Web Service Performance

SOAP/XML Web Service Performance

gSOAP has a very low latency, with performance reaching over 3000 roundtrip SOAP/XML messages per second! See gSOAP performance versus the performance of Apache Axis 1.1 on various platforms compiled by Martin Kuba.

The paper Toward Characterizing the Performance of SOAP Toolkits in the IEEE/ACM International Workshop on Grid Computing shows that gSOAP is the fastest toolkit compared to .NET, Apache Axis Java and C++, and xSOAP.

Below are results obtained in previous studies.

Performance results of different HTTP transport choices and SOAP encodings of a square integer matrix (N by N, where N is the rank) on PIII 550MHz Red Hat Linux

gSOAP (top/left) versus Java RMI (bottom/right) performance

See ICWS 2003 paper for more details.

Load stress test results (with previous gSOAP releases)

Also see SAC 2004 paper on gSOAP's code generation techniques with emphasis on embedded systems (PDF) and ICWS 2004 paper on Finite State Automata for High-Performance XML parsing (PDF).

Not-so-recent performance results can be found in this independent research paper that presents an investigation to the performance of SOAP for scientific computing. The authors implemented a special-purpose implementation for processing arrays of doubles in SOAP/XML and compared the results to gSOAP. The results show that gSOAP's performance is somewhat lower than the special-purpose implementation (but magnitude higher than Apache Axis). The reason for the lower performance is that gSOAP is a full implementation of SOAP and therefore performs more tasks at run time to process a SOAP message compared to the special-purpose implementation. In addition, the study did not enable gSOAP's keep-alive support which accounts for up to 25% of the performance difference. Other differences that explain the performance results are:

gSOAP traverses the entire data structure first to check for multi-referenced objects and cycles in the data graph. Hence, every array element is visited at least twice and its memory address is checked to see if it is multi-referenced by another node.
gSOAP uses XML namespaces and validates with namespaces.
gSOAP uses a callback mechanism for I/O operations which adds flexibility but can impact performance slightly.
The gSOAP messages studied in the paper contain xsi:type attributes for all array elements which increased the message size (these attributes can be omitted).

gSOAP's performance would be close to the reported performance in the paper by eliminating the above features. However, removing the features would result in an implementation that is not fully SOAP compliant.