Block-based grid caching for grid datafarm

Conventional grid cache follows web-based caching concept, which stores the whole object as one piece in the cache storage. Although this object-based approach can be very effective for web-based files, which are typically small, grid data files are usually large. Thus, conventional grid cache performs rather poor and becomes less-effective as the cache storage is not well-managed. This paper presents a new block-based grid caching model which divide caching space into fixed-size cache blocks for storing data. We evaluate the performance of our block-based approach and the impact of different block sizes using real-world data on our ns-based simulator. The results strongly indicate that our block-based grid caching perform reasonably well comparing to the conventional file-based grid caching.