SimpleWorker is a job queuing and scheduling system so while a lot of work comes in at random times usually based on some event in their system (ie: a user clicks a button), there are a lot of scheduled jobs too and those scheduled jobs have a huge affect on the capacity requirements for SimpleWorker. Check out this graph of CPU usage for the past 24 hours across the entire SimpleWorker server farm.
The big spikes start on the hour every hour and this is because most developers schedule their recurring jobs on the hour, not somewhere in the middle of the hour. It would be much better for us if they did so it could spread the load evenly and we could make better use of the unused time and perhaps we will provide incentive to do that at some point, but as it stands, we always have to have enough capacity for the big spikes.
Another thing to note about this information, is that it clearly shows one of the value propositions SimpleWorker provides to our customers. If their jobs only run for a few minutes each hour, they will only pay for those few minutes, not the full hour where the majority of that is unused time.