency Datacenter Scheduling via Lightweight Virtualization
Preemptive, Low Latency Datacenter Scheduling via Lightweight Virtualization Wei Chen, Jia Rao*, and Xiaobo Zhou University of Colorado, Colorado Springs * University of Texas at Arlington Data Center
Job sample: SCOPE (VLDBJ, 2012)
Apollo High level SQL-Like language The job query plan is represented as a DAG Tasks are the basic unit of computation Tasks are grouped in Stages Execution is driven by a scheduler Job sample: SCOPE (VLDBJ,
Packing Tasks with Dependencies. Robert Grandl, Srikanth Kandula, Sriram Rao, Aditya Akella, Janardhan Kulkarni
Packing Tasks with Dependencies Robert Grandl, Srikanth Kandula, Sriram Rao, Aditya Akella, Janardhan Kulkarni The Cluster Scheduling Problem Jobs Goal: match tasks to resources Tasks 2 The Cluster Scheduling
Infiniswap. Efficient Memory Disaggregation. Mosharaf Chowdhury. with Juncheng Gu, Youngmoon Lee, Yiwen Zhang, and Kang G. Shin
Infiniswap Efficient Memory Disaggregation Mosharaf Chowdhury with Juncheng Gu, Youngmoon Lee, Yiwen Zhang, and Kang G. Shin Rack-Scale Computing Datacenter-Scale Computing Geo-Distributed Computing Coflow
Analysis in the Big Data Era
Analysis in the Big Data Era Massive Data Data Analysis Insight Key to Success = Timely and Cost-Effective Analysis 2 Hadoop MapReduce Ecosystem Popular solution to Big Data Analytics Java / C++ / R /
Apache HAWQ (incubating)
HADOOP NATIVE SQL What is HAWQ? Apache HAWQ (incubating) Is an elastic parallel processing SQL engine that runs native in Apache Hadoop to directly access data for advanced analytics. Why HAWQ? Hadoop
Using Transparent Compression to Improve SSD-based I/O Caches
Using Transparent Compression to Improve SSD-based I/O Caches Thanos Makatos, Yannis Klonatos, Manolis Marazakis, Michail D. Flouris, and Angelos Bilas {mcatos,klonatos,maraz,flouris,bilas}@ics.forth.gr
Big data systems 12/8/17
Big data systems 12/8/17 Today Basic architecture Two levels of scheduling Spark overview Basic architecture Cluster Manager Cluster Cluster Manager 64GB RAM 32 cores 64GB RAM 32 cores 64GB RAM 32 cores
Improving the MapReduce Big Data Processing Framework
Improving the MapReduce Big Data Processing Framework Gistau, Reza Akbarinia, Patrick Valduriez INRIA & LIRMM, Montpellier, France In collaboration with Divyakant Agrawal, UCSB Esther Pacitti, UM2, LIRMM
Tuning Enterprise Information Catalog Performance
Tuning Enterprise Information Catalog Performance Copyright Informatica LLC 2015, 2018. Informatica and the Informatica logo are trademarks or registered trademarks of Informatica LLC in the United States
Announcements. Reading. Project #1 due in 1 week at 5:00 pm Scheduling Chapter 6 (6 th ed) or Chapter 5 (8 th ed) CMSC 412 S14 (lect 5)
Announcements Reading Project #1 due in 1 week at 5:00 pm Scheduling Chapter 6 (6 th ed) or Chapter 5 (8 th ed) 1 Relationship between Kernel mod and User Mode User Process Kernel System Calls User Process
SRM-Buffer: An OS Buffer Management SRM-Buffer: An OS Buffer Management Technique toprevent Last Level Cache from Thrashing in Multicores
SRM-Buffer: An OS Buffer Management SRM-Buffer: An OS Buffer Management Technique toprevent Last Level Cache from Thrashing in Multicores Xiaoning Ding The Ohio State University dingxn@cse.ohiostate.edu
Varys. Efficient Coflow Scheduling. Mosharaf Chowdhury, Yuan Zhong, Ion Stoica. UC Berkeley
