Grading and predicting networked application behaviour

The user-perceived quality of an application operating over a communication network has a considerable influence on the usefulness of that application . Intuitively, it may be assumed that this quality will be related to curre nt network performance, but in practice the relationship is often complex a nd difficult to determine. A scheme has been developed whereby the performa nce of network applications can be assessed and empirically graded for vari ous controlled network loading conditions. Given the current network loadin g conditions on an operational network and information generated by the gra ding process, it is possible to pr-edict the performance of the network app lication before the application is actually run. This has the potential for reducing the amount of traffic forced on a network as a consequence of abo rted connections.