Use this URL to cite or link to this record in EThOS:
Title: Financial market models for the grid
Author: Martínez Ortuno, Fernando
ISNI:       0000 0004 2701 8614
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2011
Availability of Full Text:
Access from EThOS:
Access from Institution:
The existing network of computing devices around the world created by the Internet gives the possibility of establishing a global market for computing power, where anybody connected to this network can acquire computing power or sell his own spare computing resources in exchange for real money. This potential global market for computing power, which does not exist yet, is what we study in this thesis. Specifically, we study the market with both analytic and simulated models. This thesis predicts how a future global market for Grid computing will behave. We give arguments that such a large market, together with its potential indefinite growth, would not be able to scale if it were organized with a central server, and therefore we study a peer-to-peer market model in our simulations. We create a high-level model with the most relevant characteristics of the market, where buyers and sellers trade a single commodity. In our simulations, the parameters of the volume of contracts, proportion of satisfied agents and number of messages in the network achieve stable values in the long run. We also derive analytically the conditions that make the price get stable over time; we then implement these conditions in the simulation as local mechanisms of the market participants, which make the whole system achieve a stable price evolution. We are also confident that, as soon as the Grid market emerges, a parallel market of derivatives will be created as well. This market of derivatives will be important due to the non-storability nature of computing power. We develop a futures market for computing power based on Markov chains, where we initially model the behaviour of each participant with a particular Markov chain, and then we derive a global transition probability matrix that models the market as a whole. Furthermore, we analyse the performance of a futures trader operating in such a market, and we obtain an optimal trading strategy with the use of Markov Decision Processes. We finally develop a stochastic differential equation model that captures the essence of the spot price evolution of computing power observed in our market simulations. This model is based on a previously one proposed for the electricity market, and consists of the use of a Markov regime-switching mechanism in order to model the existence of spikes in the spot price. We then estimate the parameters in the model with the output data of our simulation program; the estimation is carried out both by maximum likelihood and the generalised method of moments.
Supervisor: Harrison, Peter Sponsor: EPSRC
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral