We present an exact solution technique for analyzing the performance of such applications. Formally, the problem can be stated as follows: A job arrives at a system with \(M\) identical servers in a Poisson fashion. After some initial computation, the job spawns \(K\) statistically identical subtasks. All these subtasks can be executed in parallel but only one of them is required to complete in order to complete the entire job. The service times are assumed to be exponentially distributed. The solution is obtained by defining a set of partial generating functions and exploiting their properties. The solution technique appears to be applicable for the general problem, but we have the formal proofs of correctness only for the cases where \(K=M\) and when the mean service demands of the initial computation and the subtasks are the same. For these cases, the results show that the average number of busy processors is independent of \(K\) while the job response time decreases with \(K\). Some conjectures are presented for more general situations.