Great plug-in, thank you so much for all the great work!
The time interval from when a job lands in the build queue until a new slave node is started to handle the job is rather long. It would be useful if we could tune how frequently the build queue is checked and how quickly new slaves are started up so we can reduce the wait time.
There is an option for nodelayprovisioning which starts slaves as soon as required without waiting. The standard strategies have logic around when to start slaves.
Don't forget that to keep the OS small enough to be fast to boot, the generic algorithm used to raise new node is already quite fast. How much is your startup time ?
With the structure that we have we are not able to reduce the startup time, already we start parallel node base on the queue size. More operation will result in a deadlock of the queue.