Memory consumption, like always, is dependent on the required bandwidth in terms of parallel robots needed for a speciﬁc project.
However, some robots consume more memory than others. Large Excel spreadsheets for example or large XML ﬁles consume incredible amounts of memory.
One major pitfall in terms of memory footprint is that RoboServer is not restricted by the memory assigned to the JVM, when you run WebKit robots. That means that where you could be led to believe that conﬁguring a dedicated RoboServer-machine to let the RoboServer JVM consume all the available physical memory is a good approach, in fact it is not (if you want to run any WebKit robots). You need to leave some physical memory available for the kapowbrowser processes.
Between 9.3 and 9.4 the memory footprint (in terms of virtual memory) of the individual kapow (WebKit) browser process was signiﬁcantly reduced.
Kapow 9.5 features monitoring (through Management Console) of the individual RoboServer’s total memory and CPU consumption, including the memory consumed by the kapowbrowser sub-processes. That should help system integrators determine at which velocity robots can be safely run and/or how much memory to assign to the machine running RoboServer.
Keywords: Kapow memory, bandwidth, parallel, WebKit, KapowbrowserCPU