Server Performance Tuner (Step 1 of 2)

Using the Performance Tuner is a two-step process:

You can configure performance tuning for the entire domain, or for individual GlassFish Server instances or clusters. Ensure that the domain, server instance, or cluster for which you want to perform performance tuning is running.


Note - If you add server instances to a cluster, you should retune the configuration for that cluster.


Server Performance Tuner (Step 1 of 2) page contains the following options. After specifying your entries for these options, click the Next button to proceed to the Confirm Changes (Step 2 of 2) page.

Configuration to Tune

Select the server instance or cluster you want to tune from the Configuration to Tune drop-down list. The available configurations will vary depending on the server instances or clusters that are currently defined in the domain.

Number of Threads (Logical CPUs)

The default is 1.

You do not need to specify the total number of CPUs on the machine, only the number you want to allocate for use by each Enterprise Server instance running on it.

On an Oracle SPARC Enterprise T-Series, the number of logical CPUs (or processor threads) is returned by a command such as psrinfo. An Oracle SPARCT3-4 Server can have up to 512 processor threads. For more information, consult your hardware vendor's data sheet.

Maximum Heap Memory Available

Specify the size of the heap memory for the GlassFish Server's JVM server instance. The default is 1024 MB.

The specified value for heap memory may not be applied if the tuner determines that it is not optimal.

Allocation Time for Database Access

Specify that the allocation time is Low (0% - 10%), Medium (10% - 20%), or High (>= 20%). The default is Medium.

This is the time in the database tier spent by the primary application deployed on the instance or cluster being tuned. If you have multiple applications deployed on the cluster, this should be the sum of the time spent in the database tier by all applications. You should also take into account the time spent per web or EJB request.

Garbage Collection Strategy

Specify the Garbage Collection Strategy.

Specify either Maximum Peak Throughput (Higher GC Pause) or Low GC Pause (Lower Peak Throughput). The default is Maximum Peak Throughput.

This strategy determines what happens to garbage collection in the JVM. Obviously garbage collection should take minimal time, but this may not always be possible. The Enterprise Server's JVM can be configured to have many short GC pauses or a small number of relatively longer GC pauses. When a small number of GC pauses is your strategy, the throughput is generally greater, with 1–2% of requests experiencing latency.

Disable Security Manager?

Specify whether Security Manager should be disabled.

Disabling the security manager may improve performance but may decrease system security. The default is Yes.


Note - This option is grayed out if Security Manager is already disabled on the Configuration->Security settings page for the selected configuration. The Performance Tuner controls only enable you to disable Security Manager if it is currently enabled. If it is currently disabled, you can only enable it from the Configuration->Security page.


Disable Access Logger?

Specify whether Access Logger should be disabled.

Disabling access logging may improve performance but removes some system information. The default is Yes.


Note - This option is grayed out if Access Logging is already disabled on the Configuration->HTTP Service settings page for the selected configuration. The Performance Tuner controls only enable you to disable Access Logging if it is currently enabled. If it is currently disabled, you can only enable it from the Configuration->HTTP Service page.


Running on a SPARC Enterprise T-Series Server?

Check this box if the server instances are running on a SPARC Enterprise T-Series Server. The default is No.

For more information, see http://www.oracle.com/us/products/servers-storage/servers/sparc-enterprise/t-series/index.html.

Copyright © 2010, 2011, Oracle and/or its affiliates. All rights reserved. Legal Notices