Tunning Application Server

33

Tuning the Oracle Application Server Middle Tier

Copyright © 2004, Oracle. All rights reserved.

Objectives After completing this lesson, you should be able to do the following: • Describe Oracle Application Server performance methodology • Monitor and tune the performance of Oracle HTTP Server (OHS) • Monitor and tune performance of OC4J processes and applications • Monitor and tune the performance of OracleAS Web Cache • Monitor and tune OracleAS Forms Services

33-2

Copyright © 2004, Oracle. All rights reserved.

Performance Tuning Methodology •

Performance tuning involves – Determining measurable goals of performance – Translating the goals into observable metrics for components – Regular monitoring of metrics – Evolving optimization techniques – Implementing the appropriate tuning solution

•

Some of the parameters used in evaluating performance are: – Response Time, System Throughput, Wait Time

•

33-3

Periodic monitoring helps in anticipating and overcoming performance challenges. Copyright © 2004, Oracle. All rights reserved.

Monitoring the Performance of Oracle Application Server •

• •

•

33-4

The Oracle Application Server performance metrics are measured automatically and continuously. These metrics can be accessed using the Application Server Control. Application Server Control is the central tool to configure, manage monitor and tune the components of Oracle Application Server. Other tools such as dmstool

Copyright © 2004, Oracle. All rights reserved.

OracleAS Hi-Av Tool (iHAT) •

iHAT is a monitoring tool that provides a graphic snapshot of all processes managed by OPMN: – Grid View: Provides status of all OracleAS instances in a single window – Instance View: Provides the complete process topology view of Oracle Application Server – Routing View: Monitors routing relationships between OHS and OC4J

•

•

33-5

To invoke iHAT, use the command: java –jar ihat.zip :<port> For example: java -jar ihat.zip edcdr2p1:6003 A full list of iHAT options can be obtained with: java -jar ihat.zip -h Copyright © 2004, Oracle. All rights reserved.

Notes Page Only

33-6

Copyright © 2004, Oracle. All rights reserved.

iHAT: Using Web Browser

33-7

Copyright © 2004, Oracle. All rights reserved.

Other Monitoring Tools Oracle Application Server also provides the following monitoring tools: • AggreSpy – A prepackaged servlet that reports performance metrics – Reports performance data for an OracleAS instance – Can be used only when the home OC4J component is active

•

dmstool – Is located in the $ORACLE_HOME/bin directory – Allows you to monitor a specific, a set, or all performance metrics – Allows you to specify a reporting interval

33-8

Copyright © 2004, Oracle. All rights reserved.

Comparison of Monitoring Tools •

Application Server Control: – Is the preferred tool for management and monitoring – Is capable of enterprisewide monitoring – Includes detailed as well as summarized information as required

•

AggreSpy can be used: – For monitoring OC4J and OHS only – To obtain detailed information about specific component

•

dmstool: – Command Line tool – Useful for support

33-9

Copyright © 2004, Oracle. All rights reserved.

Monitoring the Oracle HTTP Server •

Monitoring of the Oracle HTTP Server involves: – Assessing the workload – Investigating Oracle HTTP Server errors – Categorizing Oracle HTTP Server problems

• •

33-10

Use the Performance region to drill down to detailed monitoring. When assessing work load, ensure that you are assessing a representative load.

Copyright © 2004, Oracle. All rights reserved.

Detailed Monitoring of OHS

33-11

Copyright © 2004, Oracle. All rights reserved.

Notes Page Only

33-12

Copyright © 2004, Oracle. All rights reserved.

Detailed Monitoring of OHS Logging

33-13

Copyright © 2004, Oracle. All rights reserved.

Notes Page Only

33-14

Copyright © 2004, Oracle. All rights reserved.

Categorized Monitoring with Application Server Control •

Application Server Control allows you to analyze the performance metrics in three categories: – Module Metrics – Virtual Host Metrics – Child Server Metrics

•

33-15

This categorization enables you to isolate performance bottlenecks and fix them.

Copyright © 2004, Oracle. All rights reserved.

Module Metrics

• •

33-16

If OracleAS Web Cache is enabled, then the use of http_core.c should be reduced. If mod_oc4j.c shows high activity, then drill down to the J2EE Applications. Copyright © 2004, Oracle. All rights reserved.

Virtual Host Metrics

33-17

Copyright © 2004, Oracle. All rights reserved.

Monitoring Child Server Processes You should analyze the OHS child server processes if your application server experiences delays or deadlocks.

33-18

Copyright © 2004, Oracle. All rights reserved.

