All about the system administration and application development behind a local linux-based company
I just finished a presentation at the Niagara Frontier LUG on Xen virtualization and how it applies to a managed hosting infrastructure. Here is a copy of the presentation for anyone who’s interested (sorry, it’s MS powerpoint. Yes, I do appreciate the irony).
It goes over some of the pros and cons of all virtualization, different types of virtualization, strategies for achieving levels of availability, and finally, some xen-specific configuration options and tips.
Edit: Mark asked me to clarify the benchmark parameters, so here’s an excpert of an email that I sent to another reader about this.
as an FYI, the benchmarks were done with:
core 2 duo 1.8Ghz
single 7200RPM 2.5″ driveThe image was done as a .img file on the same partition as the DomU, and the
partition was a separate DOS partition on the same drive. The only VMWare test
I ran used an image file. All of the OS installs were cloned from the same directory tree
(stay tuned for a future post on converting a Xen image to vmware)The Moodle benchmarks were a snapshot of a production moodle install,
where I copied the install to my test system, logged in and clicked a
few things, and replayed the session logs through jmeter several hundred
times with five concurrent threads.The images benchmark is transfer speed of randomly selecting 1 of 2500
10kb-40kb images. 8 concurrent users and 1500 iterations.The Mysql benchmark is the results from running all the tests in the
mysql benchmark suite.Of note, the image file generally performed a little bit better than the
raw partition. This is counter to what the Xen documentation and common
sense would say, and I think a lot of it has to do with my pretty
limited tests. The image file was ending up in memory cache, whereas
the block device wasn’t. I doubt the same comparitive performance would
play out in a production system where a lot more’s going on.