Varys Efficient Coflow Scheduling Mosharaf Chowdhury, Yuan Zhong, Ion Stoica UC Berkeley Communication is Crucial Performance Facebook analytics jobs spend 33% of their runtime in communication 1 As in-memory
BigDataBench-MT: Multi-tenancy version of BigDataBench
BigDataBench-MT: Multi-tenancy version of BigDataBench Gang Lu Beijing Academy of Frontier Science and Technology BigDataBench Tutorial, ASPLOS 2016 Atlanta, GA, USA n Software perspective Multi-tenancy
The Google File System
The Google File System Sanjay Ghemawat, Howard Gobioff, and Shun-Tak Leung SOSP 2003 presented by Kun Suo Outline GFS Background, Concepts and Key words Example of GFS Operations Some optimizations in
CS Project Report
CS7960 - Project Report Kshitij Sudan kshitij@cs.utah.edu 1 Introduction With the growth in services provided over the Internet, the amount of data processing required has grown tremendously. To satisfy
MATE-EC2: A Middleware for Processing Data with Amazon Web Services
MATE-EC2: A Middleware for Processing Data with Amazon Web Services Tekin Bicer David Chiu* and Gagan Agrawal Department of Compute Science and Engineering Ohio State University * School of Engineering
Farewell to Servers: Resource Disaggregation
Farewell to Servers: Hardware, Software, and Network Approaches towards Datacenter Resource Disaggregation Yiying Zhang 2 Monolithic Computer OS / Hypervisor 3 Can monolithic Application Hardware servers
Hadoop Beyond Batch: Real-time Workloads, SQL-on- Hadoop, and thevirtual EDW Headline Goes Here
Hadoop Beyond Batch: Real-time Workloads, SQL-on- Hadoop, and thevirtual EDW Headline Goes Here Marcel Kornacker marcel@cloudera.com Speaker Name or Subhead Goes Here 2013-11-12 Copyright 2013 Cloudera
Recovering Disk Storage Metrics from low level Trace events
Recovering Disk Storage Metrics from low level Trace events Progress Report Meeting May 05, 2016 Houssem Daoud Michel Dagenais École Polytechnique de Montréal Laboratoire DORSAL Agenda Introduction and
An Exploration of Designing a Hybrid Scale-Up/Out Hadoop Architecture Based on Performance Measurements
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI.9/TPDS.6.7, IEEE
Tuning Intelligent Data Lake Performance
Tuning Intelligent Data Lake Performance 2016 Informatica LLC. No part of this document may be reproduced or transmitted in any form, by any means (electronic, photocopying, recording or otherwise) without
Pocket: Elastic Ephemeral Storage for Serverless Analytics
Pocket: Elastic Ephemeral Storage for Serverless Analytics Ana Klimovic*, Yawen Wang*, Patrick Stuedi +, Animesh Trivedi +, Jonas Pfefferle +, Christos Kozyrakis* *Stanford University, + IBM Research 1
The Elasticity and Plasticity in Semi-Containerized Colocating Cloud Workload: a view from Alibaba Trace
The Elasticity and Plasticity in Semi-Containerized Colocating Cloud Workload: a view from Alibaba Trace Qixiao Liu* and Zhibin Yu Shenzhen Institute of Advanced Technology Chinese Academy of Science @SoCC
Network Requirements for Resource Disaggregation
Network Requirements for Resource Disaggregation Peter Gao (Berkeley), Akshay Narayan (MIT), Sagar Karandikar (Berkeley), Joao Carreira (Berkeley), Sangjin Han (Berkeley), Rachit Agarwal (Cornell), Sylvia
ibench: Quantifying Interference in Datacenter Applications
