Progress Report Meeting

Slides:



Advertisements
Similar presentations
Autonomic Scaling of Cloud Computing Resources
Advertisements

Copyright © 2008 SAS Institute Inc. All rights reserved. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks.
An Interactive-Voting Based Map Matching Algorithm
Feature Selection as Relevant Information Encoding Naftali Tishby School of Computer Science and Engineering The Hebrew University, Jerusalem, Israel NIPS.
Decision Trees and MPI Collective Algorithm Selection Problem Jelena Pje¡sivac-Grbovi´c,Graham E. Fagg, Thara Angskun, George Bosilca, and Jack J. Dongarra,
Cloud Computing Resource provisioning Keke Chen. Outline  For Web applications statistical Learning and automatic control for datacenters  For data.
Vision Based Control Motion Matt Baker Kevin VanDyke.
SBSE Course 3. EA applications to SE Analysis Design Implementation Testing Reference: Evolutionary Computing in Search-Based Software Engineering Leo.
ANALYZING STORAGE SYSTEM WORKLOADS Paul G. Sikalinda, Pieter S. Kritzinger {psikalin, DNA Research Group Computer Science Department.
1 Learning to Detect Objects in Images via a Sparse, Part-Based Representation S. Agarwal, A. Awan and D. Roth IEEE Transactions on Pattern Analysis and.
ADVISE: Advanced Digital Video Information Segmentation Engine
1 Lecture 6 Performance Measurement and Improvement.
1. 2 General problem Retrieval of time-series similar to a given pattern.
Jacinto C. Nascimento, Member, IEEE, and Jorge S. Marques
Chapter 5 Data mining : A Closer Look.
Alert Correlation for Extracting Attack Strategies Authors: B. Zhu and A. A. Ghorbani Source: IJNS review paper Reporter: Chun-Ta Li ( 李俊達 )
Vulnerability-Specific Execution Filtering (VSEF) for Exploit Prevention on Commodity Software Authors: James Newsome, James Newsome, David Brumley, David.
ROOT: A Data Mining Tool from CERN Arun Tripathi and Ravi Kumar 2008 CAS Ratemaking Seminar on Ratemaking 17 March 2008 Cambridge, Massachusetts.
Describing Methodologies PART II Rapid Application Development*
Pipelines for Future Architectures in Time Critical Embedded Systems By: R.Wilhelm, D. Grund, J. Reineke, M. Schlickling, M. Pister, and C.Ferdinand EEL.
The Electronic Geometry Textbook Project Xiaoyu Chen LMIB - Department of Mathematics Beihang University, China.
SYNAR Systems Networking and Architecture Group Scheduling on Heterogeneous Multicore Processors Using Architectural Signatures Daniel Shelepov and Alexandra.
Chapter 10 Information Systems Analysis and Design
UHD::3320::CH121 DESIGN PHASE Chapter 12. UHD::3320::CH122 Design Phase Two Aspects –Actions which operate on data –Data on which actions operate Two.
Exploiting Context Analysis for Combining Multiple Entity Resolution Systems -Ramu Bandaru Zhaoqi Chen Dmitri V.kalashnikov Sharad Mehrotra.
Chapter 5: Business Intelligence: Data Warehousing, Data Acquisition, Data Mining, Business Analytics, and Visualization DECISION SUPPORT SYSTEMS AND BUSINESS.
A Machine Learning Approach to Programming. Agenda Overview of current methodologies. Disadvantages of current methodologies. MLAP: What, Why, How? MLAP:
Chapter 11 Statistical Techniques. Data Warehouse and Data Mining Chapter 11 2 Chapter Objectives  Understand when linear regression is an appropriate.
“Isolating Failure Causes through Test Case Generation “ Jeremias Rößler Gordon Fraser Andreas Zeller Alessandro Orso Presented by John-Paul Ore.
A Generalized Architecture for Bookmark and Replay Techniques Thesis Proposal By Napassaporn Likhitsajjakul.
Fast Query-Optimized Kernel Machine Classification Via Incremental Approximate Nearest Support Vectors by Dennis DeCoste and Dominic Mazzoni International.
Dimensionality Reduction in Unsupervised Learning of Conditional Gaussian Networks Authors: Pegna, J.M., Lozano, J.A., Larragnaga, P., and Inza, I. In.
1 Munther Abualkibash University of Bridgeport, CT.
6  sixsigm a The Lean Innovation Six Sigma Black belt 5-days program will cover the most contemporary process improvement practices.
Analyze Wrap Up and Action Items
Introduction to Machine Learning, its potential usage in network area,
P.Demestichas (1), S. Vassaki(2,3), A.Georgakopoulos(2,3)
Big data classification using neural network
Auburn University
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Machine Learning with Spark MLlib
Reducing OLTP Instruction Misses with Thread Migration
Hiba Tariq School of Engineering
The Design and Analysis of Algorithms
DATA MINING © Prentice Hall.
Improved Speed Estimation in Sensorless PM Brushless AC Drives
Cristian Ferent and Alex Doboli
The Problem Finding a needle in haystack An expert (CPU)
Runtime Analysis of Hotspot Java Virtual Machine
Model-Driven Analysis Frameworks for Embedded Systems
A Graph-based Framework for Image Restoration
Machine Learning Week 1.
Concurrent Graph Exploration with Multiple Robots
Department of Computer Science Northwestern University
Cause and Effect Graphing
World-Views of Simulation
Project Champion: Process Owner: Organization: Project Location:
3.1.1 Introduction to Machine Learning
DMAIC Roadmap DMAIC methodology is central to Six Sigma process improvement projects. Each phase provides a problem solving process where-by specific tools.
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Adaptive Data Refinement for Parallel Dynamic Programming Applications
Measure Phase Wrap Up and Action Items
Timing analysis research
Learning Incoherent Sparse and Low-Rank Patterns from Multiple Tasks
Kostas Kolomvatsos, Christos Anagnostopoulos
NON-NEGATIVE COMPONENT PARTS OF SOUND FOR CLASSIFICATION Yong-Choon Cho, Seungjin Choi, Sung-Yang Bang Wen-Yi Chu Department of Computer Science &
Machine Learning in Business John C. Hull
Srinivas Neginhal Anantharaman Kalyanaraman CprE 585: Survey Project
Six Sigma (What is it?) “Six sigma was simply a TQM process that uses process capabilities analysis as a way of measuring progress” --H.J. Harrington,
Presentation transcript:

