Application-level performance prediction tools for network-based svstems

Performance measurement and monitoring in a networked computing system is a multi-level activity. At the lowest levels, performance monitoring provides a measure of “point-to-point” efficiency: one hopes to determine how well a network can move information between two arbitrary network stations under a variety of load and error configurations, usually without regard for the contents of the information being moved. At the middle levels, performance monitoring provides a measure of “end-to-end” network efficiency: one hopes to determine how well a network can move information between two cooperating stations that are working together as agents to implement a solution to a particular application's problems, although, again, without regard for the contents of the information being moved. At the highest levels, performance monitoring provides a measure of “process-to-process” efficiency: one hopes to measure how well the processes that define a network-based application interact, share information, and work together to solve some processing goal. Point-to-point and end-to-end efficiency depend almost exclusively on the physical and logical characteristics of the network architecture used by an application. Process-to-process efficiency, which is derived from the interactions of all processes residing on all stations, depends on the underlying network architecture and the logical interactions of the application itself.