A study is made of the partitioning of a global database and the allocation of the partitions in a bus computer system. The authors deal with systems where database partitioning and the allocation of the partitions to processing nodes can be solved simultaneously. The combined decision is referred to as data allocation. The objective is to allocate data so that the mean transaction response time is minimized. A queuing optimization model for solving the problem is formulated as a nonlinear programming problem. The uniqueness of the optimal policy is proved and an optimization procedure is developed. The model is illustrated with a sample system.
Proc. of the IEEE Int. Conf. on Computers and Communications, pp.394 - 399