Developing a trusted computational grid
Brennan, John David
Within institutional computing infrastructure, currently available grid mid-dlewares are considered to be overly complex. This is largely due to behaviours required for untrusted networks. These behaviours however are an integral part of grid systems and cannot be removed. Within this work the development of a grid middleware suitable for unifying institu-tional resources is proposed. The proposed system should be capa-ble of interfacing with all Linux based systems within the QueensGate Grid (QGG) campus grid, automatically determining the best resource for a given job. This allocation should be done without requiring any ad-ditional user effort, or impacting established user workﬂows. The frame-work was developed to tackle this problem. It was simulated, utilising real usage data, in order to assess suitability for deployment. The re-sults gained from simulation were encouraging. There is a close match between real usage data and data generated through simulation. Fur-thermore the proposed framework will enable better utilisation of cam-pus grid resources, will not require modiﬁcation of user workﬂows, and will maintain the security and integrity of user accounts.