DeepAI AI Chat
Log In Sign Up

Characterization of the Gittins index for sequential multistage jobs

by   Samuli Aalto, et al.

The optimal scheduling problem in single-server queueing systems is a classic problem in queueing theory. The Gittins index policy is known to be the optimal preemptive nonanticipating policy (both for the open version of the problem with Poisson arrivals and the closed version without arrivals) minimizing the expected holding costs. While the Gittins index is thoroughly characterized for ordinary jobs whose state is described by the attained service, it is not at all the case with jobs that have more complex structure. Recently, a class of such jobs, the multistage jobs, were introduced, and it was shown that the computation of Gittins index of a multistage job reduces into separable computations for the individual stages. The characterization is, however, indirect in the sense that it relies on the recursion for an auxiliary function (so called SJP function) and not for the Gittins index itself. In this paper, we answer the natural question: Is it possible to compute the Gittins index for a multistage job more directly by recursively combining the Gittins indexes of its individual stages? According to our results, it seems to be possible, at least, for sequential multistage jobs that have a fixed (deterministic) sequence of stages. We prove this for sequential two-stage jobs that have monotonous hazard rates in both stages, but our numerical experiments give an indication that the result could possibly be generalized to any sequential multistage jobs. Our approach, in this paper, is based on the Whittle index originally developed in the context of restless bandits.


page 1

page 2

page 3

page 4


Optimal Scheduling and Exact Response Time Analysis for Multistage Jobs

Scheduling to minimize mean response time in an M/G/1 queue is a classic...

Scheduling for Multi-Phase Parallelizable Jobs

With multiple identical unit speed servers, the online problem of schedu...

Scheduling to Optimize Sojourn Time of Successful Jobs

Deep neural networks training jobs and other iterative computations freq...

Towards Optimality in Parallel Scheduling

To keep pace with Moore's law, chip designers have focused on increasing...

Covert queueing problem with a Markovian statistic

Based on the covert communication framework, we consider a covert queuei...

Learning to Schedule

This paper proposes a learning and scheduling algorithm to minimize the ...

10-millisecond Computing

Despite computation becomes much complex on data with an unprecedented s...