Fair access to scarce resources in ad-hoc grids using an economic-based approach

Abstract

In ad-hoc Grids where the availability of resources and tasks changes over the time, distributing the tasks among the scarce resources in a balanced way is a challenging task. In this paper, we present how using an economic-based approach, the tasks can be distributed evenly among the scarce resources. Such that all nodes which have tasks to be executed, receive more or less equal utilization from the Grid resources. We consider a continuous double auction protocol as the economic approach to share idle cpu cycles among the nodes in a local ad-hoc Grid. Each node is given a limited budget when joining the Grid. The node can use it to buy required resources or increases its budget by selling idle resources. Consumers and producers of resources determine their bid and ask prices using a dynamic pricing strategy and the auctioneer adopts a discriminatory pricing policy which sets the transaction price individually for each matched buyer-seller pair. We experiment in a network where the resources are scarce and the tasks outnumber the resources. The spread of task utilization at individual nodes and the level of load balancing are measured and compared with a non-economic approach. The results show that economic-based approach provides a fair and balanced basis for access to Grid resources for everyone while in the same time gives a satisfactory level of system performance.