ibench: Quantifying Interference in Datacenter Applications Christina Delimitrou and Christos Kozyrakis Stanford University IISWC September 23 th 2013 Executive Summary Problem: Increasing utilization
Key aspects of cloud computing. Towards fuller utilization. Two main sources of resource demand. Cluster Scheduling
Key aspects of cloud computing Cluster Scheduling 1. Illusion of infinite computing resources available on demand, eliminating need for up-front provisioning. The elimination of an up-front commitment
CLOUD-SCALE FILE SYSTEMS
Data Management in the Cloud CLOUD-SCALE FILE SYSTEMS 92 Google File System (GFS) Designing a file system for the Cloud design assumptions design choices Architecture GFS Master GFS Chunkservers GFS Clients
EECS 482 Introduction to Operating Systems
EECS 482 Introduction to Operating Systems Winter 2018 Baris Kasikci Slides by: Harsha V. Madhyastha OS Abstractions Applications Threads File system Virtual memory Operating System Next few lectures:
Consolidating Complementary VMs with Spatial/Temporalawareness
Consolidating Complementary VMs with Spatial/Temporalawareness in Cloud Datacenters Liuhua Chen and Haiying Shen Dept. of Electrical and Computer Engineering Clemson University, SC, USA 1 Outline Introduction
Multi-tenancy version of BigDataBench
Multi-tenancy version of BigDataBench Gang Lu Institute of Computing Technology, Chinese Academy of Sciences BigDataBench Tutorial MICRO 2014 Cambridge, UK INSTITUTE OF COMPUTING TECHNOLOGY 1 Multi-tenancy
Improving DRAM Performance by Parallelizing Refreshes with Accesses
Improving DRAM Performance by Parallelizing Refreshes with Accesses Kevin Chang Donghyuk Lee, Zeshan Chishti, Alaa Alameldeen, Chris Wilkerson, Yoongu Kim, Onur Mutlu Executive Summary DRAM refresh interferes
Dynamic Resource Allocation for Distributed Dataflows. Lauritz Thamsen Technische Universität Berlin
Dynamic Resource Allocation for Distributed Dataflows Lauritz Thamsen Technische Universität Berlin 04.05.2018 Distributed Dataflows E.g. MapReduce, SCOPE, Spark, and Flink Used for scalable processing
HiTune. Dataflow-Based Performance Analysis for Big Data Cloud
HiTune Dataflow-Based Performance Analysis for Big Data Cloud Jinquan (Jason) Dai, Jie Huang, Shengsheng Huang, Bo Huang, Yan Liu Intel Asia-Pacific Research and Development Ltd Shanghai, China, 200241
Flash Storage Complementing a Data Lake for Real-Time Insight
Flash Storage Complementing a Data Lake for Real-Time Insight Dr. Sanhita Sarkar Global Director, Analytics Software Development August 7, 2018 Agenda 1 2 3 4 5 Delivering insight along the entire spectrum
SCALABLE DISTRIBUTED DEEP LEARNING
SEOUL Oct.7, 2016 SCALABLE DISTRIBUTED DEEP LEARNING Han Hee Song, PhD Soft On Net 10/7/2016 BATCH PROCESSING FRAMEWORKS FOR DL Data parallelism provides efficient big data processing: data collecting,
Extreme Storage Performance with exflash DIMM and AMPS
Extreme Storage Performance with exflash DIMM and AMPS 214 by 6East Technologies, Inc. and Lenovo Corporation All trademarks or registered trademarks mentioned here are the property of their respective
Nutanix Tech Note. Virtualizing Microsoft Applications on Web-Scale Infrastructure
Nutanix Tech Note Virtualizing Microsoft Applications on Web-Scale Infrastructure The increase in virtualization of critical applications has brought significant attention to compute and storage infrastructure.
