On 28/04/15 16:51, Antti Kantee wrote:
Can we get a re-run of the benchmarks? I get a >30% performance boost for the first 20s siege run and don't observe slowdown over multiple long 200s runs. However, since I'm running on baremetal/qemu instead of Xen, the numbers aren't at least directly comparable.
I tested it with the guest running as qemu -cpu host -enable-kvm on this prehistoric laptop (core2duo).
With nginx running on a Linux guest, I get ~4k t/s. With nginx on rumprun I get ~3.9k t/s (~3.95k if I do the build with -V RUMP_DIAGNOSTIC=no). IOW, Linux performance is 1-2% better in this setup. I think that's decent considering we didn't even think of performance until last week.
The interesting thing to note is that the longest request time for Linux is 0.02s, while for rumprun it's 0.07s. So rumprun is hitting some pathological case somewhere. Just figuring out what that is and fixing it should help us a lot.
nb. the Linux guest is running on qemu-system-x86_64, while rumprun is running on qemu-system-i386. I have no idea which way that tilts the numbers, if any.
I'm still interested in the Xen numbers.
