Title :
Minimizing flow time nonclairvoyantly
Author :
Kalyanasundaram, Bala ; Pruhs, Kirk R.
Author_Institution :
Dept. of Comput. Sci., Pittsburgh Univ., PA, USA
Abstract :
We consider the problem of scheduling a collection of dynamically arriving jobs with unknown execution times so as to minimize the average response/flow time. This is the classic CPU scheduling problem faced by time sharing operating systems. In the standard 3-field scheduling notation this is the nonclairvoyant version of 1|pmtn, rj|ΣFj. Its easy to see that every algorithm that doesn´t unnecessarily idle the processor is at worst n-competitive, where n is the number of jobs. Yet there is no known nonclairvoyant algorithm, deterministic or randomized, with a competitive ratio provably o(n). We present a randomized nonclairvoyant algorithm, RMLF, that has competitive ratio θ(lognloglogn) against an adaptive adversary. RMLF is a slight variation of the multi level feedback (MLF) algorithm used by the Unix operating system, further justifying the adoption of this algorithm. R. Motwani et al. (1994) showed that every randomized nonclairvoyant algorithm is Ω2(log n)competitive, and that every deterministic nonclairvoyant algorithm is Ω2(n1/3 )-competitive
Keywords :
computational complexity; minimisation; processor scheduling; randomised algorithms; RMLF; Unix operating system; adaptive adversary; average response/flow time; classic CPU scheduling problem; competitive ratio; deterministic nonclairvoyant algorithm; dynamically arriving jobs; flow time minimization; multi level feedback algorithm; nonclairvoyant version; randomized nonclairvoyant algorithm; scheduling; standard 3-field scheduling notation; time sharing operating systems; unknown execution times; Computer science; Dynamic scheduling; Feedback; Kirk field collapse effect; Operating systems; Processor scheduling; Random variables; Scheduling algorithm; Time factors; Time sharing computer systems;
Conference_Titel :
Foundations of Computer Science, 1997. Proceedings., 38th Annual Symposium on
Conference_Location :
Miami Beach, FL
Print_ISBN :
0-8186-8197-7
DOI :
10.1109/SFCS.1997.646123