Parallel Slackness
•Valiant called the amount of “excess” parallelism needed to cover latency parallel slackness •In the “best case” a parallel slackness of logP is required because in the best case latency will be proportional to log P •Any additional delays require further slackness