In this paper, the authors define a network model as a constrained queueing network. This is a stochastic queueing network system with service-level constraints. Packets arrive at each queue buffer in the network via an exogenous stochastic process and leave the system after being served. At most one packet in each nonempty queue can be served at a time, and all packets have a unit service time. However, because of service constraints, not all nonempty queues can transmit their packets simultaneously, and only certain subsets of the queues can serve packets at the same time. The authors call these subsets schedules, and every constrained queueing network has its own collection of schedules. In this context, they study how to design efficient scheduling algorithms that require little computational time at each service epoch while still maintaining high performance. Here the performance metric is related to the number of packets in each queue. Specifically, they present a generic framework for designing a throughput-optimal and low-complexity algorithm that finds its current schedule via interaction with an oracle system.
Recensione dell'articolo:(Suk, T.; Shin, J. - " Scheduling using interactive optimization oracles for constrained queueing networks. " - Math. Oper. Res. 42 (2017), no. 3, 723–744.) MR3685263 MathSciNet ISSN 2167-5163
Pasini Leonardo
2018-01-01
Abstract
In this paper, the authors define a network model as a constrained queueing network. This is a stochastic queueing network system with service-level constraints. Packets arrive at each queue buffer in the network via an exogenous stochastic process and leave the system after being served. At most one packet in each nonempty queue can be served at a time, and all packets have a unit service time. However, because of service constraints, not all nonempty queues can transmit their packets simultaneously, and only certain subsets of the queues can serve packets at the same time. The authors call these subsets schedules, and every constrained queueing network has its own collection of schedules. In this context, they study how to design efficient scheduling algorithms that require little computational time at each service epoch while still maintaining high performance. Here the performance metric is related to the number of packets in each queue. Specifically, they present a generic framework for designing a throughput-optimal and low-complexity algorithm that finds its current schedule via interaction with an oracle system.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.