Abstract.One of the main challenges of emerging high speed telecommunication networks is the integration of services. Both ATM as well as the Internet have been evolving so as to accommodate both best-effort type traffic (such as file transfer) as well as real time traffic requiring guaranteed performance. The co-existence of thesc two service types on the same network carries important hencfits such as resource sharing betwecn service classcs, and the ability of the user to easily select an appropriate service class according to requiremcnts and preferences. These benefits depend on efticient network management and resource sharing strategies. The aim of this paper is twofold. First, we study [he performance measures of both scrvice types, as a function of the nctwork managcmcnt strategy. This study allows us at a second phase to design efficient network managcmcnt schemes. This includes questions such as whcther best-effort trat'fic should use only bandwidth Icft-over by the guarantccd pcrformance ones, or whcthcr (and how much) bandwidth needs to be reserved to best-effort tratfic. Another managcmcnt issue that we study is thc pricing. We allow Tor diffcrent classes of ~uarantced-pcrformancc traffic to have difkrcnt priorities with rcspcct to the rejection probabilities.
We consider the problem of optimal routing of messages into two queues. The aim is to minimise some cost, composed of a holding cost which b linear in the queue length0 plus an admionion cost. As often happens in high-speed communication networks, we srsume that the decbion maker has a delayed information on the state of the network. We study this problem in the framework of Markov Decision Processes with a countable state space and unbounded cost; both the discounted and the average cost are considered. We establish some supermodularity prop erties of the value function using "value-iteration" arguments, this in turn enables us to characterise the optimal policy. It is shown to be monotone, and is characterised by some switching curves.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.