Task assignment with unknown duration

We consider a distributed server system and ask which policy should be used for assigning tasks to hosts. In our server tasks are not preemptible. Also, the task's service demand is not known a priori. We are particularly concerned with the case where the workload is heavy-tailed, as is characteristic of many empirically measured computer workloads. We analyze several natural task assignment policies and propose a new one TAGS (Task Assignment based on Guessing Size). The TAGS algorithm is counterintuitive in many respects, including load unbalancing, non-work-conserving and fairness. We find that under heavy-tailed workloads, TAGS can outperform all task assignment policies known to us by several orders of magnitude with respect to both mean response time and mean slowdown, provided the system load is not too high.

[1]  M. Crovella,et al.  Heavy-tailed probability distributions in the World Wide Web , 1998 .

[2]  Randolph D. Nelson,et al.  An Approximation for the Mean Response Time for Shortest Queue Routing with General Inerarrival and Service Times , 1993, Perform. Evaluation.

[3]  Kenneth C. Sevcik,et al.  Implementing Multiprocessor Scheduling Disciplines , 1997, JSSPP.

[4]  David L. Peterson,et al.  Fractal Patterns In DASD I/O Traffic , 1996, Int. CMG Conference.

[5]  Dror G. Feitelson,et al.  Improved Utilization and Responsiveness with Gang Scheduling , 1997, JSSPP.

[6]  Larry Rudolph,et al.  Valuation of Ultra-scale Computing Systems , 2000, JSSPP.

[7]  Sally Floyd,et al.  Wide-area traffic: the failure of Poisson modeling , 1994 .

[8]  Teunis J. Ott,et al.  Load-balancing heuristics and process behavior , 1986, SIGMETRICS '86/PERFORMANCE '86.

[9]  R. Weber.  On the optimal assignment of customers to parallel servers , 1978, Journal of Applied Probability.

[10]  Ronald W. Wolff,et al.  Stochastic Modeling and the Theory of Queues , 1989 .

[11]  Uwe Schwiegelshohn,et al.  Theory and Practice in Parallel Job Scheduling , 1997, JSSPP.

[12]  P. Sparaggis,et al.  Minimizing response times and queue lengths in systems of parallel queues , 1999 .

[13]  ShaikhAnees,et al.  Load-sensitive routing of long-lived IP flows , 1999 .

[14]  Mor Harchol-Balter.  Task assignment with unknown duration , 2002, JACM.

[15]  S. A. Nozaki,et al.  Approximations in finite-capacity multi-server queues by Poisson arrivals , 1978, Journal of Applied Probability.

[16]  Randolph D. Nelson,et al.  An approximation to the response time for shortest queue routing , 1989, SIGMETRICS '89.

[17]  Mor Harchol-Balter,et al.  Task assignment in a distributed system (extended abstract): improving performance by unbalancing load , 1997, SIGMETRICS '98/PERFORMANCE '98.

[18]  Tapani Lehtonen,et al.  On the optimality of the shortest line discipline , 1984 .

[19]  Azer Bestavros,et al.  Self-similarity in World Wide Web traffic: evidence and possible causes , 1996, SIGMETRICS '96.

[20]  Anthony Ephremides,et al.  A simple dynamic routing problem , 1980 .

[21]  Mor Harchol-Balter,et al.  On Choosing a Task Assignment Policy for a Distributed Server System , 1998, J. Parallel Distributed Comput..

[22]  Allen B. Downey,et al.  A parallel workload model and its implications for processor allocation , 1996, Proceedings. The Sixth IEEE International Symposium on High Performance Distributed Computing (Cat. No.97TB100183).

[23]  Azer Bestavros,et al.  Load Profiling in Distributed Real-Time Systems , 1997, Information Sciences.

[24]  Ward Whitt,et al.  Deciding Which Queue to Join: Some Counterexamples , 1986, Oper. Res..

[25]  Sally Floyd,et al.  Wide area traffic: the failure of Poisson modeling , 1995, TNET.

[26]  Anees Shaikh,et al.  Load-sensitive routing of long-lived IP flows , 1999, SIGCOMM '99.

[27]  Robert C. Wang.  Computing optimal quality control policies — two actions , 1976 .

[28]  Felix Pollaczek,et al.  Über eine Aufgabe der Wahrscheinlichkeitstheorie. I , 1930 .