Motivation VM VM VM VM VMM 2 15-02-02 The Storage Performance - - PowerPoint PPT Presentation

▶
motivation
SMART_READER_LITE
LIVE PREVIEW

Motivation VM VM VM VM VMM 2 15-02-02 The Storage Performance - - PowerPoint PPT Presentation

Nov 11, 2022 427 likes •571 views

Storage Performance Analyzer (SPA) Measuring, Monitoring, and Modeling of I/O Performance in Virtualized Environments Feb 2, 2015 Qais Noorshams, Axel Busch , Samuel Kounev, Ralf Reussner Austin, Texas, USA SOFTWARE DESIGN AND QUALITY GROUP

slide-1
SLIDE 1

KIT – University of the State of Baden-Wuerttemberg and National Research Center of the Helmholtz Association SOFTWARE DESIGN AND QUALITY GROUP INSTITUTE FOR PROGRAM STRUCTURES AND DATA ORGANIZATION, FACULTY OF INFORMATICS

www.kit.edu

Storage Performance Analyzer (SPA) –

Measuring, Monitoring, and Modeling of I/O Performance in Virtualized Environments

Qais Noorshams, Axel Busch, Samuel Kounev, Ralf Reussner

Feb 2, 2015

Austin, Texas, USA

slide-2
SLIDE 2

2 15-02-02

Motivation

The Storage Performance Analyzer (SPA)

VMM VM VM VM VM

slide-3
SLIDE 3

3 15-02-02

Motivation

The Storage Performance Analyzer (SPA)

VMM VM VM VM VM

slide-4
SLIDE 4

4 15-02-02

Motivation

The Storage Performance Analyzer (SPA)

VMM VM VM VM VM

slide-5
SLIDE 5

5 15-02-02

Motivation

The Storage Performance Analyzer (SPA)

VMM VM VM VM VM

slide-6
SLIDE 6

6 15-02-02

Motivation

The Storage Performance Analyzer (SPA)

VMM VM VM VM VM

Simultaneous Requests 20 40 60 80 100 R e q u e s t S i z e ( K B ) 20 40 60 R e s p

  • n

s e T i m e ( m s ) 5 10 15

slide-7
SLIDE 7

7 15-02-02

Motivation

The Storage Performance Analyzer (SPA)

VMM VM VM VM VM

Simultaneous Requests 20 40 60 80 100 R e q u e s t S i z e ( K B ) 20 40 60 R e s p

  • n

s e T i m e ( m s ) 5 10 15

slide-8
SLIDE 8

8 15-02-02

Storage Performance Analyzer I

Two part framework i) Benchmark harness with included I/O benchmarks ii) Tailored analysis library to analyze the results Tailored and pre-packaged for performance evaluation storage systems in distributed and virtualized environments Accepted in SPEC RG repository of peer-reviewed tools

http://research.spec.org/tools/

The Storage Performance Analyzer (SPA)

℠

Research

slide-9
SLIDE 9

9 15-02-02

Storage Performance Analyzer II

Design aspects and Practical benefits

Synchronized execution on multiple targets Automation of experiment runs with parameter variation Persisting results and avoid piles of log files Easy and powerful statistical evaluation

The Storage Performance Analyzer (SPA)

slide-10
SLIDE 10

10 15-02-02

Benchmark harness

Coordinates the execution of attached benchmarks and monitors Written in high-level programming language (Java) à Easy to debug

The Storage Performance Analyzer (SPA)

Architecture

Analysis Library Benchmark Harness

Benchmark Controller Benchmark Driver DataStore Interface Persistence Component Regression Modeling Regression Optimization Regression Techniques Benchmarking Component Performance Modeling Component

SQLite

Monitoring Driver

R Libraries Running on Target(s) Optional

Benchmark Benchmark Monitor Remote Execution

Easy data persistence

Lightweight SQLite Database

Tailored analysis library

Processes and evaluates the collected data and measurements Integrated into statistics tool R (http://www.r-project.org/) àFull control over the analysis

slide-11
SLIDE 11

11 15-02-02

SPA in Action – Case Studies

Run the benchmark and analyze the results

  • I. System Analysis and Evaluation
  • II. Statistical Analysis and Statistical Modeling
  • III. System Analysis and Explicit Modeling

The Storage Performance Analyzer (SPA)

Storage-Performance-Influencing Factors Workload Requests Size Mix Pattern Locality System Operating System File System I/O Scheduler Hardware

Derived from Noorshams et al. (2012) IBM System z Processors, Memory PR/SM (Hypervisor) LPAR1 LPAR2 z/Linux z/Linux App. App. IBM DS8700 Harddisks (RAID) Storage Controller Volatile Cache Non-Volatile Cache

Switched Fibre Channel Fibre Channel

LRM MARS CART M5 Cubist Piecewise linearity Tree structure Step function Combination Boosting Inst.-based

Simultaneous Requests 20 40 60 80 100 Request Size (KB) 20 40 60 Response Time (ms) 5 10 15

  • S

i m u l t a n e

  • u

s R e q u e s t s 20 40 60 80 100 Request Size (KB) 20 40 60 Response Time (ms) 5 10 15

  • S

i m u l t a n e

  • u

s R e q u e s t s 20 40 60 80 100 Request Size (KB) 20 40 60 Response Time (ms) 5 10 15

  • S

i m u l t a n e

  • u

s R e q u e s t s 20 40 60 80 100 R e q u e s t S i z e ( K B ) 20 40 60 Response Time (ms) 5 10 15

  • Simultaneous Requests

20 40 60 80 100 R e q u e s t S i z e ( K B ) 20 40 60 Response Time (ms) 5 10 15

slide-12
SLIDE 12

12 15-02-02

Conclusion

The Storage Performance Analyzer (SPA)

  • Systematic analysis of I/O performance in virtualized environments
  • Measuring, Monitoring, and Modeling of I/O Performance
  • Peer-reviewed tool allowing analysis with high degree of automation

Storage Performance Analyzer (SPA)

  • SPA Project Website


http://storageperformanceanalyzer.github.io/SPA/

  • Sources and prepared drops for common platforms
  • Documentation and examples
  • SPEC RG Tool Repository


http://research.spec.org/tools/

  • Peer-reviewed tools

Download

℠

Research

http://www.descartes-research.net/tools/