Optimality of full replication for log-convex service distributions
Let be the number of servers, let be the survival function of a task's service time, and let denote the service-time distribution. For a fork-join system, let be the job latency and be the total computational cost, with each job forked to servers. Full-replication optimality conjecture. For log-convex and any arrival rate , forking a job to all servers is optimal in both and . The preceding discussion establishes the corresponding high-traffic behavior and observations suggest that the claim holds for every arrival rate, but no proof or resolution is supplied here.
References
- Gauri Joshi, Emina Soljanin and Gregory Wornell, “Efficient Replication of Queued Tasks for Latency Reduction in Cloud Systems”, arXiv:1510.04731 (2015).
Progress summary
Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.
Solutions 0
No solutions have been posted yet.