<html><body><div style="color:#000; background-color:#fff; font-family:HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif;font-size:12pt"><div id="yiv8225798884"><div><div class="" style="color:#000;background-color:#fff;font-family:HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif;font-size:12pt;"><div id="yiv8225798884yui_3_13_0_ym1_20_1389772026800_8"><span id="yiv8225798884yui_3_13_0_ym1_20_1389772026800_14">Dear Carlo,<br><br>Thanks for your suggestions, I will add my comments and details below regarding the various performance points you mention.<br><br>So yes I am using QXL as video driver, for that purpose I have added to my ONE template the following RAW data:<br></span><br><span id="yiv8225798884yui_3_13_0_ym1_20_1389772026800_14"><devices><video><model type="qxl" heads="1"></model></video></devices><br><br>As far as I remember AIO=native is the default and as
such I should already be using native AIO.<br><br>As data storage I am using an NFS server with a SATA JBOD in hardware RAID6 (HP hardware). I am sharing the datastores from that server using the following export:<br><br>/data/one/datastores/101 192.168.1.0/24(rw,sync,no_subtree_check,root_squash)<br><br>Then on my ONE hosts I mount them using the following fstab entry:<br><br>192.168.1.2:/data/one/datastores/101 /var/lib/one/datastores/101 nfs soft,intr,rsize=8192,wsize=8192<br><br>Regarding the cache setting in the ONE template I always use none as I read quite often that in most of the cases cache should be set to none.<br><br>Using iotop (great tool btw!) to measure IO activity and using the image at the same time does not show much IO activity so I don't think IO is the bottleneck, also it is the only VM I am running on a test setup.<br><br>Other info to mention maybe: I am using a test VM with 4 GB of RAM
and 4 CPU running on a host with dual Quad-Core Intel E5504 @2GHz (approx. 3 years old CPU). The resolution I am using for the desktop is 1680x1050. For your reference, below is the full KVM process which is running on the ONE host:<br><br>/usr/bin/kvm -S -M pc-1.1 -enable-kvm -m 4096 -smp 4,sockets=4,cores=1,threads=1 -name one-70 -uuid 8429d2bf-553f-d503-83e4-c44fc626de5c -no-user-config -nodefaults -chardev socket,id=charmonitor,path=/var/lib/libvirt/qemu/one-70.monitor,server,nowait -mon chardev=charmonitor,id=monitor,mode=control -rtc base=utc -no-shutdown -device piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2 -drive file=/var/lib/one//datastores/101/70/disk.0,if=none,id=drive-virtio-disk0,format=qcow2,cache=none -device virtio-blk-pci,scsi=off,bus=pci.0,addr=0x4,drive=drive-virtio-disk0,id=virtio-disk0,bootindex=1 -drive file=/var/lib/one//datastores/101/70/disk.1,if=none,id=drive-ide0-0-0,readonly=on,format=raw -device
ide-cd,bus=ide.0,unit=0,drive=drive-ide0-0-0,id=ide0-0-0 -netdev tap,fd=20,id=hostnet0 -device virtio-net-pci,netdev=hostnet0,id=net0,mac=02:00:b9:0f:1f:29,bus=pci.0,addr=0x3 -spice port=5970,addr=0.0.0.0,disable-ticketing -vga qxl -global qxl-vga.vram_size=67108864 -device virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x5<br><br>Let me know if you need any more infos... Looking forward to your feedback.<br><br>Regards,<br>M.L.<br></span></div><div class="" id="yiv8225798884yqt53956"><div class="" id="yiv8225798884yui_3_13_0_ym1_20_1389772026800_10" style="display: none;"> <br clear="none"> <br clear="none"> <div class="" style="font-family:HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif;font-size:12pt;"> <div class="" style="font-family:HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif;font-size:12pt;"> <div dir="ltr"> <font face="Arial" size="2"> On Tuesday, January 28, 2014 4:50 AM, Carlo Daffara
<carlo.daffara@cloudweavers.eu> wrote:<br clear="none"> </font> </div> <div class="">There
are several things you can do to improve basic performance. First of all, if you use SPICE as a protocol use QXL as the KVM video card, it will provide a substantial<br clear="none">speedup (the emulated video card will be much more intelligent and the job of protocol compression will be much simpler). Then, IO speedup is the second low-hanging fruit. We<br clear="none">use aio=native, and this alone substantially increases io rates. The next step depend on the kind of datastore you use- NFS, an external SAN, etc. If you can provide some additional details I can give some more options. For example, for machines that require high random R/W we use cache=none as it does decrease double buffering, thus improving performance. <br clear="none">You can use iotop on the machine that executes the KVM process to check how much does it write, and why.<br clear="none">In our tests, using OpenNebula on top of the MFS distributed filesystem we easily get VDI
performance near to that of the hardware with limited tuning.<br clear="none">regards,<br clear="none">carlo daffara<br clear="none">CloudWeavers<br clear="none"><div class="" id="yiv8225798884yqtfd17117"><br clear="none">----- Messaggio originale -----<br clear="none">Da: "ML mail" <<a rel="nofollow" shape="rect" ymailto="mailto:mlnospam@yahoo.com" target="_blank" href="mailto:mlnospam@yahoo.com">mlnospam@yahoo.com</a>><br clear="none">A: "users" <<a rel="nofollow" shape="rect" ymailto="mailto:users@lists.opennebula.org" target="_blank" href="mailto:users@lists.opennebula.org">users@lists.opennebula.org</a>><br clear="none">Inviato: Luned́, 27 gennaio 2014 16:51:19<br clear="none">Oggetto: [one-users] Using OpenNebula for Linux desktops<br clear="none"><br clear="none">Hi, <br clear="none"><br clear="none">I am currently evaluating OpenNebula for running KVM virtual machines which will be used as Linux desktops with the SPICE protocol. For
now I have tried Fedora 20 as a desktop system with KDE and
SPICE (using remote viewer) but I am quite disappointed with the performance. The desktop reacts quite slowly even basic things like displaying the main menu and using the terminal even on a 1 Gig LAN. <br clear="none"><br clear="none">So does anyone out there really use OpenNebula for this purpose with success? and are there any recommendations for speeding up things in order to make it really usable. <br clear="none"><br clear="none">Cheers, <br clear="none">M.L. </div><br clear="none"><br clear="none">_______________________________________________<br clear="none">Users mailing list<br clear="none"><a rel="nofollow" shape="rect" ymailto="mailto:Users@lists.opennebula.org" target="_blank" href="mailto:Users@lists.opennebula.org">Users@lists.opennebula.org</a><br clear="none"><a rel="nofollow" shape="rect" target="_blank"
href="http://lists.opennebula.org/listinfo.cgi/users-opennebula.org">http://lists.opennebula.org/listinfo.cgi/users-opennebula.org</a><br clear="none">_______________________________________________<br clear="none">Users mailing list<br clear="none"><a rel="nofollow" shape="rect" ymailto="mailto:Users@lists.opennebula.org" target="_blank" href="mailto:Users@lists.opennebula.org">Users@lists.opennebula.org</a><br clear="none"><a rel="nofollow" shape="rect" target="_blank" href="http://lists.opennebula.org/listinfo.cgi/users-opennebula.org">http://lists.opennebula.org/listinfo.cgi/users-opennebula.org</a><div class="" id="yiv8225798884yqtfd15067"><br clear="none"></div><br clear="none"><br clear="none"></div> </div> </div> </div></div> </div></div></div></div></body></html>