Scheduling and queuing models for dynamic resource environment, differentiated and mobile services
Raj S. Acharya, John Tsiligaridis · 2003
In the essential area of Network Management we have dealt with the subject of the scheduler reaction and adjustment in a dynamic resource environment. A number of scheduling models are proposed. A new approach for adjustment of the weights of the Deficit Round Robin (DRR) is developed. We are dealing with the service of various large size packets and we use gradual or direct weight increases that result to the decrease of the borrowing of credits. A family of algorithms, with O(1) complexity, that use different and systematic weight request methods, is presented. The problem of the design and adjustment of the weights is examined. Periodically instantly changes of the size of quanta optimize the performance of the RR scheduler. Our solution basically enhances the service ability of the DRR server, avoiding the “empty rounds” and additionally offers a framework for temporary economical adjustment of the weights at the routers using various types of RR schedulers. Lightweight mechanisms are included in the differentiated services architecture in order to provide DiffServ in the Internet. A lightweight scheme of Premium service based on the delay ratio criterion is provided. A conservative control mechanism is periodically developed to ensure the stability of the delay ratio. We expand the model into three queues and the protocols of home domain. In order to avoid the sorting operation, in the various packet-scheduling policies, a new algorithm, called the Left and Right Move Algorithm (LRMA) is proposed. The LRMA can be considered as an implementation tool especially useful for sorted-priority algorithms (like SFQ) diminishing their complexity to O(1), and as well as frame-based server for ATM networks. A new prediction scheme is developed with the support of the Data Mining technique. Based on the mobility prediction, taking into account the time factor, the bandwidth is reserved for the paths with the maximum support, so that the service of the handoff calls can be guaranteed. The nodes using the Steps Forward with Distance Algorithm (SFDA) and the Direct Group Method (DGM) enable the BSs to avoid the congestion to the handoff users' benefit.