The allocation of resources among various project, units. or users is accomplished through the use of a systematic mechanism called resource allocation. The types of resources vary, depending upon the system under consideration. For instance, frequency spectrum and transmitter power might be the resources needed to allocate in an efficient manner on a cellular network system, so that the number of mobile users attended is maximized. On a Grid computing system, one needs to allocate resources such as processors, memory, disk space, and so on, in order that computational tasks run in the most efficient manner.
In order to evaluate resource allocation techniques performances, we first need to evaluate whether a particular resource allocation problem can be cast in the mathematical formu-lation we are exploring. We also need to decide which mechanism will be used, or if a new one needs to be constructed to solve the particular formulation. Finally, we need to evaluate whether the solution obtained is better than those obtained from other techniques that might express and solve the allocation problem in different ways.
In this dissertation, we propose a new resource allocation technique for a system described only by a formulation known as the Multichoice Multidimensional Knapsack Problem, or MMKP. We also propose and evaluate resource allocation techniques on two other sys¬tems: a cellular network and a Grid computing system; in this regard, the resource alloca¬tion problem is not expressed as an MMKP, although the formulations used are particular cases of the MMKP. The MMKP formulation is not applied because its use would not have allowed us to make a fair performance comparison with other more commonly used allo¬cation techniques. However, we believe that as more complex tasks are demanded from systems where resource allocation mechanisms are needed, an MMKP formulation could more suitably represent the allocation problem.
Numerical results indicate that the resource allocation techniques explored in this work present a better performance than previous techniques. Numerical results also indicate that the use of the proposed techniques and the use of suitable optimization criteria can be used to achieve a number of resource allocation goals.
| Identifer | oai:union.ndltd.org:uvic.ca/oai:dspace.library.uvic.ca:1828/1931 |
| Date | 02 December 2009 |
| Creators | Parra Hernández, Rafael |
| Contributors | Dimopoulos, Nikitas J. |
| Source Sets | University of Victoria |
| Language | English, English |
| Detected Language | English |
| Type | Thesis |
| Rights | Available to the World Wide Web |
Page generated in 0.0076 seconds