Output details
11 - Computer Science and Informatics
Imperial College London
Model-Driven System Capacity Planning under Workload Burstiness
<07>The paper studies queueing models of distributed systems processing workloads affected by burstiness. Previously, these models were thought intractable and thus studied only by approximations or
simulation. We found instead linear programming methods to analyse them without approximation. The proposed methods address the state explosion of the underlying Markov process and return provable upper
and lower bounds on the system steady-state performance. The paper extends a work presented at ACM SIGMETRICS 2008 and it was later generalised in a paper at IEEE/IFIP DSN 2010. In the collaboration, I have driven the theoretical part and authored the theorems.