Hadoop MapReduce Framework
Hadoop MapReduce Framework Contents Hadoop MapReduce Framework Architecture Interaction Diagram of MapReduce Framework (Hadoop 1.0) Interaction Diagram of MapReduce Framework (Hadoop 2.0) Hadoop MapReduce
Enabling NVMe I/O Scale
Enabling NVMe I/O Determinism @ Scale Chris Petersen, Hardware System Technologist Wei Zhang, Software Engineer Alexei Naberezhnov, Software Engineer Facebook Facebook @ Scale 800 Million 1.3 Billion 2.2
SRM-Buffer: An OS Buffer Management Technique to Prevent Last Level Cache from Thrashing in Multicores
SRM-Buffer: An OS Buffer Management Technique to Prevent Last Level Cache from Thrashing in Multicores Xiaoning Ding et al. EuroSys 09 Presented by Kaige Yan 1 Introduction Background SRM buffer design
Experimental Study of Virtual Machine Migration in Support of Reservation of Cluster Resources
Experimental Study of Virtual Machine Migration in Support of Reservation of Cluster Resources Ming Zhao, Renato J. Figueiredo Advanced Computing and Information Systems (ACIS) Electrical and Computer
Coflow. Recent Advances and What s Next? Mosharaf Chowdhury. University of Michigan
Coflow Recent Advances and What s Next? Mosharaf Chowdhury University of Michigan Rack-Scale Computing Datacenter-Scale Computing Geo-Distributed Computing Coflow Networking Open Source Apache Spark Open
LITE Kernel RDMA. Support for Datacenter Applications. Shin-Yeh Tsai, Yiying Zhang
LITE Kernel RDMA Support for Datacenter Applications Shin-Yeh Tsai, Yiying Zhang Time 2 Berkeley Socket Userspace Kernel Hardware Time 1983 2 Berkeley Socket TCP Offload engine Arrakis & mtcp IX Userspace
SANDPIPER: BLACK-BOX AND GRAY-BOX STRATEGIES FOR VIRTUAL MACHINE MIGRATION
SANDPIPER: BLACK-BOX AND GRAY-BOX STRATEGIES FOR VIRTUAL MACHINE MIGRATION Timothy Wood, Prashant Shenoy, Arun Venkataramani, and Mazin Yousif * University of Massachusetts Amherst * Intel, Portland Data
BIG DATA AND HADOOP ON THE ZFS STORAGE APPLIANCE
BIG DATA AND HADOOP ON THE ZFS STORAGE APPLIANCE BRETT WENINGER, MANAGING DIRECTOR 10/21/2014 ADURANT APPROACH TO BIG DATA Align to Un/Semi-structured Data Instead of Big Scale out will become Big Greatest
CEC 450 Real-Time Systems
CEC 450 Real-Time Systems Lecture 6 Accounting for I/O Latency September 28, 2015 Sam Siewert A Service Release and Response C i WCET Input/Output Latency Interference Time Response Time = Time Actuation
What s New in VMware vsphere 4.1 Performance. VMware vsphere 4.1
What s New in VMware vsphere 4.1 Performance VMware vsphere 4.1 T E C H N I C A L W H I T E P A P E R Table of Contents Scalability enhancements....................................................................
