Performance impacts of look-ahead execution in the conversation scheme

The look-ahead execution approach, which allows participant processes that exist early to exit from a conversation before other participants finish their conversation activities, is presented here as a fundamental approach to reducing the synchronization overhead. Queuing network models are developed for both the system operating under the basic conversation execution scheme and the system operating under the look-ahead execution scheme. Based on the models, various performance indicators such as the system throughput, the average number of processors idling inside a conversation due to the synchronization required, and the average time spent in conversation, are evaluated numerically for different application environments. The performance under the look-ahead execution scheme is compared with that under the basic conversation execution scheme. The results provide insights into the extent of benefits provided by the look-ahead execution approach