First, the load has to be polled periodically within acceptable time intervals. Since usually large numbers of computers are involved (ten and more), this cannot be done sequentially, because not all hosts can be polled within one interval. This is due to the fact that each host can take a couple of seconds until it responds. Moreover, sequential polling is highly susceptible to corrupted computers, since it has to wait for each computer's response until it is able to proceed. Hence, the load of each host needs to be polled in parallel using asynchronous communication.
Secondly, the load which is reported by an operating system, is usually subject to a delay. This delay is not acceptable, since it results in an unstable system which will lead to an oscillation of the computers workload. Therefore, the reported load needs to undergo some processing in order to account for this delay.