DBSonar - Slice and Dice Performance Tuning and Management for IDS 7.x, 9.x and 10.x Eric Lam CTO/Cobrasonic
3 Difficulties in database performance tuning Massive data with unknown definition No relationship between DATA Lack of Interpretation knowledge No historical evidence information is kept It happened, but when? Why? What’s the Impact? Fighting among AP, DB, and HW teams Human is not designed for routine checking…
4 Collect Integrate Narrow Down Root Cause Provides analysis and suggestion about real-time/historical resource consumption Cross-linking data from Net, OS, DB and applications Performs slice and dice analysis as OLAP tool to bring out a clear picture of bottleneck Tells what happened inside the user application itself Continuously monitors database systems for potential problems and issues precautious alerts for you to solve these problems before they impact end-users. Allowing system administrators to make informed decisions in handling exceptions or planning resources. Make you an expert to work productively instead of routinely. Mission: A non-stop expertise providing tight linked information for browsing and discovering
5 OSAP DBSonar Server Collecting, Integrating, Interpreting OSAP DBSonar Server Collecting, Integrating, Interpreting DBSonar — System Architecture Repository DBSonar Client Interactive Analysis Tool Alert Deliver Service
6 Fatal ErrorMonitoring and Alert OS PerformanceIndicatorDatabase ServerPerformance Trend SQL DB Profile Network Server Switch Panel
7 Integrates OS, DB and Application Top down analysis Integrated Performance Analysis
8 The power of cross-linking What if System Busy now???
9 Tracking Individual Busy Connection A language independent DB application inspector Easily pin-point the bottleneck instead of tracing the source code
10 SQL Statement Normalization seleCt * from CUSTOMER where fname=‘Eric’ and age = 36; Select * from cuStomer WHERE Fname=‘Brian’ AND age = 32 select * from customer where fname=‘Spark’ and age =35 SQLFrequencyCost select * from customer where fname=‘Eric’ and age = The ONLY way to do meaningful SQL tracing It does matter whatever “condition” have been changed in the where claust, or formatting, all will be normalized to be an identical one!
11 Cost-evaluation of problematic SQL statements Statements are not executed (even on IDS 7.x)
12 Why did the performance drop recently? Check recently outstanding SQL from repository High Execution Cost Heavy Sequential Scan High Frequency
13 Why did performance drop recently? Is the SQL optimizer working properly? How are resources being used and by whom? Any conflict occurred? Is data structure physically optimal?
14 Table Update Statistics Analysis Out of date Table Statistics Use change rate to verify which table need to do update statistics Incomplete Update Statistics Table Distribution is missing?
15 How the table is being used?
16 Check SQL against busy table Evaluate the criteria, whether it choose the right path? Whether additional index need to added or restructured? Drill down to see SQL Analysis Information
17 Customizable Multilayer Trend-chart Display Events are connected based on point in time SQL at point in time can also be retrieved to check where the bottleneck came!
18 Real-Time & Historical Trend Chart Real-Time System Monitoring Trend Chart Disk I/O 、 Memory 、 CPU 、 SQL Zoom in for Detail Utilization Trend Chart Historical Utilization Trend Chart OS – Disk I/O 、 Memory 、 CPU 、 SQL Database – DBSpace I/O 、 Buffer read/write 、 Checkpoint … Both for Hourly and Daily Trend Chart The time-series based chart analysis brings the relationship between event and resource utilization Service as source for capacity planning
19 Expert system for onconfig parameter optimization Parameters Effective Value Suggestion Original Value
20 Maintain a continuous service environment Event Based Alert Subsystem ( , SMS) When/Why/Who causes the issue? Fast drill down approach to narrow down to the causing to ensure service continuous and recover fast Log event and associated info for long term issue elimination
21 Instability causing Dead Lock Long Transaction Chunk down Assertion Fail Logical Log Full Disk Full Table Failure Backup Aborted Database resource overflow Unable to allocate memory
22 Facility for problem resolving Problem Resolve Quickly
23 Proactive Resource Monitoring
24 Display of the IDS Online Logs Important Messages -> highlighted Unimportant Messages -> f i l t e re d o u t
Analysis Reporting
26 Database Health Check Reports Analysis Action Guide – System Tuning iteratively Outstanding issues list both system and AP Time series based historical trend charts integrate and form a clear picture among OS, DB and AP Hardware Capacity Plan for Future
27 Database Performance Analysis Report (4)
28 Speed-up Your Business with Intelligent Administration DBSonar ™ offers Higher Overall System Availability Optimized Hardware Utilization Faster Application Response Time Reduced System Administration Effort Reduced Software Development Costs
29 DBSonar - Product Development Focus Disaster Recovery Reliability Security IDS 7.x/9.x/10.x Performance
30 Session Title Session: ##### Eric Lam Cobrasonic Inc.
31 Product Plan Deliver new Rule and Drill Down Tree SNMP support GUI enhancement, optimize in content aspect SQL optimizer rewriter Access shared memory instead of sysmaster Report in PDF formant, support report customization Support Save Directives Generate update statistics SQL command Tuning on Data Dictionary