On the mediation of program allocation in high-demand environments

TL;DR

This paper proposes an automatic, server-side method for determining the number of processors in distributed cloud computations, challenging the traditional user-specified approach and analyzing its computational complexity.

Contribution

It introduces a weakly NP-hard allocation problem and a cost model for automatic processor determination in cloud environments.

Findings

The allocation problem is weakly NP-hard, solvable in pseudo-polynomial time.

Automatic processor determination can be effectively modeled and implemented.

The approach critiques the user-driven sizing paradigm in cloud computing.

Abstract

In this paper we challenge the widely accepted premise that, in order to carry out a distributed computation, say on the cloud, users have to inform, along with all the inputs that the algorithm in use requires, the number of processors to be used. We discuss the complicated nature of deciding the value of such parameter, should it be chosen optimally, and propose the alternative scenario in which this choice is passed on to the server side for automatic determination. We show that the allocation problem arising from this alternative is NP-hard only weakly, being therefore solvable in pseudo-polynomial time. In our proposal, one key component on which the automatic determination of the number of processors is based is the cost model. The one we use, which is being increasingly adopted in the wake of the cloud-computing movement, posits that each single execution of a program is to be subject to current circumstances on both user and server side, and as such be priced independently of all others. Running through our proposal is thus a critique of the established common sense that sizing a set of processors to handle a submission to some provider is entirely up to the user.