Peter, I'm going to recommend you to look at the various blog posts by Felipe Franciosi on xenserver.org ( http://xenserver.org/blog/blogger/listings/franciozzy.html). He is one of the performance engineers on the team, and has some pretty deep understanding of why things behave the way they do. There have also been a number of storage performance threads on xs-devel at xenserver.org covering results obtained with Creedence. While some of the observations on't apply to your situation, it will be a good starting point.
-tim On Wed, Dec 3, 2014 at 7:02 PM, Erdősi Péter <[email protected]> wrote: > Hi, > > I faced with a strange problem, when configured iscsi storage one of my > VM-s. > The Vm has 2 dedicated vif (which is connected to 10GB/s NICs), and i > successfully logged in to iSCSI with them. > When I started some dd write test, it's going with 200-230MB/s (on whole > 100GB size, and the VM has 4G mem, so it was not cacheing) > After that, I started read test (simple copy from multipathed block device > to /dev/null) and the read speed is around 30MB/s. > > I know, it's not exactly a CS question, but may somebody faced this kind > of problem too. > > (cause the LUN is 48TB, I cannot add it to CS as primary storage, or if I > do it, I need to add the whole 48T in ~2TB chunks... However if i connect > iSCSI to Xen, the read and write speed is normal...) > > Thanks, > Peter >
