Wireless LANs not only provide an effective means of communication, but also allow to extract information about the location of mobile stations. Numerous Wireless LAN location systems have been proposed in the past, yet it is difficult to compare the performance of different systems, since the conditions under which these systems are evaluated differ considerably. Hence, the accuracy information presented in literature varies widely, even for conceptually identical systems. This paper proposes to evaluate the performance of location systems in standardised test environments. To this end, existing location systems are discussed to assess their information requirements and their reported accuracies. Subsequently, the requirements for benchmarks in the context of Wireless LAN location systems are established. To conclude, a publicly available benchmark is presented against which different Wireless LAN location systems can be compared.

Abstract. As mobile agents have the ability to operate autonomously and in a disconnected way they are considered to suit mobile computing environments. Mobile users can dispatch agents into the fixed network where the agents operate in the users behalf. Thus, in contrast to client/server interactions agents do not suffer from poor performing wireless access networks. In this paper the performance of mobile agents and client/server interactions are analysed with respect to heterogeneous networks and server resources. It is argued that without a certain knowledge of the available resources agents can hardly decide whether they should migrate or just apply client/server calls to access a remote service. To this end, it is proposed that agents should access server selection systems in order to plan their migration strategy. However, while server selection systems process agent requests the agents are waiting idle. Thus, access to server selection systems comes at a cost and therefore agents must be careful about it. To solve this decision problem an algorithm is proposed which estimates the benefits of accessing server selection systems. Finally, the decision algorithm is evaluated with the help of a simulation model.

No abstract

Abstract. Since agents have the ability to migrate to outperforming resources they can potentially balance the load of heterogeneaus systems. However, to balance resources efficiently agents must take the load into account. Thus, to support the agent migration strategy the application of server selection systems has been proposed recently. Server selection systems keep track of the load of network and host resources and hence predict the performance of different migration strategies. Yet, server selection comes at a cost and therefore agents must take care when applying it. This paper presents a decision strategy for the agent's decision problem. The performance of the approach is analysed with the help of a simple queuing model. lntroductionBasically, the advantage of mobile agent technology is that it allows application designers to decide where an agent is processed. In that sense, developers can optimise the performance of an application by carefully selecting processing resources. Yet, approaches proposing an automatic selection of destination systems [1] [2] [3] have not gained much attention. A similar problern is server selection in the Internet. The deployment of mirrar sites has motivated users to select a mirrar affering the best performance. With gaining significance of bulk document, audio, and video file transfer, smart mirrar site selection has become a compelling task and thus motivated numerous automatic server selection approaches. Lately, we discussed in [4] the application of server selection to mobile agents. It has been pointed out that due to the rather small resource requirements of mobile agents, these must be careful when applying server selection. That is, since agents are idle while the server selection system processes their requests, server selection comes at a cost. Thus, agents will only access a server selection system if the utility function U is positive:(1) where d0 is the average service time, drnin is the service time of a server recommended by a selection system, and 8 is the agent idle time while the selection system processes its request. In the remainder of the paper, this problern will be [4] and analyses the performance of the algorithm with the help of a queuing model.

No abstract