XSLT benchmarks

To compare Gregor's performance against several other XSLT processors we have used two publicly available benchmarks: DataPower's XSLTMark and Sarvega's XSLTBench. Both benchmarks are collections of XML transformation tasks each consisting of an example XML input and an accompanying XSLT transform. Each benchmark runs the tasks multiple times using supplied XSLT processors and then summarizes the results. Each benchmark exercises a wide variety of XSLT mechanisms striving to provide an unbiased model of processors' expected real life performance.

Gregor/XSLT was initially compared with several other processors using XSLTMark (results).

Informed by experiences with real life examples from Gregor's early adopters we now strongly prefer Sarvega's XSLTBench to the earlier XSLTMark; here is a comparison:

The biggest differences are

1. parsing input: although XML enables multipurposing by design our overwhelming experience has been that documents are almost always parsed before transformation; transformation results are then used and input discarded
2. stylesheet size and complexity: all of our customers' stylesheets are large and complex
3. namespace processing can strongly affect overall performance: namespace use is now the norm

Gregor/XSLT results for the XSLTBench suite are here.

[ Home ] [ Up ]

Send mail to info@ambrosoft.com with questions or comments about this web site.
Copyright © 2008 Ambrosoft, Inc.
Last modified: March 07, 2008