Resource Pricing In A Dynamic Multi-Commodity Market For Computational Resources

TL;DR

This paper explores a market-based approach to resource pricing in computational grids, using a multi-commodity model with CPU categories to optimize resource allocation through supply-demand equilibrium.

Contribution

It introduces a multi-commodity pricing model with substitutable CPU resources and analyzes equilibrium algorithms under dynamic market conditions.

Findings

Effective algorithms for computing supply-demand equilibrium.

Demonstrated adaptability to dynamic consumer and provider populations.

Improved resource allocation efficiency in grid environments.

Abstract

The adoption of market-based principles in resource management systems for computational infrastructures such as grids and clusters allows for matching demand and supply for resources in a utility maximizing manner. As such, they offer a promise of producing more efficient resource allocations, compared to traditional system-centric approaches that do not allow consumers and providers to express their valuations for computational resources. In this paper, we investigate the pricing of resources in grids through the use of a computational commodity market of CPU resources, where resource prices are determined through the computation of a supply-and-demand equilibrium. In particular, we introduce several categories of CPUs characterized by their execution speed. These differ in cost and performance but may be used interchangeably in executing jobs and thus represent so-called substitutable resources. We investigate the performance of the algorithms for computing the supply-and-demand equilibrium in this multi-commodity setting under dynamically varying consumer and provider populations.