DB-Time-based Oracle Performance Tuning: Theory and Practice

Graham Wood, Uri Shaft, John Beresniewicz Oracle Corporation RMOUG Feb 2008

The following is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions.

The development, release, and timing of any features or functionality described for Oracle’s products remains at the sole discretion of Oracle.

Agenda

• • • • Brief History of Oracle Performance Tuning Methods DB Time: What is it?

DB Time: Data Sources DB Time Method

Oracle Tuning Methods

• Prehistory • Debug code • Dark Ages • • • Counters/Ratios BSTAT/ESTAT SQL*Trace • Renaissance • • • Increased instrumentation inc. Wait Events Move from counters to timers STATSPACK

Oracle Tuning Methods

• More Recent Methods – Time-based methods • YAPP • Instance tuning - instance statistics • • Non intrusive Always on • Method R • Session tuning – sql trace based • Tightly scoped • Must be highly selective • Modern advances • DB Time Tuning • • Instrumentation improvements ASH, AWR,ADDM

Why Do We Care About Time?

Performance Is Always About Time

• Human time is critical to the enterprise • System time includes human and IT resource time to accomplish business goals • System performance affects business goals • “Time is money.” • Performance improvement usually means doing things faster • Method: find where system time is spent – reduce it!

The DB Time Method

• • • • • Uses combination of cumulative and sampled DB Time • ‘Always on’ data only Combines best of current methods • • Low intrusion Detailed data No scope necessary for collections No requirement to reproduce problem Works for concurrency problems such as locking

The DB Time Method

• Supports multiple scopes for diagnosis • • • • • • Database Instance Session Client id Module/Action SQL ID • More inclusive, less intrusive

Database time (DB Time)

• Time spent in database calls by foreground sessions • Includes CPU time, IO time and wait time • Excludes idle wait time • The lingua franca for Oracle performance analysis

Database time is total time spent by user processes either actively working or actively waiting in a database call.

A Single Session Single session with Database Black Box server Browse Books Read Reviews For One Book Add to Cart Checkout = time spent in database TIME

Fundamental concepts Active Session = Session currently spending time in a database call Database Time (DB Time) = Total time session spent in all database calls Average Activity of the Session (% Activity) = The ratio of time active to total wall clock time Browse Books Read Reviews For One Book Add to Cart Checkout = time spent in database TIME

Active sessions

• Foreground • sessions in a database call Backgrounds are also interesting • Either on CPU , waiting for IO , or waiting (not idle) • V$ACTIVE_SESSION_HISTORY is a collection of timed regular samples of active session attributes

Active sessions are foreground sessions contributing to DB time in any given moment.

Multiple Sessions DB Time = Sum of DB Time Over All Sessions Avg. Active Sessions = Sum of Avg. Activity Over All Sessions At time t we have 2 active sessions User 1 User 2 User 3 TIME User

n

= time spent in database

t

The Basic Relationship Database Time Avg. Active Sessions = Wall Clock (Elapsed) Time Browse Books Read Reviews For One Book Add to Cart Checkout = time spent in database TIME

DB time

System load and DB time

• More users • => More calls • => DB time increases • Larger transactions • => Longer calls • => DB time increases

DB time increases as system load increases.

System performance and DB time

• IO performance degrades • => IO time increases • => DB time increases • Application performance degrades • => Wait time increases • => DB time increases

DB time increases as system performance degrades.

System performance and DB time

Where to find DB time?

• V$SYS_TIME_MODEL • • STAT_NAME = ‘DB time’ Accumulated value over entire instance • V$WAITCLASSMETRIC_HISTORY • AVERAGE_WAITER_COUNT • It is precisely Average Active Sessions • V$SYSMETRIC_HISTORY • • • “Database Time Per Second”, “CPU Usage Per Sec” Units are Centi-seconds per second Value is 100 x Average Active Sessions

Where to find DB time?

• V$SQL • ELAPSED_TIME • Also wait class times • V$ACTIVE_SESSION_HISTORY • • Sample per second Count = time

t0 Estimating DB time with ASH

Active sessions DB time is area under curve ASH sample count is value of active sessions function at sample times  t = 1 sec DB Time time

t1

Integral approximation using ASH

DB time

 t0 t1 

ActiveSess ions

(

t

)

dt



lim



t

 0  

k n

  1

ActiveSess ions

(

t

k

) *



t

  

sampletime

 

sampletime



t

1

ASHsamples

0

t

* 1 (where



t



1 second)

EM Top Activity page

• • ASH-estimated DB time by wait class Aggregated over 15 second intervals

Sampled vs. cumulative DB time

Where is DB time used?

• ADDM • AWR and AWR compare periods reports • EM Performance page and drill downs • ASH report • Server-generated Alerts

EM Performance page

• • • Cumulative DB time by wait class v$waitclassmetric_history and v$sysmetric_history 1 minute intervals

DB Time Tuning

• DB Time can be aggregated at multiple levels: • • • • Database / instance Service / module / action Session / user / client id SQL id / rowsource •

Performance improvement for Oracle database means doing the same work in less DB Time

Performance Problem Resolution 101

• • • Discover the problem: User phone call or other complaint Metric threshold alert or system monitoring • • • • Scope the problem: How widespread is it?

How severe is it?

In other words: Who or what is wasting DB Time and how much is being wasted?

• Diagnose the problem • • Scope the solution How much of the pain can be relieved?

The DB Time Method

1.

2.

3.

4.

5.

6.

Scope Set Goal Investigate DB time distribution Identify the largest potential for improvement Modify system Evaluate against Goal

The DB Time Method

• • • Scope • • • What is the problem?

Business Requirements Resource capacity Resource contention System wide or individual Business Function

The DB Time Method

• • • • Set Goal Quantitative Establishes the STOP TUNING criteria Should be business driven for applications (X per day)

The DB Time Method

• • • Investigate DB time distribution Identify major contributors to DB time at the selected scope • • • • • System scope V$SYS_TIME_MODEL V$ACTIVE_SESSION_HISTORY V$SQL Identify high load service, sessions and SQL Identify resource constraints or contention

The DB Time Method

• • Investigate DB time distribution • • • • • • Session scope V$SESS_TIME_MODEL V$ACTIVE_SESSION_HISTORY Identify if database is the problem Identify high load SQL Identify application efficiency issues Identify resource constraints or contention

The DB Time Method

• • Identify the largest potential for improvement • • • • • What can be changed that will produce the greatest reduction in ‘scoped’ DB time?

Parameters System Application SQL Design • Modify system

The DB Time Method

• • • • Evaluate against Goal Did our changes to the system achieve our goal?

If not return to step 3 and repeat If we have reached our goal STOP

Summary

DB Time is the fundamental performance metric The DB Time Method uses many different sources of DB time within the database to allow many different scopes of performance tuning Time based diagnosis removes ‘value judgments’ from performance analysis

New In 11g – Enhancements for RAC

• • • ADDM has “Database” analysis mode New AWR “Database” report EM Performance screens for RAC enhanced to support new server capabilities.