The arrow of time

Ivan Voras' blog

Some simple Xen performance measurements

When it doesn't do IO, virtualization is pretty good performance-wise these days. Here's how Xen looks like  on FreeBSD 8-stable AMD64 XENHVM (pysysbench).

Without Xen (no vrtualization):

Running 8 benchmarks on 2 threads for 10 seconds each
Running benchmark: Warmup-Ignore ... score: 230.2 average: 115.1
Running benchmark: Hash-SHA256 ... score: 231.2 average: 115.6
Running benchmark: Hash-SHA512 ... score: 364.4 average: 182.2
Running benchmark: Zlib-Compress ... score: 157.0 average: 78.5
Running benchmark: Zlib-DeCompress ... score: 315.5 average: 157.7
Running benchmark: Socket-Syscalls ... score: 343.7 average: 171.9
Running benchmark: Socket-OneSocket ... score: 96.3 average: 48.2
Running benchmark: Socket-OnePipe ... score: 151.8 average: 75.9
Total score: 1890.1

With Xen:

Running 8 benchmarks on 2 threads for 10 seconds each
Running benchmark: Warmup-Ignore ... score: 228.2 average: 114.1
Running benchmark: Hash-SHA256 ... score: 228.9 average: 114.5
Running benchmark: Hash-SHA512 ... score: 361.8 average: 180.9
Running benchmark: Zlib-Compress ... score: 132.8 average: 66.4
Running benchmark: Zlib-DeCompress ... score: 312.5 average: 156.2
Running benchmark: Socket-Syscalls ... score: 109.7 average: 54.9
Running benchmark: Socket-OneSocket ... score: 38.9 average: 19.5
Running benchmark: Socket-OnePipe ... score: 153.6 average: 76.8
Total score: 1566.5

Curiously, pushing a single byte through a pipe takes almost no performance hit in virtualization but pushing it through a socket becomes very expensive (I'm looking at the last two benchmarks).

 

Post your comment here!

Your name:
Comment title:
Text:
Type "xxx" here:

Comments are subject to moderation and will be deleted if deemed inappropriate. All content is © Ivan Voras. Comments are owned by their authors... who agree to basically surrender all rights by publishing them here :)