Efficient Data Allocation Heuristics for Data Center Server Consolidation

摘要

Data centers have been widely used for many data intensive applications. A data center consists of multiple servers, where data are distributed and stored. Users connect to data centers to access data frequently. Initially, data should be loaded to the data center. Server consolidation is a common strategy for data center to reduce costs, by consolidating resources on as few servers as possible. In this paper, we investigate the data allocation for server consolidation in a data center to minimize the number of active servers. We formulate the data allocation problem as Integer Programming. We make use of a recently developed technique of Vector Bin Packing and propose both offline and online heuristic algorithms. To further speedup the data allocation, we also design a grouping based approach to reduce the total number of iterations in the algorithm. Performance evaluation shows our approaches can efficiently allocate data.