On 27 March 2015 at 05:06, Matthew Law <[email protected]> wrote: > Obvioulsy, I would expect a VM to be slower than bare metal but perhaps not > this slow? The load on the host when the build is running peaks at around > 0.8. Shutting down the other few tiny VMs on the host has little effect. > How do I go about debugging and optimising the performance?
You should probably start by trying to figure out if the bottleneck is CPU, or disk I/O, or some other factor. Because you're using hardware virtualisation, it likely makes sense to look (first) at whatever performance monitoring you can do within the guest OS. Then, try and compare that to what you see with similar tools on the host. For example, if a tool like top or htop in the Linux guest shows you that you are always using 100% of every virtual CPU, then you are probably CPU bound. It makes sense to then look at, say, "prstat -mL -p <qemu_pid> 1" in the host and see what the threads are doing there: are they 100% split across USR and SYS work? Are they mostly blocked up waiting for things (SLP/LAT)? If you are _not_ CPU bound, you should move on to other resources. It may help to follow a methodology like the USE Method[1], looking at both the guest[2] and the host[3] OS. [1] http://www.brendangregg.com/usemethod.html [2] http://www.brendangregg.com/USEmethod/use-linux.html [3] http://www.brendangregg.com/USEmethod/use-smartos.html -- Joshua M. Clulow UNIX Admin/Developer http://blog.sysmgr.org ------------------------------------------- smartos-discuss Archives: https://www.listbox.com/member/archive/184463/=now RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00 Modify Your Subscription: https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb Powered by Listbox: http://www.listbox.com
