A common problem in client-server based systems is the uncertainity of response times as a consequence of variable bandwidth. By offering a predictable service it is possible for the client to obtain some information about the progress in request handling. Even in a system where no guaranteed links are supported.
In this thesis a server is presented that is either able to answer requests in a given time defined by the client or can detect and inform clients about delays to be extpected as early as possible, e.g. before the limit on the response time has been reached.
To perform delivery of required data within the given time, data has to be adapted dynamically to the state of the system. If bandwidth between server and client goes down data has to be reduced, if bandwidth raises data can be expanded. As reduction of data can be fairly CPU intensive, a mechanism for the resource management has to be added.
This thesis covers developement, implementation and experimental validation of a concurrent adaptive server, which is able to distribute request handling within a cluster of workstations.