Computer Science 1 Resource Overbooking and Application Profiling in Shared Hosting Platforms Bhuvan Urgaonkar Prashant Shenoy Timothy Roscoe † UMASS Amherst.

Slides:



Advertisements
Similar presentations
Pricing for Utility-driven Resource Management and Allocation in Clusters Chee Shin Yeo and Rajkumar Buyya Grid Computing and Distributed Systems (GRIDS)
Advertisements

Distributed Multimedia Systems Tarek Elshaarani Vahid Rafiei.
Full-System Timing-First Simulation Carl J. Mauer Mark D. Hill and David A. Wood Computer Sciences Department University of Wisconsin—Madison.
Adaptive QoS Control Based on Benefit Optimization for Video Servers Providing Differential Services Ing-Ray Chen, Sheng-Yun Li, I-Ling Yen Presented by.
1 Scoped and Approximate Queries in a Relational Grid Information Service Dong Lu, Peter A. Dinda, Jason A. Skicewicz Prescience Lab, Dept. of Computer.
Detecting Transient Bottlenecks in n-Tier Applications through Fine- Grained Analysis Qingyang Wang Advisor: Calton Pu.
Memory Buddies: Exploiting Page Sharing for Smart Colocation in Virtualized Data Centers Timothy Wood, Gabriel Tarasuk-Levin, Prashant Shenoy, Peter Desnoyers*,
U NIVERSITY OF M ASSACHUSETTS, A MHERST – Department of Computer Science Dynamic Provisioning for Multi-tier Internet Applications Bhuvan Urgaonkar, Prashant.
Computer Science Deadline Fair Scheduling: Bridging the Theory and Practice of Proportionate-Fair Scheduling in Multiprocessor Servers Abhishek Chandra.
U NIVERSITY OF M ASSACHUSETTS, A MHERST – Department of Computer Science Quantifying the Benefits of Resource Multiplexing in On-Demand Data Centers Abhishek.
U NIVERSITY OF M ASSACHUSETTS, A MHERST – Department of Computer Science Dynamic Resource Allocation for Shared Data Centers Using Online Measurements.
Handling Web Hotspots at Dynamic Content Web Sites Using DotSlash Weibin Zhao Henning Schulzrinne Columbia University NYMAN’04.
Fair Scheduling in Web Servers CS 213 Lecture 17 L.N. Bhuyan.
Energy Efficient Web Server Cluster Andrew Krioukov, Sara Alspaugh, Laura Keys, David Culler, Randy Katz.
Handling Web Hotspots at Dynamic Content Web Sites Using DotSlash Weibin Zhao Henning Schulzrinne Columbia University Dagstuhl.
Adaptive Content Delivery for Scalable Web Servers Authors: Rahul Pradhan and Mark Claypool Presented by: David Finkel Computer Science Department Worcester.
Operating Systems Operating System Support for Multimedia.
Bandwidth Allocation in a Self-Managing Multimedia File Server Vijay Sundaram and Prashant Shenoy Department of Computer Science University of Massachusetts.
Computer Science Surplus Fair Scheduling: A Proportional-Share Scheduling Algorithm for Symmetric Multiprocessors Abhishek Chandra Micah Adler Pawan Goyal.
© 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice Automated Workload Management in.
A Prediction-based Real-time Scheduling Advisor Peter A. Dinda Prescience Lab Department of Computer Science Northwestern University
Resource Management in Virtualization-based Data Centers Bhuvan Urgaonkar Computer Systems Laboratory Pennsylvania State University Bhuvan Urgaonkar Computer.
Computer Science Cataclysm: Policing Extreme Overloads in Internet Applications Bhuvan Urgaonkar and Prashant Shenoy University of Massachusetts.
AGILE, DYNAMIC PROVISIONING OF MULTITIER INTERNET APPLICATIONS Bhuvan Urgaonkar, Prashant Shenoy, Abhishek Chandray, and Pawan Goyal ACM Transactions on.
U NIVERSITY OF M ASSACHUSETTS, A MHERST Department of Computer Science Black-box and Gray-box Strategies for Virtual Machine Migration Timothy Wood, Prashant.
Adaptive Control of Virtualized Resources in Utility Computing Environments HP Labs: Xiaoyun Zhu, Mustafa Uysal, Zhikui Wang, Sharad Singhal University.
Dynamic and Decentralized Approaches for Optimal Allocation of Multiple Resources in Virtualized Data Centers Wei Chen, Samuel Hargrove, Heh Miao, Liang.
SEDA: An Architecture for Well-Conditioned, Scalable Internet Services
1 An SLA-Oriented Capacity Planning Tool for Streaming Media Services Lucy Cherkasova, Wenting Tang, and Sharad Singhal HPLabs,USA.
U NIVERSITY OF M ASSACHUSETTS, A MHERST – Department of Computer Science An Analytical Model for Multi-tier Internet Services and its Applications Bhuvan.
Virtual Machine Scheduling for Parallel Soft Real-Time Applications
Profile Driven Component Placement for Cluster-based Online Services Christopher Stewart (University of Rochester) Kai Shen (University of Rochester) Sandhya.
