WebSphere Application Server V8.5 Administration and Configuration Guide for the Full Profile

Chapter 16. Monitoring distributed systems 575

Advisors are best used when:

 A reasonable load can be driven to the application server utilizing significant CPU (50+%).

 You want help in tuning a server while establishing initial performance benchmarks.

For more information about the advisors, see the following website:

http://pic.dhe.ibm.com/infocenter/wasinfo/v8r5/topic/com.ibm.websphere.nd.doc/ae/c

prf_whyuseperfadvisors.html

To view the Advisors window, click the Advisor link in the TPV menu window. The advisor

provides advice to the administrator and from the advisors view, the different advice

statements can be selected and further analyzed by the administrator, as shown in

Figure 16-20.

Figure 16-20 Advice window

16.4 Monitoring examples

In this section, a variety of monitoring data is used to provide examples and information about

what that data tells an administrator.

Important: Although the statistical data that is observed in PMI and request metrics

provides a powerful understanding of what is occurring in the environment, it is important

to remember that the counters and statistics are mostly averages. You must always verify

that the statistics make sense for your environment. For example, if an EJB seems to be

having good response time, but it is throwing and catching an exception shortly after

entering the bean, it still appears to have excellent response time. Response time alone is

not enough to indicate a healthy system.

576 WebSphere Application Server V8.5 Administration and Configuration Guide for the Full Profile

16.4.1 JVM memory and CPU usage

Environmentally, the way the applications and application servers use memory and CPU is

important to the overall server performance. The application server PMI data provides some

basic level monitoring capabilities.

The JVM Used memory and ProcessCpuUsage counters can be monitored and displayed in

the TPV graph. Figure 16-21 shows an example of this data.

Figure 16-21 JVM memory and CPU usage

The report can be viewed by clicking Monitoring and Tuning  Performance Viewer 

Current Activity  Performance Modules  JVM Runtime.

This graphical view is a good way to visually check that memory is not constantly growing and

is stabilized over time by garbage collection. It also gives a good indication of how frequently

garbage collection might occur.

However, when tuning memory sizes, use verbose garbage collection and a combination of

the application server support tools available in IBM Support Assistant for the analysis of the

memory usage.

Chapter 16. Monitoring distributed systems 577

Verbose garbage collection

Assuming there is appropriate disk space, a monitoring strategy must include verbose

garbage collection.

To enable verbose garbage collection, click Servers  Server Types  WebSphere

application servers  <your server>  Server Infrastructure  Java and process

management  Process definition  Additional Properties  Java Virtual Machine and

select the Verbose garbage collection option.

When this field is enabled, a report is written to the output stream each time the garbage

collector runs. The information can usually be found in native log files. In Version 7,

WebSphere used the optimal throughput (optthruput) algorithm for garbage collection. Now

generational concurrent garbage (gencon) collection is being used, which allows performance

improvements. Because optthruput uses a large contiguous heap shared by all threads, when

garbage collection is invoked, all of this area is scanned. This policy is beneficial for

applications that demand optimal throughput, but it has long pause times as a side effect.

Generational concurrent divides the heap memory in two pieces:

 Nursery, for new objects

 Tenured, for aged objects

So, in this policy, the time spent to scan one of the areas is smaller.

IBM Support Assistant provides tools that make the garbage collection analyses easier. For

more information about the garbage collection on WebSphere, read the following articles:

http://www.ibm.com/developerworks/websphere/techjournal/1106_bailey/1106_bailey.ht

ml

http://www.ibm.com/developerworks/websphere/techjournal/1108_sciampacone/1108_scia

mpacone.html

Note: TPV provides a good representation of what occurs with the JVM memory, but the

IBM Garbage Collection and Memory Visualizer tool and the IBM Pattern Modelling and

Analysis for Java Garbage Collector tool provide more insightful help with memory-related

configurations to assist tuning memory. For example, you can be provided with advice on

heap sizes and garbage collection algorithms based on the patterns observed in the

garbage collection log.

These tools are available as add-ons to IBM Support Assistant. Use IBM Support Assistant

set up wizards to download these and other Support tools. For more information, refer to

the following web sites:

 IBM Support Assistant:

http://www.ibm.com/software/support/isa/

 Complete list of ISA available add-ons:

http://www.ibm.com/support/docview.wss?uid=swg27013116

 IBM Education Assistant module for The IBM Garbage Collection and Memory

Visualizer:

http://publib.boulder.ibm.com/infocenter/ieduasst/v1r1m0/index.jsp?topic=/co

m.ibm.iea.was_v7/was/7.0/ProblemDetermination/WASv7_GCMVOverview/player.html