Martin wrote: > Please excluse that my comment is cross-posted to both the Xen and zfs > discuss forums, I didn't notice it on the Xen forum until I had posted. > Anyway, here goes. Any info, testing, etc please contact me. > > > Regarding the following that I also hit, see > http://www.opensolaris.org/jive/thread.jspa?messageID=180995 > and if any further details or tests are required, I would be happy to assist.
> I set /etc/system's zfs:zfs_arc_max = 0x10000000 and it seems better now. > > I had previously tried setting it to 2Gb rather than 256Mb as above without > success... I should have tried much lower! > > It "seems" that when I perform I/O though a WindowsXP hvm, I get a > "reasonable" I/O rate, but I'm not sure at this point in time. When a write > is made from within the hvm VM, would I expect for the same DMA issue to > arise? (I can't really tell either way aty the moment because it's not super > fast anyway) A couple of things here.. I have found that ~ 2G is good for a dom0 if you use zfs. If your using zfs in dom0, you should use a dom0_mem entry in grub's menu.lst and fix the amount of memory dom0 starts with. e.g. my entry looks like... title 64-bit dom0 root (hd0,0,d) kernel$ /boot/$ISADIR/xen.gz dom0_mem=2048M com1=9600,8n1 console=com1 module$ /platform/i86xpv/kernel/$ISADIR/unix /platform/i86xpv/kernel/$ISADIR/unix -k module$ /platform/i86pc/$ISADIR/boot_archive Then, make sure you don't auto-balloon dom0 down (by creating guests which would take some of that memory from dom0). The best way to do this is to set config/dom0-min-mem in xvm/xend (e.g. svcprop xvm/xend). zfs and auto-ballooning down don't seem to work great together (I haven't done a lot of testing to characterize this though). If you do want to auto-balloon, or want to use less memory in a zfs based dom0, setting zfs_arc_max to a low value seems to work well.. I/O performance in a Windows HVM guests is not good at this point. It won't be until we have Windows PV drivers available. Thanks, MRJ _______________________________________________ xen-discuss mailing list [email protected]