IISWC 2007 Panel Benchmarking in the Web 2.0 Era Prashant Shenoy UMass Amherst.
Workload-driven Analysis of File Systems in Shared Multi-Tier Data-Centers over InfiniBand K. Vaidyanathan P. Balaji H. –W. Jin D.K. Panda Network-Based.
© 2006 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice Profiling and Modeling Resource Usage.
MClock: Handling Throughput Variability for Hypervisor IO Scheduling in USENIX conference on Operating Systems Design and Implementation (OSDI ) 2010.
Computer Science 1 Resource Overbooking and Application Profiling in Shared Hosting Platforms Bhuvan Urgaonkar Prashant Shenoy Timothy Roscoe † UMASS Amherst.
A dynamic optimization model for power and performance management of virtualized clusters Vinicius Petrucci, Orlando Loques Univ. Federal Fluminense Niteroi,
Providing Differentiated Levels of Service in Web Content Hosting Jussara Almeida, etc... First Workshop on Internet Server Performance, 1998 Computer.
Design and Evaluation of a Model for Multi-tiered Internet Applications Bhuvan Urgaonkar Internship project talk – Services Management Middleware Dept,
1 Virtual Machine Memory Access Tracing With Hypervisor Exclusive Cache USENIX ‘07 Pin Lu & Kai Shen Department of Computer Science University of Rochester.
1 Admission Control and Request Scheduling in E-Commerce Web Sites Sameh Elnikety, EPFL Erich Nahum, IBM Watson John Tracey, IBM Watson Willy Zwaenepoel,
Flexibility, Manageability and Performance in a Grid Storage Appliance John Bent, Venkateshwaran Venkataramani, Nick Leroy, Alain Roy, Joseph Stanley,
Empirical Quantification of Opportunities for Content Adaptation in Web Servers Michael Gopshtein and Dror Feitelson School of Engineering and Computer.
Computer Science Dynamic Resource Management in Internet Data Centers Prashant Shenoy University of Massachusetts.
Application Performance in the QLinux Multimedia Operating System Sundaram, A. Chandra, P. Goyal, P. Shenoy, J. Sahni and H. Vin Umass Amherst, U of Texas.
DynamicMR: A Dynamic Slot Allocation Optimization Framework for MapReduce Clusters Nanyang Technological University Shanjiang Tang, Bu-Sung Lee, Bingsheng.
U NIVERSITY OF M ASSACHUSETTS, A MHERST – Department of Computer Science Dynamic Resource Management in Internet Data Centers Bhuvan Urgaonkar Laboratory.
1 Agility in Virtualized Utility Computing Hangwei Qian, Elliot Miller, Wei Zhang Michael Rabinovich, Craig E. Wills {EECS Department, Case Western Reserve.
Profiling, Prediction, and Capping of Power in Consolidated Environments Bhuvan Urgaonkar Computer Systems Laboratory The Penn State University Talk at.
Managing Web Server Performance with AutoTune Agents by Y. Diao, J. L. Hellerstein, S. Parekh, J. P. Bigus Presented by Changha Lee.
Making the “Box” Transparent: System Call Performance as a First-class Result Yaoping Ruan, Vivek Pai Princeton University.
Providing Differentiated Levels of Service in Web Content Hosting J ussara Almeida, Mihaela Dabu, Anand Manikutty and Pei Cao First Workshop on Internet.
Friendly Virtual Machines Zhang,Bestavros etc., Boston Univ. ACM/USENIX VEE 2005 CSE 598c April 17, 2006 Bhuvan Urgaonkar CSE 598c April 17, 2006 Bhuvan.
Capsule Placement in the Service Platform Bhuvan Urgaonkar Timothy Roscoe Systems Group, Sprint ATL.
Architecture for Resource Allocation Services Supporting Interactive Remote Desktop Sessions in Utility Grids Vanish Talwar, HP Labs Bikash Agarwalla,
Dynamic Resource Allocation for Shared Data Centers Using Online Measurements By- Abhishek Chandra, Weibo Gong and Prashant Shenoy.
Towards an integrated multimedia service hosting overlay Dongyan Xu Xuxian Jiang Proceedings of the 12th annual ACM international conference on Multimedia.
A Practical Performance Analysis of Stream Reuse Techniques in Peer-to-Peer VoD Systems Leonardo B. Pinho and Claudio L. Amorim Parallel Computing Laboratory.
DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING CLOUD COMPUTING
Abhinav Kamra, Vishal Misra CS Department Columbia University
Distributed Multimedia Systems
Dynamic Provisioning for Multi-tier Internet Applications
Cluster Resource Management: A Scalable Approach
Integrated Resource Management for Cluster-based Internet Services
Zhen Xiao, Qi Chen, and Haipeng Luo May 2013
CLUSTER COMPUTING.
DotSlash: An Automated Web Hotspot Rescue System
Admission Control and Request Scheduling in E-Commerce Web Sites
Cataclysm: Handling Extreme Overloads in Internet Services
Presentation transcript:

