|
|
Analysis Of Multi-tasking |
|
For my Discrete Stochastic Processes class at MIT (course number 6.262) I wrote a paper analyzing the performance advantage of multi-tasking in the presence of disk access latency. My results probably aren't original and I'm sure you could find a more in depth analysis in the literature. However, I thought my results were quite interesting so the paper I wrote is available if you want to read it. Abstract:In this paper we analyze the advantages of multi-tasking systems in the presence of disk latency. Specifically we analyze a processor which can work on M tasks concurrently. Also we assume that for every small time unit, delta, when a task is being processed, there is a probability lambda * delta of that task requiring a disk access independent of past and future disk accesses of all tasks. We allow the time for the disk drive to complete a disk access to be arbitrarily distributed. From this fairly general model we obtain results for how much faster a multi-tasking system will be compared to a non-multi-tasking system. A surprising result is that the multi-tasking system is not more than twice as fast as the non-multi-tasking system for many reasonable distributions of the disk latency regardless of M.Click here for the full paper in gzipped postscript form. |
|