Modeling Resource Prices in Computational Commodity Markets

In the context of grids, economic principles can be introduced in resource management systems to shift from away a system-centric towards a user-centered focus. Individual valuations of resource usage can be expressed on a per job basis through resource prices. In a commodity market model for CPU resources determination of an equilibrium price balances supply and demand. Because of the possible disruptive effects on operational grids, such approaches need to be investigated with simulators. We use a grid economics simulator that supports the commodity market model to find out whether a temporal neural networks can be used to predict price evolution. This study extends previous work, by including several categories of CPU's differing in their performance levels. Our investigation shows that the neural network is still capable of adequate price predictions in this context.