aurora-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mark Chu-Carroll <mchucarr...@apache.org>
Subject Vagrant provider statistics
Date Tue, 04 Feb 2014 16:19:37 GMT
I've been experimenting with using the VMWare provider for Vagrant, and
I've gathered some stats on relative performance.

General comments:

   - VMware consistently performs *slightly* better, but provisioning is
   slower.
   - VMware appears to be more resilient in the face of configuration
   errors. For example, if you assign the same IP address to two virtual
   machines, VirtualBox will sometimes hang with a lost lock, and require the
   machine to be rebooted. VMware appears to have no trouble with this issue.
   - Overall, it's probably not worth the $200 it takes to purchase the
   necessary licenses to use VMware. The different just isn't big enough. Both
   work well.

Statistics: Using the current 2-slave tests. Machine is mostly idle.
 Memory use was indistinguishable between the two alternatives.  Timings
were gathered from the command-line, using "time", with real time reported.
 CPU percentage was gathered with "top -c a", run for five minutes in an
idle state, with the 6 VM processes averaged.

VMWare:

   - Vagrant up, with no prebuilt vms: 13m 54s
   - Vagrant halt: 0m30.557s
   - Vagrant up, with prebuilt vms: 2m34.104s
   - Vagrant reload: 3m10.957s
   - Test run: 2m57.714s
   - Machine load percentage idle: 1.0-2.9, ave 2.0


VirtualBox:

   - Vagrant up, with no prebuilt vms: 10m 22s
   - Vagrant halt: 0m28.255s
   - Vagrant up, with prebuilt vms: 2m44.747s
   - Vagrant reload: 3m12.084s
   - Test run: 3m9.324s
   - Machine load percentage idle: 1.9-3.7, ave 2.6

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message