MassConf: Automatic Configuration Tuning By Leveraging User Community Information Computer Science Wei Zheng, Ricardo Bianchini, Thu Nguyen Rutgers University.

Slides:



Advertisements
Similar presentations
Copyright © SoftTree Technologies, Inc. DB Tuning Expert.
Advertisements

Automatic Configuration of Internet Services Wei Zheng, Ricardo Bianchini, and Thu Nguyen Department of Computer Science Rutgers University.
Managing Web server performance with AutoTune agents by Y. Diao, J. L. Hellerstein, S. Parekh, J. P. Bigu Jangwon Han Seongwon Park
SkewReduce YongChul Kwon Magdalena Balazinska, Bill Howe, Jerome Rolia* University of Washington, *HP Labs Skew-Resistant Parallel Processing of Feature-Extracting.
SLA-Oriented Resource Provisioning for Cloud Computing
Autonomic Systems Justin Moles, Winter 2006 Enabling autonomic behavior in systems software with hot swapping Paper by: J. Appavoo, et al. Presentation.
TI: An Efficient Indexing Mechanism for Real-Time Search on Tweets Chun Chen 1, Feng Li 2, Beng Chin Ooi 2, and Sai Wu 2 1 Zhejiang University, 2 National.
5/17/20151 Adaptive RED: An Algorithm for Increasing the Robustness of RED’s Active Queue Management or How I learned to stop worrying and love RED Presented.
Energy Conservation in Datacenters through Cluster Memory Management and Barely-Alive Memory Servers Vlasia Anagnostopoulou Susmit.
Artificial Intelligence in Game Design Introduction to Learning.
Efficient Autoscaling in the Cloud using Predictive Models for Workload Forecasting Roy, N., A. Dubey, and A. Gokhale 4th IEEE International Conference.
Masoud Valafar †, Reza Rejaie †, Walter Willinger ‡ † University of Oregon ‡ AT&T Labs-Research WOSN’09 Barcelona, Spain Beyond Friendship Graphs: A Study.
Karl Schnaitter and Neoklis Polyzotis (UC Santa Cruz) Serge Abiteboul (INRIA and University of Paris 11) Tova Milo (University of Tel Aviv) Automatic Index.
Paper Title Your Name CMSC 838 Presentation. CMSC 838T – Presentation Motivation u Problem paper is trying to solve  Characteristics of problem  … u.
11/14/05ELEC Fall Multi-processor SoCs Yijing Chen.
Hardware-based Load Generation for Testing Servers Lorenzo Orecchia Madhur Tulsiani CS 252 Spring 2006 Final Project Presentation May 1, 2006.
Retrieval Evaluation. Brief Review Evaluation of implementations in computer science often is in terms of time and space complexity. With large document.
Retrieval Evaluation: Precision and Recall. Introduction Evaluation of implementations in computer science often is in terms of time and space complexity.
1 Porcupine: A Highly Available Cluster-based Mail Service Yasushi Saito Brian Bershad Hank Levy University of Washington Department of Computer Science.
Retrieval Evaluation. Introduction Evaluation of implementations in computer science often is in terms of time and space complexity. With large document.
The Relevance Model  A distribution over terms, given information need I, (Lavrenko and Croft 2001). For term r, P(I) can be dropped w/o affecting the.
CoolAir Temperature- and Variation-Aware Management for Free-Cooled Datacenters Íñigo Goiri, Thu D. Nguyen, and Ricardo Bianchini 1.
Take An Internal Look at Hadoop Hairong Kuang Grid Team, Yahoo! Inc
Performance of Web Applications Introduction One of the success-critical quality characteristics of Web applications is system performance. What.
Dynamic Resource Allocation Using Virtual Machines for Cloud Computing Environment.
Surface Simplification Using Quadric Error Metrics Michael Garland Paul S. Heckbert.
Introduction and Overview Questions answered in this lecture: What is an operating system? How have operating systems evolved? Why study operating systems?
Computer Science Department University of Pittsburgh 1 Evaluating a DVS Scheme for Real-Time Embedded Systems Ruibin Xu, Daniel Mossé and Rami Melhem.
TRACEREP: GATEWAY FOR SHARING AND COLLECTING TRACES IN HPC SYSTEMS Iván Pérez Enrique Vallejo José Luis Bosque University of Cantabria TraceRep IWSG'15.
Protecting Sensitive Labels in Social Network Data Anonymization.
Adaptive Data Aggregation for Wireless Sensor Networks S. Jagannathan Rutledge-Emerson Distinguished Professor Department of Electrical and Computer Engineering.
Frontiers in Massive Data Analysis Chapter 3.  Difficult to include data from multiple sources  Each organization develops a unique way of representing.
