A Study And Implementation Of Resource Usage Accounting In Grid System

Computational grid is an open distributed system, which integrates abundant geographic distributed HPC resources by various grid technologies and is capable of large scale computing. It is significantly worthy to accounting and calculating resource usage in a large grid system, since it can provide knowledge of resource contribution and usage, as well as the steps to avoid waste. Meanwhile, we can use accounting to archive load balance on every computing site. A study of accounting system not only involves metering resource usage such as CPU time, wall time, memory usage, etc., but also includes study of computing market and pricing mechanism. This has tremendous effect on development of grid computing.Our study is focused on resource usage accounting and implementation of an easy-to-use and lightweight grid accounting system. Firstly, we introduce the background and previous research result in grid accounting area. Secondly, we promote the three key issues of our consideration, which are timely accounting, an implementation of OGF-RUS, and easy deployment configuration, management and using. Thirdly, based on these works, we depict the architecture of our accounting system. We invent a line based parsing framework to get logs parsed timely. Then we make an implementation of OGF-UR and OGF-RUS to archive inter-operation. We propose SSQT algorithm as our solution to large amount of XML data storage, and that innovation is a big help to the implementation of resource usage service. At last, some testing and analyze on our system is present to verify the performance improvement as well as easy-to-use features of our implementation.