On Dependent Data Fetching in Cloud Environment

This work was motivated by the study of the file fetching process in a cloud system, in particular by the recent progress in the model of transparent computing. A transparent computing system may have many clients, each requesting a significant set of files from the server, including user data and many commonly used softwares (operating systems and apps). These files may have inherent dependence relations so should be received by the clients in a specific topological order. On the other hand, since many of these files are commonly used softwares, many clients may request copies of the same files. This proposes an interesting problem on the server side of how this kind of requests should be handled efficiently to improve the performance of the system. In particular, we are interested in the processes that significantly reduce the disk IO operations in the server, which are in general very time-consuming. We propose a formal model for this problem and study its validity and correctness. Heuristic algorithms for the problem are proposed and studied. Simulation results are presented to compare the proposed heuristics and algorithms based on known techniques in scheduling literature. 7% - 20% of the total disk IO can be reduced via the optimizations proposed in this work