ecms_neu_mini.png

Digital Library

of the European Council for Modelling and Simulation

 

Title:

A Simple Dispatching Policy For Minimizing Mean Response Time In Non-Observable Queues With SRPT Policy Operating In Parallel

Authors:

Mikhail Konovalov, Rostislav Razumchik

Published in:

 

 

2020). ECMS 2020 Proceedings Edited by: Mike Steglich, Christian Muller, Gaby Neumann, Mathias Walther, European Council for Modeling and Simulation.

 

DOI: http://doi.org/10.7148/2020

ISSN: 2522-2422 (ONLINE)

ISSN: 2522-2414 (PRINT)

ISSN: 2522-2430 (CD-ROM)

 

ISBN: 978-3-937436-68-5
ISBN: 978-3-937436-69-2(CD)

 

Communications of the ECMS , Volume 34, Issue 1, June 2020,

United Kingdom

 

Citation format:

Mikhail Konovalov, Rostislav Razumchik (2020). A Simple Dispatching Policy For Minimizing Mean Response Time In Non-Observable Queues With SRPT Policy Operating In Parallel, ECMS 2020 Proceedings Edited By: Mike Steglich, Christian Mueller, Gaby Neumann, Mathias Walther European Council for Modeling and Simulation. doi: 10.7148/2020-0398

DOI:

https://doi.org/10.7148/2020-0398

Abstract:

Consider a non-observable system with a single dis-patcher and N ≥ 2 single server queues operating in parallel and independently. Each queue uses shortest remaining processing time discipline for scheduling the jobs. Jobs from a single flow arrive one by one to the dispatcher, which must immediately route them to one of the queues. Each decision is irrevocable. The dispatcher does not have any online information about the system and can base its decisions only on the job size distribution, job’s inter-arrival time distribution, server’s speeds, time instants of previously arrived jobs and pre-vious routing decisions. Under these conditions, one is interested in the routing policies, which minimize the job’s long-run mean response time. New simple single-parameter policy is proposed which is applicable in case of i.i.d. arrivals and i.i.d. service times and which, according to the numerical experiments, always outper-forms the optimal probabilistic policy and may outperform the deterministic policy (under medium and low load).

 

Full text: