Service provider selection in an open services environment

The authors consider the case of a large number of servers that offer services of a given type, differing only in quality and cost. Each service request may be directed to any of those servers. The authors consider rules for choosing the most suitable server for a given request and present a taxonomy of server selection strategies with respect to both information acquisition and decision making. The authors consider in detail the influence of significantly delayed state information on the selection process and demonstrate by means of simulation models that simply selecting the server with minimal load is worse than random selection. An alternative selection mechanism, performing well for a large range of delays, is developed. The possible contradiction between the socially optimal and individual optimal approach resulting from autonomy in choice of server selection rules is recognized and investigated