On Tue, Oct 01, 2013 at 04:58:26PM +0000, Chiradeep Vittal wrote: > [Apologize for the crosspost, but I think there may be answers from the dev > perspective] > [Also, this is poll / opinion question] > > What is the maximum latency / minimum bandwidth folks have between their > Management Server and the hypervisor? > What works? What needs to be tweaked? > > TIA
FWIW, $dayjob has taken the approach of mgmt server clusters being geographically local to the zones under management (so sub 5ms typically). Bandwidth requirements from mgmt servers to hosts is actually so small in the aggregate, that it's not even a consideration really. Production traffic / storage IO / secondary storage activity are what drive the BW requirements, and mgmt traffic gets to come along for the ride. The regions feature makes it perfectly reasonable to have multiple mgmt clusters, and frankly they have a small footprint... so there isn't much downside to this approach. Why push the limits of latency tolerance if not required. ;-) This doesn't directly answer you question, but it's one approach. -chip