SharePoint 2010 Technical Case Study: Microsoft SharePoint Server 2010 Social Environment
SharePoint 2010 Technical Case Study: Microsoft SharePoint Server 2010 Social Environment This document is provided as-is. Information and views expressed in this document, including URL and other Internet
Locality-Aware Dynamic VM Reconfiguration on MapReduce Clouds. Jongse Park, Daewoo Lee, Bokyeong Kim, Jaehyuk Huh, Seungryoul Maeng
Locality-Aware Dynamic VM Reconfiguration on MapReduce Clouds Jongse Park, Daewoo Lee, Bokyeong Kim, Jaehyuk Huh, Seungryoul Maeng Virtual Clusters on Cloud } Private cluster on public cloud } Distributed
Data Processing at the Speed of 100 Gbps using Apache Crail. Patrick Stuedi IBM Research
Data Processing at the Speed of 100 Gbps using Apache Crail Patrick Stuedi IBM Research The CRAIL Project: Overview Data Processing Framework (e.g., Spark, TensorFlow, λ Compute) Spark-IO Albis Pocket
Native-Task Performance Test Report
Native-Task Performance Test Report Intel Software Wang, Huafeng, Huafeng.wang@intel.com Zhong, Xiang, xiang.zhong@intel.com Intel Software Page 1 1. Background 2. Related Work 3. Preliminary Experiments
Building A Data-center Scale Analytics Platform. Sriram Rao Scientist/Manager, CISL
Building A Data-center Scale Analytics Platform Sriram Rao Scientist/Manager, CISL CISL: Cloud and Information Services Lab Started in May 2012 Mission Statement: Applied research lab working on Systems
Bring x3 Spark Performance Improvement with PCIe SSD. Yucai, Yu BDT/STO/SSG January, 2016
Bring x3 Spark Performance Improvement with PCIe SSD Yucai, Yu (yucai.yu@intel.com) BDT/STO/SSG January, 2016 About me/us Me: Spark contributor, previous on virtualization, storage, mobile/iot OS. Intel
StoreApp: Shared Storage Appliance for Efficient and Scalable Virtualized Hadoop Clusters
: Shared Storage Appliance for Efficient and Scalable Virtualized Clusters Yanfei Guo, Jia Rao, Dazhao Cheng Dept. of Computer Science University of Colorado at Colorado Springs, USA yguo@uccs.edu Changjun
SEDA: An Architecture for Well-Conditioned, Scalable Internet Services
SEDA: An Architecture for Well-Conditioned, Scalable Internet Services Matt Welsh, David Culler, and Eric Brewer Computer Science Division University of California, Berkeley Operating Systems Principles
davidklee.net gplus.to/kleegeek linked.com/a/davidaklee
@kleegeek davidklee.net gplus.to/kleegeek linked.com/a/davidaklee Specialties / Focus Areas / Passions: Performance Tuning & Troubleshooting Virtualization Cloud Enablement Infrastructure Architecture
The Google File System
October 13, 2010 Based on: S. Ghemawat, H. Gobioff, and S.-T. Leung: The Google file system, in Proceedings ACM SOSP 2003, Lake George, NY, USA, October 2003. 1 Assumptions Interface Architecture Single
CCA-410. Cloudera. Cloudera Certified Administrator for Apache Hadoop (CCAH)
Cloudera CCA-410 Cloudera Certified Administrator for Apache Hadoop (CCAH) Download Full Version : http://killexams.com/pass4sure/exam-detail/cca-410 Reference: CONFIGURATION PARAMETERS DFS.BLOCK.SIZE
ZBD: Using Transparent Compression at the Block Level to Increase Storage Space Efficiency
ZBD: Using Transparent Compression at the Block Level to Increase Storage Space Efficiency Thanos Makatos, Yannis Klonatos, Manolis Marazakis, Michail D. Flouris, and Angelos Bilas {mcatos,klonatos,maraz,flouris,bilas}@ics.forth.gr
Farewell to Servers: Hardware, Software, and Network Approaches towards Datacenter Resource Disaggregation
