2020/05/27 by Youri Raaijmakers, Raaijmakers, Youri, Sem Borst +3
Business, Management and Accounting · Computer Science · #Advanced Queuing Theory Analysis #Age of Information Optimization #Distributed systems and fault tolerance #FOS: Computer and information sciences #FOS: Mathematics #Performance (cs.PF) #Probability (math.PR)
paper · pdf · doi:10.48550/arxiv.2005.13353
openalex publication_date 2020/05/27 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
We consider a system with several job types and two parallel server pools.\nWithin the pools the servers are homogeneous, but across pools possibly not in\nthe sense that the service speed of a job may depend on its type as well as the\nserver pool. Immediately upon arrival, jobs are assigned to a server pool. This\ncould be based on (partial) knowledge of their type, but such knowledge might\nnot be available. Information about the job type can however be obtained while\nthe job is in service; as the service progresses, the likelihood that the\nservice speed of this job type is low increases, creating an incentive to\nexecute the job on different, possibly faster, server(s). Two policies are\nconsidered: reroute the job to the other server pool, or replicate it there.\n We determine the effective load per server under both the rerouting and\nreplication policy for completely unknown as well as partly known job types. We\nalso examine the impact of these policies on the stability bound, and find that\nthe uncertainty in job types may significantly degrade the performance. For\n(highly) unbalanced service speeds full replication achieves the largest\nstability bound while for (nearly) balanced service speeds no replication\nmaximizes the stability bound. Finally, we discuss how the use of\nthreshold-based policies can help improve the expected latency for completely\nor partly unknown job types.\n