Progress Report Meeting Francisco de Melo Jr December 12, 2016 École Polytechnique de Montréal Laboratoire DORSAL

1 Outline 1. Research Questions 2. Review of Enhanced Calling Context Tree (ECCT) 3. Analysis methods 4. Results 5. Conclusion and future work 6. General View of ECCTView POLYTECHNIQUE MONTREAL – Francisco de Melo

2 1.1 Research Questions 1.1 Hypothesis Is it possible to find useful information by analyzing segments of traces from calling context trees? 1.2 Methodology Which methods can be used to do this process? With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

3 1.2 Research Questions 1.2 Automation of a task in Tracecompare Automate the process of finding the root cause of a performance problem based on performance metrics With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

4 1.2 Stages 1.2 One image explanation With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT Trace Compare view Trace Compare view POLYTECHNIQUE MONTREAL – Francisco de Melo

5 1.2 Stages 1.3 Diagram With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT Overview of the system POLYTECHNIQUE MONTREAL – Francisco de Melo

6 2.1 ECCT Review Review of the Enhanced Calling Context Tree Aggregate calls with the same context More efficient than Call Graph and Dynamic Call Tree Can aggregate metrics from the context With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

7 2.2 ECCT Review In one image Dynamic tree Calling Context A A C B C With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT D E D D E POLYTECHNIQUE MONTREAL – Francisco de Melo

8 3.1 ECCT Model View of the model Implementation details: Read the trace Create a summary tree Record the information from the context With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

9 3.2 ECCT Model Chart Summary of ECCT With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

10 4. Analyzing the ECCT 1.2 Grouping techniques 1.1 Regression models With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

11 4. Analyzing the ECCT 4.1 Regression analysis 4.1.1 Statistical process for estimating the relations among variables 4.1.2 Group the data 4.1.3 Analysis of association between groups 4.1.4 Build a regression rule of those relations 4.1.5 Demonstration With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

12 4. Results for Linear Model 5.2 Analysis MRL Y= 213764395.0B0 +4.9281714M1+1.3205339M2 PCA Associated right the workload metric with the duration Explanation The model is able to explain ~ 80% of the duration With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT Linear output POLYTECHNIQUE MONTREAL – Francisco de Melo

13 4. Analyzing the ECCT 4.2 Automatic classification 4.2.1 Classifying a set a metric by using a non-supervised method (k-means and elbow method) 4.2.2 Do this process for each metric 4.2.3 Analysis of the classification 4.2.4 Use the current information to deduce the problem 4.2.5 Demonstration With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

14 4. Results for automatic classification 5.1 Representation scheme With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT Classification explanation POLYTECHNIQUE MONTREAL – Francisco de Melo

14 4. Results for automatic classification 5.1 Workload analysis - Instructions With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT Classification output POLYTECHNIQUE MONTREAL – Francisco de Melo

15 4. Results for automatic classification 5.2 Second workload analysis Chart 1 – Workload cache misses With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT Classification output POLYTECHNIQUE MONTREAL – Francisco de Melo

16 5. Conclusion and Future work 5.1 Conclusion (i) It is possible to classify the data (ii) There still an analysis phase, which may not be deterministic 5.1 Conclusion With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

17 5. Conclusion and Future work 5.2 Future work 1. Testing with real use cases and software 2. Applying a more advanced technique for 2D and matrix With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

18 6. Demonstration of the View General view of ECCTView + features With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

Questions 19 isnaldo-francisco.de-melo-jr@polymtl.ca Any other info? POLYTECHNIQUE MONTREAL – Francisco de Melo

20 References Doray, F, and M. R. Dagenais, "Diagnosing Performance Variations by Comparing Multi-Level Execution Traces", IEEE Transactions on Parallel and Distributed Systems, vol. pp,issue: 99 no. 1, 2016. Andrea Adamoli and Matthias Hauswirth. Trevis: A context tree visualization & analysis framework and its use for classifying performance failure reports. In SoftVis ’10: Proceedings of the ACM Symposium on Software Visualization, 2010. J. M. Spivey. Fast, Accurate Call Graph Profiling. Softw. Pract. Exper., 34(3):249–264, 2004. W. N. Sumner, Y. Zheng, D. Weeratunge, and X. Zhang. Precise Calling Context Encoding. In ACM International Conference on Software Engineering, 2010. With latest techniques and work of pioneers, we have achieved very high tracing speeds and minimum overhead – well and good But adding more features, newer techniques will drag down the desired performance of tracers My goal is to attack those underlying techniques and algorithms so that tracers become future and feature ready and have uniformity JIT really improvesJIT only when necessary – method or trace Explore opportunities for optimizing – like specializing bytecode or improve JITing techniques Like determine instruction type, using specialized instructions. Similar to LuaJIT POLYTECHNIQUE MONTREAL – Francisco de Melo

21 Obrigado POLYTECHNIQUE MONTREAL – Francisco de Melo