Farewell to Servers: Hardware, Software, and Network Approaches towards Datacenter Resource Disaggregation Yiying Zhang Datacenter 3 Monolithic Computer OS / Hypervisor 4 Can monolithic Application Hardware
History-Based Harvesting of Spare Cycles and Storage in Large-Scale Datacenters
History-Based Harvesting of Spare Cycles and Storage in Large-Scale Datacenters Yunqi Zhang, George Prekas, Giovanni Matteo Fumarola, Marcus Fontoura, Íñigo Goiri, Ricardo Bianchini Datacenters are underutilized
Enabling Cost-effective Data Processing with Smart SSD
Enabling Cost-effective Data Processing with Smart SSD Yangwook Kang, UC Santa Cruz Yang-suk Kee, Samsung Semiconductor Ethan L. Miller, UC Santa Cruz Chanik Park, Samsung Electronics Efficient Use of
YARN: A Resource Manager for Analytic Platform Tsuyoshi Ozawa
YARN: A Resource Manager for Analytic Platform Tsuyoshi Ozawa ozawa.tsuyoshi@lab.ntt.co.jp ozawa@apache.org About me Tsuyoshi Ozawa Research Engineer @ NTT Twitter: @oza_x86_64 Over 150 reviews in 2015
PowerVault MD3 SSD Cache Overview
PowerVault MD3 SSD Cache Overview A Dell Technical White Paper Dell Storage Engineering October 2015 A Dell Technical White Paper TECHNICAL INACCURACIES. THE CONTENT IS PROVIDED AS IS, WITHOUT EXPRESS
OPERATING SYSTEM. Chapter 9: Virtual Memory
OPERATING SYSTEM Chapter 9: Virtual Memory Chapter 9: Virtual Memory Background Demand Paging Copy-on-Write Page Replacement Allocation of Frames Thrashing Memory-Mapped Files Allocating Kernel Memory
Clash of the Titans: MapReduce vs. Spark for Large Scale Data Analytics
Clash of the Titans: MapReduce vs. Spark for Large Scale Data Analytics Presented by: Dishant Mittal Authors: Juwei Shi, Yunjie Qiu, Umar Firooq Minhas, Lemei Jiao, Chen Wang, Berthold Reinwald and Fatma
TIMELINE SCHEDULING FOR OUT-OF-CORE RAY BATCHING. SGVR Lab KAIST
TIMELINE SCHEDLIN FOR OT-OF-CORE RAY BATCHIN Myungbae Son SVR Lab KAIST Sung-EuiYoon Our Scenario Complex scenes Out-of-core model: Too big data! Cannot be stored in main / memory Complex device configurations
TRASH DAY: COORDINATING GARBAGE COLLECTION IN DISTRIBUTED SYSTEMS
TRASH DAY: COORDINATING GARBAGE COLLECTION IN DISTRIBUTED SYSTEMS Martin Maas* Tim Harris KrsteAsanovic* John Kubiatowicz* *University of California, Berkeley Oracle Labs, Cambridge Why you should care
CS425 Computer Systems Architecture
CS425 Computer Systems Architecture Fall 2017 Thread Level Parallelism (TLP) CS425 - Vassilis Papaefstathiou 1 Multiple Issue CPI = CPI IDEAL + Stalls STRUC + Stalls RAW + Stalls WAR + Stalls WAW + Stalls
Exploiting Inter-Flow Relationship for Coflow Placement in Data Centers. Xin Sunny Huang, T. S. Eugene Ng Rice University
Exploiting Inter-Flow Relationship for Coflow Placement in Data Centers Xin Sunny Huang, T S Eugene g Rice University This Work Optimizing Coflow performance has many benefits such as avoiding application
Moneta: A High-performance Storage Array Architecture for Nextgeneration, Micro 2010
Moneta: A High-performance Storage Array Architecture for Nextgeneration, Non-volatile Memories Micro 2010 NVM-based SSD NVMs are replacing spinning-disks Performance of disks has lagged NAND flash showed
Predictable Time-Sharing for DryadLINQ Cluster. Sang-Min Park and Marty Humphrey Dept. of Computer Science University of Virginia
Predictable Time-Sharing for DryadLINQ Cluster Sang-Min Park and Marty Humphrey Dept. of Computer Science University of Virginia 1 DryadLINQ What is DryadLINQ? LINQ: Data processing language and run-time
April Copyright 2013 Cloudera Inc. All rights reserved.
