Bugzilla – Bug 5612
Improving Container/Dispatch Reliability and Performance
Last modified: 2007-11-13 11:45:33
You need to
before you can comment on or make changes to this bug.
Profile the Java WS Core code base, specifically the stand-alone
container and the dispatch code, to identify and analyze performance
issues in the code.
Based on some preliminary testing with GRAM, single client access to
the container had limits of around 150 concurrent submissions on a two
processor machine (TeraGrid machine). The focus of this work would be
profile similar access patterns, establish base line numbers, and
effect improvements on the code.
Develop a test suite that helps isolate various aspects of the code
base, with increasing complexity, to mimic use cases like GRAM
usage. The suite must be reusable, added to the test code base and
provide a means for repeatable performance measurements.
Profile the code base using the test suite to identify issues in the
container and dispatch code. The profiling might indicate issues with
other aspects of the core code, potentially security and separate
action items will be created for such work.
Fix relevant code to improve performance or analyze alternative
solutions as needed. Provide recommendations on hosting environments
for use in production.
The tasks need to be on the 4.0.x code base first and followed up with
trunk code profiling and fixes.