Download presentation
Presentation is loading. Please wait.
1
One Picture is Worth a Thousand Words
Amihay Gonen, Oracle Domain expert, Dell
2
Who Am I More than 20 years work as an Oracle DBA and developer
Worked with Oracle since version 7 Most of the time in R&D departments on BIG DATA projects Managed several DBA teams for more than 10 years. Currently works as an Oracle domain expert in DELL, as part of the FogLight development team. Hobbies : Playing chess , running and developing database-oriented tools 9/20/2018
3
Preface David McCandless said in his TED talk:
"By visualizing information, we turn it into a landscape that you can explore with your eyes, a sort of information map. And when you're lost in information, an information map is kind of useful." 9/20/2018
4
Preface 9/20/2018
5
Agenda Oracle performance Tools Data visualization methods
Poor man’s Heat Map from AWR Root cause analysis Future 9/20/2018
6
Oracle performance Tools
9/20/2018
7
The Evolution of Oracle Performance Tools
UtlBstat-UtlEstat (Oracle 7) StatsPack (Oracle 8,8i) Automatic Workload Repository – (Oracle 10G and above) Active Session History (Oracle 10) Automatic Database Diagnostic Monitor The Evolution of Oracle Performance Tools utlBstat/Etstat – collect basic key metrics like system wait statstics , data dictionary statstics and library cache StatsPack – add extra information include TopSql … to complete … 9/20/2018
8
Oracle performance Tools- StatsPack
Collects more data, including high-resource SQL. Pre-calculates many ratios such as cache hit ratios, rates, and transaction statistics. Separates data collection from report generation. No extra license ($$) Support only 9i metrics No built-in retention scheme. Focus on history only Farther reading: 9/20/2018
9
The evolution of Oracle performance utilities - AWR
ADDM Server alerts AWR/ ASH Advisors Confidential 9/20/2018
10
The evolution of Oracle performance Tools - AWR
A DWH like repository for all performance metrics Repository tables are part of the database (SYSAUX tablespace ) Data is retrieved in a “snapshot” once an hour and retained for 7 days (configurable) Baseline enables to compare between different periods (snapshot range) Basic Report - awrrpt.sql 9/20/2018
11
The Evolution of Oracle Performance Tools - Server Alerts
Database Self-Monitoring Alerts triggered when A problem occurs Data does not match expected values for metrics Alert thresholds are configurable (OEM or DBMS_SERVER_ALERT) View alerts OEM DBA_OUTSTANDING_ALERTS Asynchronous notification (for example: ) 9/20/2018
12
Oracle performance Tools- ADDM
A report from the AWR to identify potential performance bottlenecks (“human like”) Report is combined of the following parts : Finding issues Description of the issue Fix it Recommendation Symptoms which lead to its discovery ADDM report (task) generation is part of AWR snapshot (default one an hour) 9/20/2018
13
The Evolution of Oracle Performance Tools - Advisory
(Automatic) SQL Tuning Advisor – provides recommendations for improving the efficiency of an SQL query or a set of SQL queries Memory Advisors Memory SGA Buffer Cache PGA Undo Advisor (Automatic) Segment Advisor 9/20/2018
14
Data Visualization Methods
9/20/2018
15
Data Visualization Methods Details …
Simplest form of data Compare between different metrics in a point in time Enables to detect macro patterns and also to pay attention for detail information 9/20/2018
16
Data Visualization Methods
first glance impression 9/20/2018
17
Data Visualization Methods - HeatMap
Computers are good with numbers People are good with pictures Quickly focus “hot” areas Get the “global” picture 9/20/2018
18
Data Visualization Methods - TimeLine
Answer What/When questions Know the process to have better understanding 9/20/2018
19
Data Visualization Methods - Magic Square
Compare 2-dimension easily Use common standard in the industry 9/20/2018
20
Poor Man’s Heat Map from AWR
9/20/2018
21
Data Visualization Methods - HeatMap
What are the busiest hours ? Display in one picture data of one week or more Focus in the general picture 9/20/2018
22
Data Visualization Methods - HeatMap
Detect patterns Confidential 9/20/2018
23
Data Visualization Methods - HeatMap
How to bake your “own” heat map – my receipt …. Use popular HeatMap.js library – Query the AWR views to get performance data : X axis – hour in a day Y axis – Day list Count – Value of the metric Max – Max value ( calculate heat values) 9/20/2018
24
Data Visualization Methods - HeatMap
How to bake your “own” heat map – my receipt …. Use smart queries select mod((s.snap_id-first_value(s.snap_id) over (order by s.snap_id)),24) x, trunc((s.snap_id-first_value(s.snap_id) over (order by s.snap_id))/24) y, nvl(round((value-lag(value) over (order by s.snap_id))/ ),0) count from DBA_HIST_SYS_TIME_MODEL m join DBA_HIST_SNAPSHOT s on m.snap_id=s.snap_id where stat_name='DB time'; 9/20/2018
25
Data Visualization Methods - HeatMap
9/20/2018
26
Root Cause Analysis 9/20/2018
27
Root Cause Analysis Hi I’m Bob , I’m the DBA on duty , How may I assist you ? My name is John. My DB suddenly works slower than I am used to No problem, please provide me connection details ? Sure , I’ll send by the SSH connection details Confidential 9/20/2018
28
Root Cause Analysis What to-do next ? Confidential 9/20/2018
29
Root Cause Analysis – Top Down
Data Visualization Table Chart (awr report) * Focus on the sql detail Line Chart * What are the peak time ? * Is the metric out of it’s baseline Information Visualization HeatMap * What are the busiest hours ? * is there a trend ? * What is the problematic resource (Cpu, I/O , Application behavior ) 9/20/2018
30
Root Cause Analysis Hmm… let’s focus on 11/2/2014 \ 9/20/2018
31
Hmm… Interesting wait class changed from User I/O to Application
Root Cause aAnalysis Hmm… Interesting wait class changed from User I/O to Application The system wait class has changed from User I/O to Application wait , it worth focusing on the hours of 3:00 and 4:00 am plus 6:00 and 7:00 AM. \ 9/20/2018
32
Root Cause Analysis Ok , got the problem… \ 9/20/2018
33
Future 9/20/2018
34
Future The ratio databases per DBA is increasing .
DBAs expected to achieve more in less time One of Keys to success is to pick the right tool for the right problem. For code example : 9/20/2018
35
Thank you 9/20/2018
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.