Hadoop Beyond Batch: Real-time Workloads, SQL-on- Hadoop, and the Virtual EDW Headline Goes Here Marcel Kornacker marcel@cloudera.com Speaker Name or Subhead Goes Here April 2014 Analytic Workloads on
Principal Software Engineer Red Hat Emerging Technology June 24, 2015
USING APACHE SPARK FOR ANALYTICS IN THE CLOUD William C. Benton Principal Software Engineer Red Hat Emerging Technology June 24, 2015 ABOUT ME Distributed systems and data science in Red Hat's Emerging
Columnstore and B+ tree. Are Hybrid Physical. Designs Important?
Columnstore and B+ tree Are Hybrid Physical Designs Important? 1 B+ tree 2 C O L B+ tree 3 B+ tree & Columnstore on same table = Hybrid design 4? C O L C O L B+ tree B+ tree ? C O L C O L B+ tree B+ tree
Current Topics in OS Research. So, what s hot?
Current Topics in OS Research COMP7840 OSDI Current OS Research 0 So, what s hot? Operating systems have been around for a long time in many forms for different types of devices It is normally general
SharePoint 2010 Technical Case Study: Microsoft SharePoint Server 2010 Enterprise Intranet Collaboration Environment
SharePoint 2010 Technical Case Study: Microsoft SharePoint Server 2010 Enterprise Intranet Collaboration Environment This document is provided as-is. Information and views expressed in this document, including
Performance Characterization, Prediction, and Optimization for Heterogeneous Systems with Multi-Level Memory Interference
The 2017 IEEE International Symposium on Workload Characterization Performance Characterization, Prediction, and Optimization for Heterogeneous Systems with Multi-Level Memory Interference Shin-Ying Lee
Percona Live September 21-23, 2015 Mövenpick Hotel Amsterdam
Percona Live 2015 September 21-23, 2015 Mövenpick Hotel Amsterdam TokuDB internals Percona team, Vlad Lesin, Sveta Smirnova Slides plan Introduction in Fractal Trees and TokuDB Files Block files Fractal
Peacock: Probe-Based Scheduling of Jobs by Rotating Between Elastic Queues
Peacock: Probe-Based Scheduling of Jobs by Rotating Between Elastic Queues Mansour Khelghatdoust and Vincent Gramoli The University of Sydney, Australia Abstract. In this paper, we propose Peacock, a new
CS 5523 Operating Systems: Memory Management (SGG-8)
CS 5523 Operating Systems: Memory Management (SGG-8) Instructor: Dr Tongping Liu Thank Dr Dakai Zhu, Dr Palden Lama, and Dr Tim Richards (UMASS) for providing their slides Outline Simple memory management:
Batch Processing Basic architecture
Batch Processing Basic architecture in big data systems COS 518: Distributed Systems Lecture 10 Andrew Or, Mike Freedman 2 1 2 64GB RAM 32 cores 64GB RAM 32 cores 64GB RAM 32 cores 64GB RAM 32 cores 3
Real-time Scheduling of Skewed MapReduce Jobs in Heterogeneous Environments
Real-time Scheduling of Skewed MapReduce Jobs in Heterogeneous Environments Nikos Zacheilas, Vana Kalogeraki Department of Informatics Athens University of Economics and Business 1 Big Data era has arrived!
Tatsuhiro Chiba, Takeshi Yoshimura, Michihiro Horie and Hiroshi Horii IBM Research
Tatsuhiro Chiba, Takeshi Yoshimura, Michihiro Horie and Hiroshi Horii IBM Research IBM Research 2 IEEE CLOUD 2018 / Towards Selecting Best Combination of SQL-on-Hadoop Systems and JVMs à à Application
Optimizing Datacenter Power with Memory System Levers for Guaranteed Quality-of-Service
Optimizing Datacenter Power with Memory System Levers for Guaranteed Quality-of-Service * Kshitij Sudan* Sadagopan Srinivasan Rajeev Balasubramonian* Ravi Iyer Executive Summary Goal: Co-schedule N applications
Kodewerk. Java Performance Services. The War on Latency. Reducing Dead Time Kirk Pepperdine Principle Kodewerk Ltd.
