Scheduling and optimising XMP pipeline processing.
Summary
A analyse and research into ways of improving efficiency and streamlining the performance of XML processing.
The Research Team
Mr. Tom McCormack Principal Researcher Department of Computing IT, Sligo.
Mr. Tony Partridge Joint researcher Department of Computing IT, Sligo..
Mr. Michal Sankot Research Student Department of Computing IT, Sligo.
Project Details
Research and write up an overview of the state of the art in process scheduling, parallel execution and throughput optimisation. The work includes picking various algorithms from various industries ranging from manufacturing to hardware and software and explaining the advantages and disadvantages of those algorithms/techniques. In addition, the ‘correctness’ of those real-life algorithms shall be proven along with assessing ay practical issues surrounding them.
Analyse XML processing pipeline to identify the scheduling techniques most appropriate. It may be that different approaches are better suited to particular task types. If possible, develop a framework for selecting a particular schedule technique.
This is a broad analysis of how XML processing is currently done. It then involves contrasting that with XML pipeline processing and asserting the performance and scalability factors for each.
In depth research into the kinds of document and data to be processed can then be done to assess the suitability of processing documents one way or the other. If neither is possible, it would be good to get a more informal view of where the problems are and what is not suitable for processing in XML pipelines.
Develop formal/semi-formal models for the most promising techniques and liase with XPipe development team on match with implementation.
It involves matching in intricacies of J2EE performance and scalability development with that of the theoretically perfect model supplied from tasks 1 and 2. This provides an ideal opportunity to see where the theoretically correct model is impacted by technology.
Develop formal/semi-formal mechanisms for determining XML processing pipeline computational power and structure requirements given a particular workload mix and associated throughput requirements. This is capacity planning given the model for how our system works.
Research techniques for making pipeline-processing throughput improve over time by means of feedback loops of performance data.
This may possibly involve applying artificial intelligence algorithms to the scheduling of execution of components across many machines. For example, application concepts such as backtracking and fuzzy-logic may provide more optimal ways to handle loads.