Personal mobile grids with a honeybee inspired resource scheduler

Abstract

The overall aim of the thesis has been to introduce Personal Mobile Grids (PMGrids) as a novel paradigm in grid computing that scales grid infrastructures to mobile devices and extends grid entities to individual personal users. In this thesis, architectural designs as well as simulation models for PM-Grids are developed. The core of any grid system is its resource scheduler. However, virtually all current conventional grid schedulers do not address the non-clairvoyant scheduling problem, where job information is not available before the end of execution. Therefore, this thesis proposes a honeybee inspired resource scheduling heuristic for PM-Grids (HoPe) incorporating a radical approach to grid resource scheduling to tackle this problem. A detailed design and implementation of HoPe with a decentralised self-management and adaptive policy are initiated. Among the other main contributions are a comprehensive taxonomy of grid systems as well as a detailed analysis of the honeybee colony and its nectar acquisition process (NAP), from the resource scheduling perspective, which have not been presented in any previous work, to the best of our knowledge. PM-Grid designs and HoPe implementation were evaluated thoroughly through a strictly controlled empirical evaluation framework with a well-established heuristic in high throughput computing, the opportunistic scheduling heuristic (OSH), as a benchmark algorithm. Comparisons with optimal values and worst bounds are conducted to gain a clear insight into HoPe behaviour, in terms of stability, throughput, turnaround time and speedup, under different running conditions of number of jobs and grid scales. Experimental results demonstrate the superiority of HoPe performance where it has successfully maintained optimum stability and throughput in more than 95% of the experiments, with HoPe achieving three times better than the OSH under extremely heavy loads. Regarding the turnaround time and speedup, HoPe has effectively achieved less than 50% of the turnaround time incurred by the OSH, while doubling its speedup in more than 60% of the experiments. These results indicate the potential of both PM-Grids and HoPe in realising futuristic grid visions. Therefore considering the deployment of PM-Grids in real life scenarios and the utilisation of HoPe in other parallel processing and high throughput computing systems are recommended.