Optimal allocation of programs and data in distributed systems (abstract only)
Amjad Umar · 1985
The problem of allocating programs and data to minimize total cost (communication cost + local processing cost + storage cost) subject to storage, availability, response time and policy/security constraints is formulated and solved. The problem formulation takes into account the complex interactions between data, programs and transactions, and includes the effect of update synchronization algorithms, query processing options, network topologies and queuing delays due to device-busy and locking conflicts. A wide variety of operations research techniques are integrated into a single optimization algorithm which yields, in polynomial time, exact optimal solutions for simplified cases and near optimal solutions for more complex situations. The solution approach lends itself very easily to application of “distributed” optimization, where portions of the algorithm can be executed in parallel by several computers, thus significantly reducing the total optimization time.