Presentation is loading. Please wait.

Presentation is loading. Please wait.

Deploy With Confidence Minimize risks Improve business output Optimize resources.

Similar presentations


Presentation on theme: "Deploy With Confidence Minimize risks Improve business output Optimize resources."— Presentation transcript:

1

2 Deploy With Confidence Minimize risks Improve business output Optimize resources

3 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com3 The Application Lifecycle SYSTEM PERFORMANCE O B J E C T I V E S DEVELOP OPERATE DEPLOY TESTPLAN IT GOVERNANCE DEVELOPMENT APPLICATION READINESS BUSINESS AVAILABILITY

4 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com4 Deployment Complexity DATABASEAPP SERVERWEB SERVERNETWORK Sessions Connection DNS Latency Throughput Virtual Machine Customize Code Indexes SQL Call SECURITY Encryption Firewall LOAD BALANCER Legacy CRM APPLICATION ERP No system-wide perspective Significant time pressure

5 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com5 of deployed applications are rolled back Gartner 50 % Deployment Risks 40 % of software projects fail to meet business objectives Giga 75 % of new ERP/CRM rollouts use more hardware to fix performance AMR

6 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com6 When to Optimize/Tune Initial Rollout Projects New Geographies User Expansions System Integrations Shared Infrastructures OS, App, System Updates

7 Optimization Methodology

8 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com8 Our Optimization Methodology Phase-driven approach to Performance Optimization Phase 4: Report Phase 3: Optimize Phase 2: Baseline Phase 1: Plan & Organize Fully Plan the project Organize the Team Quantify the System Performance Iteratively Isolate and Eliminate Performance Bottlenecks Assess the Performance Improvements and Report

9 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com9 Determine Business Objectives (Phase 1) Business ObjectiveSystem MetricGoal Scale the system to additional users Concurrent Users>5,000 Increase orders processedTransaction Rate>1000 TPS Decrease User Wait timeResponse Time<10s Decrease incomplete transactions Error Rate<2%

10 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com10 Organize the Team (Phase 1) Most problems result from interactions between components Team structure streamlines analysis and diagnosis Application Network ProjectManager DBA Silo-centric Cross Discipline Team Sys Admin NW Admin App Architect Dev QA ProTune Specialist Data Base Plan

11 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com11 Apply Controlled Load ProTune Baseline the System (Phase 2) Application Tier Application Servers Streaming Media Server Infrastructure Tier Fire wall Web Servers Load Balancer Fire wall Database Tier Database Servers Storage Distributed System Gather Metrics Measure Performance … ~~~ Select Business Processes

12 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com12 Compare Results to Targets (Phase 2) Business ObjectiveSystem MetricGoalBaseline Scale the system to additional users Concurrent Users>5,0001,000 Increase orders processed Transaction Rate>1000 TPS400 TPS Decrease User Wait time Response Time<10s20 s Decrease incomplete transactions Error Rate<2%5%

13 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com13 Optimize the System tier by tier (Phase 3) Diagnose by logical tiers Employ component library tests for infrastructure tiers Create specific tests for application and database tiers ProTune Application Tier Application Servers Streaming Media Server Database Tier Database Servers Storage Infrastructure Tier Fire wall Web Servers Load Balancer Fire wall Gather Metrics Measure Performance …

14 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com14 Iteratively Analyze each Tier (Phase 3) Infrastructure Tier Application Tier Database Tier ~~~ Develop Targeted Load Tests Fire wall Web Servers Load Balancer Fire wall ProTune Controlled Load to Isolate Component ~~~ Analyze Results and Recommend Changes Validate Fix Client Expert Implement Fix Move to next Constraint Optimize

15 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com15 EJB - Latency Breakdown

16 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com16 EJB - Throughput Distribution

17 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com17 Re-run Baseline and Assess (Phase 4) Quantify improvements Deliver Executive and Detailed Reports Business ObjectiveSystem MetricGoalBaselineFinal Scale the system to additional users Concurrent Users>5,0001,0005,000 Increase orders processed Transaction Rate>1000 TPS400 TPS1,200 TPS Decrease User Wait time Response Time<10s20 s8s Decrease incomplete transactions Error Rate<2%5%1.5%

18 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com18 Introducing ProTune Validate performance Optimize configuration Verify alignment with business objectives

19 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com19 Safe Deployment System ™ (SDS) Overload Protection DB SERVERS WEB SERVERS APP SERVERS NETWORKSECURITY Systematic Identification Problem Isolation Expert Recommendations Automated Improvements and Validation LOAD BALANCER Including

20 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com20 MONITOR LOAD GENERATORS How It Works CONSOLE 1.Mapped topology 2.Auto-Assigned monitoring 3.Business process emulation 4.System component validation 5.Problem detection 6.Recommended configuration 7.Automated checklist Repeated validation Audit trail

21 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com21 Our Experience and Track Record Over 3000 Optimization Engagements Average Results (performance improvements) : ERP Systems – 100% CRM Systems – 150% Web andJ2EE – 400% Complex C/S– 200%

22 ©2002 Mercury Interactive Corporationwww.mercuryinteractive.com22 Electronic Ticketing System Improves Throughput by 10x… The Case –E-ticket system required throughput of 400 tickets/hr –Web-based application, Oracle database, SABRE system integration The Findings –Infrastructure Configurations: Routers improperly configured, insufficient bandwidth, and insufficient file descriptors for Web Servers –Application Code Defects Isolated JAVA code null pointer exception The Results –Increased system throughput by 10x from 270 to over 3000 transactions/hour …and booked over $1M per day with no down-time.


Download ppt "Deploy With Confidence Minimize risks Improve business output Optimize resources."

Similar presentations


Ads by Google