Achieving Scalability, Performance and Availability on Linux with Oracle 9iR2-RAC Grant McAlister Senior Database Engineer Amazon.com Paper
CISC Machine Learning for Solving Systems Problems Presented by: Alparslan SARI Dept of Computer & Information Sciences University of Delaware
Quantitative Evaluation of Unstructured Peer-to-Peer Architectures Fabrício Benevenuto José Ismael Jr. Jussara M. Almeida Department of Computer Science.
Adaptive Web Caching CS411 Dynamic Web-Based Systems Flying Pig Fei Teng/Long Zhao/Pallavi Shinde Computer Science Department.
A Single-Pass Cache Simulation Methodology for Two-level Unified Caches + Also affiliated with NSF Center for High-Performance Reconfigurable Computing.
A Novel Local Patch Framework for Fixing Supervised Learning Models Yilei Wang 1, Bingzheng Wei 2, Jun Yan 2, Yang Hu 2, Zhi-Hong Deng 1, Zheng Chen 2.
1 Evaluation of Cooperative Web Caching with Web Polygraph Ping Du and Jaspal Subhlok Department of Computer Science University of Houston presented at.
Zibin Zheng DR 2 : Dynamic Request Routing for Tolerating Latency Variability in Cloud Applications CLOUD 2013 Jieming Zhu, Zibin.
To Tune or not to Tune? A Lightweight Physical Design Alerter Nico Bruno, Surajit Chaudhuri DMX Group, Microsoft Research VLDB’06.
Intradomain Traffic Engineering By Behzad Akbari These slides are based in part upon slides of J. Rexford (Princeton university)
Improving Disk Throughput in Data-Intensive Servers Enrique V. Carrera and Ricardo Bianchini Department of Computer Science Rutgers University.
Operating Systems: Wrap-Up Questions answered in this lecture: What is an Operating System? Why are operating systems so interesting? What techniques can.
DynamicMR: A Dynamic Slot Allocation Optimization Framework for MapReduce Clusters Nanyang Technological University Shanjiang Tang, Bu-Sung Lee, Bingsheng.
Managing Web Server Performance with AutoTune Agents by Y. Diao, J. L. Hellerstein, S. Parekh, J. P. Bigus Presented by Changha Lee.
ApproxHadoop Bringing Approximations to MapReduce Frameworks
Top-K Generation of Integrated Schemas Based on Directed and Weighted Correspondences by Ahmed Radwan, Lucian Popa, Ioana R. Stanoi, Akmal Younis Presented.
Enterprise Solutions Chapter 11 – In-memory Technology.
UNIT-3 Performance Evaluation UNIT-3 IT2031. Web Server Hardware and Performance Evaluation Key question is whether a company should host their own Web.
Image Processing A Study in Pixel Averaging Building a Resolution Pyramid With Parallel Computing Denise Runnels and Farnaz Zand.
Optimal Relay Placement for Indoor Sensor Networks Cuiyao Xue †, Yanmin Zhu †, Lei Ni †, Minglu Li †, Bo Li ‡ † Shanghai Jiao Tong University ‡ HK University.
LIOProf: Exposing Lustre File System Behavior for I/O Middleware
3M Display & Graphics © 3M All Rights Reserved. Predict. Measure. Optimize. Control. Realizing the Full Potential of Digital Communication Networks.
Data Summit 2016 H104: Building Hadoop Applications Abhik Roy Database Technologies - Experian LinkedIn Profile:
Re-Architecting Apache Spark for Performance Understandability Kay Ousterhout Joint work with Christopher Canel, Max Wolffe, Sylvia Ratnasamy, Scott Shenker.
Generating, Maintaining, and Exploiting Diversity in a Memetic Algorithm for Protein Structure Prediction Mario Garza-Fabre, Shaun M. Kandathil, Julia.
Curator: Self-Managing Storage for Enterprise Clusters
Introduction Characteristics Advantages Limitations
Introduction | Model | Solution | Evaluation
RE-Tree: An Efficient Index Structure for Regular Expressions
Decoupled Access-Execute Pioneering Compilation for Energy Efficiency
Database Performance Tuning and Query Optimization
“C” and Assembly Language- What are they good for?
Smita Vijayakumar Qian Zhu Gagan Agrawal
Chapter 11 Database Performance Tuning and Query Optimization
Realizing Closed-loop, Online Tuning and Control for Configurable-Cache Embedded Systems: Progress and Challenges Islam S. Badreldin*, Ann Gordon-Ross*,
Adaptive RED: An Algorithm for Increasing the Robustness of RED’s Active Queue Management or How I learned to stop worrying and love RED Presented by:
Area Coverage Problem Optimization by (local) Search
Coevolutionary Automated Software Correction
Presentation transcript:

