In this paper we have presented algorithms for predicting the wait time of a job after it reaches the head of the queue. We have not addressed the question of how long it will take to get there. Predictor B can be extended in a natural way to predict the behavior of jobs in queue, and thus to predict the entire queue time for a new arrival. We have implemented this extension and seen promising preliminary results. Predictor B can also be extended to handle non-FCFS scheduling strategies, such as the priority scheme used at SDSC.
As part of the Distributed Object Computation Testbed (DOCT [13]) project, we are implementing a persistent Legion [7] object that monitors the load on space-sharing parallel computers and answers queries regarding expected queue times. Application-Level Schedulers (AppLeS) will use this information for external resource selection [2].