A profit maximizing distribution strategy for context-aware services

Abstract

Large scale mobile commerce applications need to be replicated and distributed to offer an appropriate quality of service to a large audience. In this context, a very important question is when and where replicas should be placed. In this article we present a new distribution strategy for applications in content delivery networks (CDNs). It is intended to be used especially for context-aware services (CAS). The main idea is to monetarily value the benefits that an additional replica gains at a certain location and to compare this value with the costs a replica causes. In this way, the replica's contribution to the overall profit is determined. A replica's profit is dependent on the number of requests it receives per time interval. Replicas are only placed at locations where the request rate is high enough to let the profit become positive. In order to react only to persistent and significant changes of the request rate, techniques from statistical process control are used. The new strategy is evaluated by simulation. The results suggest that it consumes significantly less storage and bandwidth than currently applied alternatives, while offering a comparable performance improvement for the clients