MassConf: Automatic Configuration Tuning By Leveraging User Community Information Computer Science Wei Zheng, Ricardo Bianchini, Thu Nguyen Rutgers University

Introduction Large software is complex –May have hundreds of configuration parameters –Selecting proper values is important Configuring software is difficult –Depends on hardware, workload, load intensity, and target –Hard to understand the relationship between them –Large configuration space Existing approaches are far from ideal – Hard to find related parameters – Tuning performance involves many time-consuming experiments 2

MassConf Our approach: vendor helps new users’ configuration process – Collect configurations of existing users for new users to try – Rank configurations to minimize the number of experiments Key observations: – A configuration may work well for many users – Multiple configurations may work well for each user Main challenges: – Ranking configurations from most to least promising configurations – Incomplete info about how well each configuration would work 3

Incomplete Information Configuration SpaceUser Space C2C2 C9C9 C4C4 U5U5 U3U3 U1U1 U7U7 Configuration Space User Space C2C2 C9C9 C4C4 U5U5 U3U3 U1U1 U7U7 4 MassConf wants to rank C4 highly.

MassConf Overview Existing User 1 Existing User 2 Existing User N New User M Vendor New User 1 1. Inform environment and configuration 3. Rank configurations 6. Change ranked list 2. Inform environment and target 4. Provide ranked list of configurations 5. Try configurations in turn (resort to Simplex, if needed) 5

Adaptive Ranking Dynamically adapt to place good configurations at the top Three approaches: slow, fast, and fastest C7C7 C2C2 C3C3 C5C5 C9C9 C1C1 C8C8 C4C4 C6C6 First Configuration Last Configuration C7C7 C2C2 C3C3 C9C9 C1C1 C8C8 C6C6 C7C7 C2C2 C3C3 C9C9 C1C1 C8C8 C6C6 C6C6 Original Slow Fast Fastest C5C5 C5C5 C7C7 C2C2 C3C3 C9C9 C1C1 C8C8 C5C5 C4C4 C4C4 C4C4 1 st User 2 nd User 6

Case Study: Apache Performance Synthetic population of users due to lack of real data Workloads: small files, large files, dynamic CGI scripts A “user” is a combination of workload & performance target 219 existing users – Evenly spread in the space of workloads, intensity, and target 195 new users – Evenly spread but not overlapping with existing users 7

Configuration Popularity Some configurations work well for many users. 8

Popularity vs Meeting Users’ Target Some good configurations are not popular. 9

Evaluation MassConf: Adaptive ranking (low, fast, and fastest) Popularity ranking – the intuitive and obvious approach Simplex – a well-known optimization algorithm Metric: number of experiments to satisfy new users 10

Results MassConf successfully reached all performance targets Adaptive ranking beats popularity-based ranking Adaptive ranking: the faster, the better MassConf reaches more users’ targets than Simplex MassConf is also faster than Simplex # of Exp’sPopularity Ranking MassConf Adapt-slow MassConf Adapt-fast MassConf Adapt-fastest Total Avg Max84 11

Conclusions MassConf uses existing configurations to help new users Case study shows that MassConf efficiently achieves the performance targets MassConf can be applied to other software, types of targets Future works: Multi-tier systems In the paper and TR: bootstrapping; optimized MassConf; more experiments, analysis, and results 12

MassConf Overview Existing User 1 Existing User 2 Existing User N New User M Vendor New User 1 1. Inform environment and configuration 2. Cluster environments 4. Rank configurations 7. Store selected configuration 8. Change ranked list 3. Inform environment and target 5. Provide ranked list of configurations 6. Try configurations in turn (resort to Simplex, if needed) 9. Warn about configuration 13