Monitoring OHS Using AggreSpy http:///j2ee/servlet/AggreSpy

33-19

Copyright © 2004, Oracle. All rights reserved.

Managing Server Processes Directives in httpd.conf that affect OHS processes: • StartServers • MaxClients • MaxKeepAliveRequests • MaxRequestPerChild • MaxSpareServers • MinSpareServers • Timeout • KeepAlive • ThreadsPerChild (on Windows platforms)

33-20

Copyright © 2004, Oracle. All rights reserved.

Notes Page Only

33-21

Copyright © 2004, Oracle. All rights reserved.

Monitoring Processor Usage • •

33-22

On UNIX, you can use the sar or mpstat utilities to monitor the processor usage. On Windows, you can use the Performance Monitor.

Copyright © 2004, Oracle. All rights reserved.

Configuring OC4J for Performance Configuring of OC4J for improving the performance includes: • Providing adequate memory resources • Caching and reusing database connections • Managing concurrency and limiting connections • Load balancing • Database monitoring and tuning

33-23

Copyright © 2004, Oracle. All rights reserved.

Monitoring OC4J Process or Instance

33-24

Copyright © 2004, Oracle. All rights reserved.

Monitoring J2EE Applications

33-25

Copyright © 2004, Oracle. All rights reserved.

Setting Java Options You can set the following Java Options: • Server Option • Heap Size • Stack Size • Concurrent IO

33-26

Copyright © 2004, Oracle. All rights reserved.

Notes Page Only

33-27

Copyright © 2004, Oracle. All rights reserved.

Setting Up Data Sources Performance depends on the settings for: • connection pooling • Maximum and minimum open connections • Cached connection inactivity timeout • Wait for free connection timeout • Connection retry interval • Maximum number of connection attempts • JDBC statement cache size

33-28

Copyright © 2004, Oracle. All rights reserved.

Notes Page Only

33-29

Copyright © 2004, Oracle. All rights reserved.

Setting Up Open Connections In a pooled data source: • max-connections is the upper limit of number connections to data source • cacheScheme determines how max-connection is interpreted • Default cacheScheme is DYNAMIC_SCHEME • min-connections is the level of connections maintained • inactivity-timeout specifies the time out for inactive connections • wait-time specifies the time to wait for obtaining data source connection 33-30

Copyright © 2004, Oracle. All rights reserved.

Setting Up Open Connections

33-31

Copyright © 2004, Oracle. All rights reserved.

Setting Up Connection Attributes Using Application Server Control

33-32

Copyright © 2004, Oracle. All rights reserved.

Tips for Tuning Servlets 1. Monitor the load on your server with respect to: – The metric serviceRequest.avg – The metric processRequest.avg

2. Watch for unused sessions observing: – sessionActivation.active – sessionActivation.completed – session-timeout

3. Watch for slow servlet loading – Enable load-on-startup for servlets that require extra time to load.

33-33

Copyright © 2004, Oracle. All rights reserved.

Notes Only Page

33-34

Copyright © 2004, Oracle. All rights reserved.

Loading Servlet Classes at Startup • • • •

33-35

Servlet classes are loaded when the first request is made. This can imply varying performance for requests. The load-on-startup facility enables classes to load when the application starts up. You can use the Application Server Control to specify if a Web module is loaded at start up.

Copyright © 2004, Oracle. All rights reserved.

Tips for Tuning JSPs The JSP developer can improve JSP performance by: • Using the JESI and the Web Object Cache tag library • Tuning JSP code The administrator can improve JSP performance by: • Changing JSP parameters • Using the main_mode parameter

33-36

Copyright © 2004, Oracle. All rights reserved.

Using the main_mode parameter The main_mode parameter for each OC4J instance controls the way run time operation is performed. • recompile: This is the default value. The dispatcher will check the timestamp of the JSP. Retranslate and execute all reload if the timestamp is different. • reload: The dispatcher will check if any classes have been modified since loading. Reload if the timestamp is different. • justrun: The dispatcher will not check if any classes have been modified since loading.

33-37

Copyright © 2004, Oracle. All rights reserved.

Tips for Tuning EJBs •

• •

33-38

The timeout parameter specifies the maximum time taken for a transaction to finish before it is rolled back due to a timeout. The default value for this parameter is 60,000 milliseconds (60 seconds). The timeout should be set to a value greater than or equal to the timeouts used within individual transactions.

Copyright © 2004, Oracle. All rights reserved.