Computer Science 1 Resource Overbooking and Application Profiling in Shared Hosting Platforms Bhuvan Urgaonkar Prashant Shenoy Timothy Roscoe † UMASS Amherst and Intel Research † Fifth USENIX OSDI, Boston, Dec 2002

Computer Science 2 Introduction r Proliferation of Internet applications m E-commerce, streaming media, online games, … r Commonly hosted on clusters of servers m Cheaper alternative to large multiprocessors Clients Internet Streaming Games E-commerce cluster

Computer Science 3 Hosting Platforms r Hosting platform: server cluster that runs third-party applications r Applications pay for server resources m CPU, network bandwidth, memory, disk r Platform provider guarantees resource availability r Challenge: Maximize # hosted applications while providing resource guarantees

Computer Science 4 Design Challenges r How to determine an application’s resource needs? r How to provision resources to meet these needs? r How to map applications to servers in the platform? r How to handle dynamic variations in load?

Computer Science 5 Talk Outline þ Introduction r Inferring Resource Requirements r Provisioning Resources r Mapping Applications to Servers r Experimental Evaluation r Related Work

Computer Science 6 Terminology r Hosting platform models m Dedicated: Applications get integral # nodes m Shared: Applications may get fractional # nodes Applications Platform nodes http App serv DB serv r Capsule: component of an application running on a node

Computer Science 7 Provisioning By Overbooking r Worst-case provisioning is wasteful m Low utilization of resources r Applications may be tolerant to occasional violations m E.g., CPU guarantees should be met 99% of the time r Possible to provide useful guarantees even after provisioning less than worst-case needs ð Overbook resources to improve utilization m E.g., Airline reservations

Computer Science 8 Application Profiling r Use the Linux Trace Toolkit [Yaghmour00] time Begin CPU quantumEnd CPU quantum ON OFF r Profiling: process of determining resource usage m Run the application on an isolated set of nodes m Subject the application to a real workload m Model CPU and network usage as ON-OFF processes

