Evaluating Price Mechanisms for Clouds

Principal Investigator(s): 
Eric Friedman, Scott Shenker

Evaluating Price Mechanisms for CloudsResearchers are studying the problems that arise in cloud computing centers that use economic models to allocate resources. In these clouds, resources, such as storage, processing, and data transfer, must be allocated to different users. In economics-based clouds, artificial economies are set up; each resource is assigned a "price" and each user is given a "budget," which they spend on the resources they need. As in economies, the prices of resources fluctuate based on demand: resources in high demand have high prices, and those in low demand have low prices. In this model, resource allocation is simpler than in traditional models because it is decentralized: prices depend on natural fluctuations in demand and usage. This project looks at two issues that arise in economics-based systems in order to determine whether the simplicity gained by using them is worth the losses in efficiency they may involve.

First, to be efficient, the prices of resources must be reasonably stable. The goal of a cloud computer center is to maximize the total value of the users, and this requires appropriate and stable prices. Instabilities can arise in many ways. For example, if two resources are interchangeable, and one has a lower price, most users will buy it, which causes its price to go up and the other's to go down. This causes users to switch back, which can lead to price oscillations. This issue gets more complicated in real-world settings and requires careful analysis.

Second, users can manipulate the prices in cloud computing economies through what they buy. For example, users can drive up prices on a critical resource for a competitor, allowing them to buy other resources more cheaply.

In this project, researchers will study how significant these problems are, using methods from classical economics and contemporary computer science. Much of the research will rely on mathematical models and simulations.

Funding provided by NSF grant 1216073, ICES: Small: Evaluating Price Mechanisms for Clouds.