Monitoring the execution of query plans

Monitoring the execution of a query plan is the activity concerned with the collection of information that becomes available from the completed or on-going parts of the execution. Monitoring can be classified in three categories according to the transmission of data and messages that are related to the monitoring procedure. First, monitoring can be performed at the level of physical operators that comprise the query plan. In that case, no data is conveyed, and consequently, there is no communication overhead. Second, monitoring may require a set of operators of the query plan on a node to communicate with each other. In that case, the communication overhead can remain very low. Third, operators or sets of operators on different nodes may send data to each other in order to monitor aspects of the query plan. The third case applies to situations in which the query plan is executed in a parallel or distributed setting. In the last two cases, the monitoring incurs communication overhead. In all cases there is a CPU overhead. In order to have a complete view about the quality of the execution of the query plan, apart from the query plan, the execution environment needs to be monitored as well (which can be seen as a fourth case of monitoring). In this report, the focus is on the first case of monitoring, in which no data is conveyed. The improvements in the quality of the derived information when monitoring falls in the other two categories are also discussed.