Research Dimensions of SOA - pt3
Yet another research dimension of SOA.
While SOA abstracts IT functionality on basis of standards, these standards base themselves on XML as the core representation format. The dependence upon XML imposes performance constraints due its verbosity, and need for parsing.
How can we do research to overcome the performance challenge of XML in SOA? Read on..
While the open and text based nature of XML makes it an easy language for usage in SOA applications, however the performance limitations it imposes need to be tackled in terms of producing scalable high performance SOA.
How do we tackle this?
Couple of directions we are working on include:
1. Inducing XML compression infrastructure in SOA environments and evaluation of the performance impact thereof.
2. Next Generation XML parsers like STAX etc. and their role in SOA performance enhancement
3. Abstract machines for intermediate languages to speed XML based parsing
4. Optimizing grammars for XML languages
5. Light intermediate languages for translation of XML documents
These are interesting directions with promise for producing high performance SOA.
Comments
XML surely has come up as "the standard" for "writing standards". However, It remains simple for simple things but how about complex structures? Or, large structures? It becomes too complex!
XML fits pretty good in the present arena. But it would become quite complex in semantic scenarios(eg. RDF structures and operations like search over them). It sound pretty much like using Digital Computing for analog processing for almost all digital devices, however, is not sufficient anymore. Complex activities like recognising videos, images, cant be carried with text based meta-data-tagging, but has to involve some intelligence which has miserably failed in digital world.
Besides improving XML processing further, we shall also start looking for specialised formats suiting different requirements, or simply, different models.
String based data are too inefficient for processing. Unless standards go even further down to binary level, real-time scenarios are still far to achieve.
Posted by: Shaurabh Bharti | October 6, 2006 12:34 PM