This proposal is for seed money to enable research in the area of limited resource sharing systems. The goal is to establish enough results so as to warrant the submission of a full proposal in 12 months.
In limited resource sharing systems, the degree of sharing between jobs is limited. An example is a time-sharing (processor-sharing) system, such as a CPU, where a fixed Multi-Programming-Limit (MPL) is imposed, so that only MPL number of jobs can share the CPU at a time. Another example is the classic M/G/k multiserver queueing system, with a fixed limit, k, on the number of servers.
Although such limited resource sharing systems are quite prevalent in computer systems and manufacturing systems, their analysis has stymied researchers for decades. This proposal makes a breakthrough in the analysis of such systems, by observing that the third moment of the job size distribution plays a critical role in their analysis -- a fact that has gone unnoticed all these years. The proposal seeks to understand this effect and find more accurate analyses of these important systems.
This proposal is a collaboration between two schools, Carnegie Mellon University and Georgia Institute of Technology, and thus a portion of the funds will be used to pay for travel between the two universities.