Computer Science 9 Resource Usage Distribution time Measurement Interval Cumulative Probability Fractional usage 01 1 A 0.99 B Fractional usage ON-OFF PROCESS Probability 01 1 PDF CDF

Computer Science 10 Profiles of Server Applications r Applications exhibit different degrees of burstiness r Need to capture variability in resource usage Postgres Server, 10 clients Probability Fraction of CPU Apache Web Server, 50% cgi-bin Probability Fraction of CPU Streaming Media Server, 20 clients Probability Fraction of NW bandwidth

Computer Science 11 Capturing Burstiness: Token Bucket r Token Bucket (σ, ρ) m Resource usage over t ≤ σ.t + ρ ρ1ρ1 ρ2ρ2 time usage σ 1.t + ρ 1 σ 2.t + ρ 2 time ON-OFF PROCESS r Choose (σ, ρ) based on a high percentile

Computer Science 12 Resource Overbooking Mechanism r Applications specify overbooking tolerance O i m Probability with which capsule needs may be violated r Controlled overbooking via admission control: m Resource requirements of all capsules are met Σ K (σ k ·T min + ρ k )·(1 - O k ) ≤ C·T min m Overbooking tolerances of all capsules are met Pr (Σ K U k > C) ≤ min (O 1,…,O k ) r A node that has sufficient resources for a capsule is feasible for it

Computer Science 13 Mapping Capsules to Nodes r A bipartite graphs of capsules and feasible nodes r Greedy mapping: consider capsules in non-decreasing order of degrees r Multiple feasible nodes => random, best fit, worst fit… capsules nodes capsules nodes Final Mapping

Computer Science 14 Talk Outline þ Introduction þ Inferring Resource Requirements þ Provisioning Resources þ Mapping Applications to Servers r Experimental Evaluation r Related Work

Computer Science 15 The SHARC Prototype r A Linux-based Shared Hosting Platform m 6 Dell Poweredge 1550 servers m Gigabit Ethernet link r Software Components m Profiling  Vanilla Linux + Linux Trace Toolkit m Control plane  Overbooking, placement m QoS-enhanced Linux kernel  HSFQ schedulers

Computer Science 16 Experimental Setup r Prototype running on a 5 node cluster m Each server: 1 GHz PIII with 512MB RAM and Gigabit ethernet m Control plane runs on a dedicated node m Applications run on the other four nodes r Workload: mix of server applications m Apache web server with SPECWeb99 (static & dynamic HTTP) m PostgreSQL database server with pgbench (TPC-B) benchmark m MPEG streaming server with 1.5 Mb/s VBR MPEG-1 clients m Quake I game server with “terminator” bots

Computer Science 17 Resource Overbooking Benefits r Small amounts of overbooking can yield large gains r Bursty applications yield larger benefits Placement of Apache Web Servers No Ovb Ovb=1% Ovb=5% Web Servers Placed Number of Nodes Placement of Streaming Media Servers No Ovb Ovb=1% Ovb=5% Media Servers Placed Number of Nodes

Computer Science 18 Performance with Overbooking Isolated100th99th95thAverage Performance of Postgres Throughput (trans/s) CPU Provisioning Isolated100th99th95thAverage Performance of Apache Throughput (req/s) CPU Provisioning r Performance degradation is within specified overbooking tolerance

Computer Science 19 Handling Flash Crowds r Detect overloads by online profiling r Reacting to overloads (ongoing work) m Compute new allocations m Change allocations, move capsules, add servers

Computer Science 20 Related Work r Single node resource management m Proportional share schedulers: WFQ, SFQ, BVT, … m Reservation based schedulers: Nemesis, Rialto, … r Cluster-based resource management m Cluster Reserves [Aron00] m MUSE [Chase01]: economic approach m Oceano [IBM], Planetary computing [HP] m Clusters for high availability: Porcupine [Saito99] m Grid computing [Globus]

Computer Science 21 Concluding Remarks r Resource management in shared hosting platforms m Application profiling to determine resource usage m Controlled overbooking to improve utilization m Mapping applications to servers r Future work m Handling dynamic workloads m Managing memory and disk bandwidth r URL: