Does repacking improve performance of multiclass loss networks with overflow routing?


This paper considers multiclass loss networks with monoskill and multiskill servers and overflow routing. Accordingly, an arriving job is assigned to a corresponding vacant monoskill server, if possible. Otherwise the job is routed to a multiskill server, or rejected. We derive an efficiently computable upper bound for the system utilization by studying a modified routing where jobs can be redirected from multiskill to monoskill servers. This repacking policy improves performance also in terms of blocking probability when the service requirements of different job classes are statistically identical. Numerical simulations illustrate that our bounds provide good approximations for the performance of the original system.


Keywords: multiclass loss network, stochastic comparison, overflow routing, repacking

AMS subject classification: 68M20, 60K25, 60E15, 90B15, 90B22