Kodewerk tm Java Performance Services The War on Latency Reducing Dead Time Kirk Pepperdine Principle Kodewerk Ltd. Me Work as a performance tuning freelancer Nominated Sun Java Champion www.kodewerk.com
!! What is virtual memory and when is it useful? !! What is demand paging? !! When should pages in memory be replaced?
Chapter 10: Virtual Memory Questions? CSCI [4 6] 730 Operating Systems Virtual Memory!! What is virtual memory and when is it useful?!! What is demand paging?!! When should pages in memory be replaced?!!
Considering Resource Demand Misalignments To Reduce Resource Over-Provisioning in Cloud Datacenters
Considering Resource Demand Misalignments To Reduce Resource Over-Provisioning in Cloud Datacenters Liuhua Chen Dept. of Electrical and Computer Eng. Clemson University, USA Haiying Shen Dept. of Computer
Introduction to MapReduce
Basics of Cloud Computing Lecture 4 Introduction to MapReduce Satish Srirama Some material adapted from slides by Jimmy Lin, Christophe Bisciglia, Aaron Kimball, & Sierra Michels-Slettvet, Google Distributed
GPUfs: Integrating a File System with GPUs. Yishuai Li & Shreyas Skandan
GPUfs: Integrating a File System with GPUs Yishuai Li & Shreyas Skandan Von Neumann Architecture Mem CPU I/O Von Neumann Architecture Mem CPU I/O slow fast slower Direct Memory Access Mem CPU I/O slow
Mbench: Benchmarking a Multicore Operating System Using Mixed Workloads
Mbench: Benchmarking a Multicore Operating System Using Mixed Workloads Gang Lu and Xinlong Lin Institute of Computing Technology, Chinese Academy of Sciences BPOE-6, Sep 4, 2015 Backgrounds Fast evolution
Tuning Intelligent Data Lake Performance
Tuning Intelligent Data Lake 10.1.1 Performance Copyright Informatica LLC 2017. Informatica, the Informatica logo, Intelligent Data Lake, Big Data Mangement, and Live Data Map are trademarks or registered
Reactive NUCA: Near-Optimal Block Placement and Replication in Distributed Caches
Reactive NUCA: Near-Optimal Block Placement and Replication in Distributed Caches Nikos Hardavellas Michael Ferdman, Babak Falsafi, Anastasia Ailamaki Carnegie Mellon and EPFL Data Placement in Distributed
Don t Get Caught In the Cold, Warm-up Your JVM Understand and Eliminate JVM Warm-up Overhead in Data-parallel Systems
Don t Get Caught In the Cold, Warm-up Your JVM Understand and Eliminate JVM Warm-up Overhead in Data-parallel Systems David Lion, Adrian Chiu, Hailong Sun*, Xin Zhuang, Nikola Grcevski, Ding Yuan University
NVMf based Integration of Non-volatile Memory in a Distributed System - Lessons learned
14th ANNUAL WORKSHOP 2018 NVMf based Integration of Non-volatile Memory in a Distributed System - Lessons learned Jonas Pfefferle, Bernard Metzler, Patrick Stuedi, Animesh Trivedi and Adrian Schuepbach
vsan 6.6 Performance Improvements First Published On: Last Updated On:
vsan 6.6 Performance Improvements First Published On: 07-24-2017 Last Updated On: 07-28-2017 1 Table of Contents 1. Overview 1.1.Executive Summary 1.2.Introduction 2. vsan Testing Configuration and Conditions
Chapter 9: Virtual Memory
Chapter 9: Virtual Memory Silberschatz, Galvin and Gagne 2013 Chapter 9: Virtual Memory Background Demand Paging Copy-on-Write Page Replacement Allocation of Frames Thrashing Memory-Mapped Files Allocating