OC4J-Specific Tips for Tuning EJBs • • • • •

33-39

OC4J specific attributes applying to all EJB are stored in orion-ejb-jar.xml. The parameter call-timeout specifies maximum time to wait before calling the EJB method. Default call-timeout for entity bean is 900000 ms and for session beans 0 (always). The parameter max-tx-retries specifies the number of times to retry a transaction. The default max-tx-retries is three for both session and entity beans.

Copyright © 2004, Oracle. All rights reserved.

Configuring orion-ejb-jar.xml The orion-ejb-jar.xml contains the configuration for EJB modules. • cache-timeout: Time for which inactive stateless session bean remains in the pool • call-timeout: Time to wait for an EJB • copy-by-value: To copy all parameters in EJB calls • max-tx-tries: Number of times to retry a failed transaction • persistence-filename: File in which session details are stored for use across server restarts

33-40

Copyright © 2004, Oracle. All rights reserved.

Factors that Affect the Caching Efficiency Some of the factors that affect the caching efficiency are: • Available resources: Memory and bandwidth • Latency • Response time • Web Cache size • Size of documents • Cacheability rules • Frequency of invalidation and nature of content • Personalization and versions

33-41

Copyright © 2004, Oracle. All rights reserved.

Notes Page Only

33-42

Copyright © 2004, Oracle. All rights reserved.

Tuning Network Timeouts •

There are two types of Network Timeouts: – Keep-Alive Timeout for Browser – default 5 seconds – Origin Server Timeout – default 3600 seconds

• •

Keep Alive Timeout and Keep Alive need to be configured for the Browser and Web Cache. Tune time outs based on the load tests. Cache-Specific Configuration > Network Timeouts.

33-43

Copyright © 2004, Oracle. All rights reserved.

Connection Limit To help determine a reasonable Connection Limit, consider following factors: • The number of concurrent clients to serve • The average size of a page and the average number of requests for page • Network bandwidth • How quickly a page is processed Use the Resource Limits link in Web Cache Manager

33-44

Copyright © 2004, Oracle. All rights reserved.

Cache Size Determination • •

Cacheability rules determine if a document is to be cached Memory for document body is allocated in – 4KB chunks if the document size is less than 4KB – 32KB chunks otherwise

•

HTTP response headers use 4KB Use the Resource Limits link in Web Cache Manager

33-45

Copyright © 2004, Oracle. All rights reserved.

Invalidation Performance To expedite invalidation, consider the following: • Use basic invalidation requests for invalidating single object • Use substring matching for invalidating multiple objects in advanced invalidations • Use invalidation indexes to improve performance of content traversal

33-46

Copyright © 2004, Oracle. All rights reserved.

Invalidating Multiple Objects Using SUBSTRING To send an advanced invalidation request with substring matching do the following: 1. Specify the OTHER element in a manual invalidation request that uses the ADVANCEDSELECTOR element. 2. Specify the NAME attribute to use a value of URI, BODY, or QUERYSTRING_PARAMETER. 3. Specify the TYPE attribute to use a value of SUBSTRING.

33-47

Copyright © 2004, Oracle. All rights reserved.

Notes Page Only

33-48

Copyright © 2004, Oracle. All rights reserved.

Using Indexes to Improve Invalidation Performance •

• •

33-49

If invalidation requests use the QUERYSTRING_PARAMETER, then index would be helpful. Using index, Web Cache categorizes objects in cache, and performs faster lookup. To specify an invalidation index, add an INVALIDATIONINDEX element to the webcache.xml.

Copyright © 2004, Oracle. All rights reserved.

Monitoring OracleAS Forms Services •

Use Application Server Control to monitor Forms Services.

•

Click the Forms instance to display the Forms Services page to view information about: – – – –

33-50

Status Configuration Load Response

Copyright © 2004, Oracle. All rights reserved.

Tuning Forms Services • • •

• •

33-51

Run-time pooling for Forms Services provides quick connections at server peak times. Set prestartRuntimes to true to enable run-time pooling and pre start. Specify the number of run-time executables to be prespawned with the parameter initialRuntimes. Set the timeOut to 0 if you want all the spawned executable to exist even when inactive. Set the minRuntimes to the number of spawned executables to exist even when inactive.

Copyright © 2004, Oracle. All rights reserved.

Summary In this lesson, you should have learned to do the following: • Describe the performance methodology • Monitor and tune the performance of Oracle HTTP Server (OHS) • Monitor and tune performance of OC4J processes and applications • Monitor and tune the performance of OracleAS Web Cache • Monitor and tune OracleAS Forms Services

33-52

Copyright © 2004, Oracle. All rights reserved.