Description of problem: In current glusterfs rebalance,at beginning of rebalance a thread pool is created equal to the number of cores present on that machine and based on the throttle configuration, few stay awake and others go into sleep mode (until asked to wake up again). But this behavior can be changed where the threads are created on demand, based on the user configuration.
Currently the focus is mostly on getting overall stability, and better performance on basic high availability focus! This would be picked up later!