The workload on your system can be modelled as a set of activities that arrive at the system at a rate governed by an arrival rate distribution for activities (often measured as its inverse, the inter-arrival time distribution) and the amount of time activities spend executing in the system following a service time distribution.
Inter-arrival time is the time between the arrival of one activity and the arrival of the next activity. Service time is the time that an activity spends executing on the system. For example, if you submit a query at time 0 seconds, it spends 2 seconds in a queue, and it finishes at time 5 seconds, the service time is 5 - 2 = 3 seconds. Service time assumes no other work executing on the system (that is, it is not the observed execution time, but rather the time it would take to execute the activity in isolation). The service time distribution can be approximated for DML activities using the estimated cost in timerons, which considers both processor and I/O time for an activity.