Infonyte boosts XSLT Application Performance

Server Applications often require formatting of XML formatted data or RDBMS data to various output formats, e.g. HTML or PDF. Though XSLT processors and stylesheets provide comfortable means to define the required mappings, available Open-Source XSLT processors are often too slow or do not scale with the data source size.

Infonyte's PXSLT engine is the only XSLT accelerator which is fast and scales to almost unbound XML document sizes.

We have conducted several standardized benchmarks (XSLTMark) which show that Infonyte's PXSLT is on average 3-10 times faster than Open-Source alternatives for XSLT based XML data processing and even 7-77 times faster compared to RDBMS based XSLT processing for large data-sets.

The XML data processing benchmarks compare PXSLT with Xalan (interpreted + compiled), Saxon, and MSXSL. The large data-set benchmarks compare PXSLT+Infonyte-DB (as native XML data store) with an optimized RDBMs based solution for the FreeDB CD collection (~500MB data on both sides).

The benchmarks show that PXSLT always outperforms the best competitor. Depending on the benchmark, PXSLT is 1.5 - 50 x faster than Xalan 2.6.0 (compiled), 3 - 277 x faster than Xalan 2.6.0 (interpreted), 2.5 - 100 x faster than SAXON 6.5.3, and 1.5 - 400 x faster than MSXSL. These numbers comprise e.g. also parsing-only, and serialization-only benchmarks where all tested processors cannot outperform each other by magnitudes. For real world XSLT scenarios requiring complex transformations and/or large XML data sets the performance boost is usually 3-10 times.

As PXSLT is able to process multi-gigabyte XML documents (that naturally cannot be processed by traditional XSLT processors), two benchmarks compare PXSLT+Infonyte-DB with an MySQL+XSLT based solution. The first simple benchmark generates a HTML table from fulltext query results, selecting CD-rows from a single table (MySQL) and retrieving XML nodes via Infonyte-DB. The second benchmark extends the first one by retrieving also the track information (in the MySQL case from another table) via a nested-loop join.

Though the first simple benchmark shows the most trivial query case for MySQL, PXSLT was 7 times faster, and in the more complex second benchmark, it outperforms MySQL by factor 77.

The very fast document loading and indexing times (it was ~20 times faster to create and index the FreeDB database with Infonyte-DB than with MySQL) and XSLT based reasoning, transformation, and query facilities provide the means for establishing low-cost materialized XML warehouse based applications.

The detailed benchmark is available for download in PDF format.

Contact: sales@infonyte.com