Managers listen for network RAM requests from clients (including the client residing on the same node as the manager). When a request is received, a new entry for the parallel process represented by the thread is created in a request table. Depending on the PNR strategy, the manager may immediately act on this request or wait for other requests to come in before acting. Most strategies require that all threads within a parallel job must contact the same manager before the manager is allowed to act on any one of the threads' requests. For this to work, it is assumed that the PNR manager can discover the total number of threads belonging to the parallel job. For instance, the threads can send this information along with their requests. In any case, as each new request comes in, information on the aggregate memory request is stored in a request table.