In this talk I will discuss research leadiung to the development of Performance Engineering Technology for the Design, Management and Control of Computing Systems.
In the context of enabling such capabilities we view the computing system as consisting of several architectural layers - the application layer, the system software, processing nodes, grid platform architecture, and interconnect layers. The research discussed here and fostered by some government agencies programs is aimed to create computing systems tools that take into account the interrelation among these layers in affecting the behavior of the computing system, and guiding the design, management, operation and control of the system.
The performance engineering thrust will present notions of multilevel
methods and tools, and the notion of performance frameworks for combing
such methods and tools for a particular analysis or view of the system,
it's components and layers. The power of these new methods is in
creating the capability to analyze behavior of existing systems, as well
as enable prediction of the behavior of future systems, thus affecting
their design cycle. In addition use of this technology in the
maintenance and run-time support of computing systems will be discussed.