Exploiting Availability Prediction in Distributed Systems

  • James Mickens ,
  • Brian D. Noble

Proceedings of NSDI |

Published by USENIX

Loosely-coupled distributed systems have significant scale and cost advantages over more traditional architectures, but the availability of the nodes in these systems varies widely. Availability modeling is crucial for predicting per-machine resource burdens and understanding emergent, system-wide phenomena. We present new techniques for predicting availability and test the musing traces taken from three distributed systems. We then describe three applications of availability prediction. The first, availability-guided replica placement, reduces object copying in a distributed data store while increasing data availability. The second shows how availability prediction can improve routing in delay-tolerant networks. The third combines availability prediction with virus modeling to improve forecasts of global infection dynamics.