Re: [uml-user] UML scheduling

[Jim C. <ji...@ma...>]

On Thu, 24 Feb 2005, Gordon Russell wrote:

> We have a cluster of 5 machines, each running about 15 UMLs. Things
> seem to run great for a while, then performance of the UMLs seems to
> die for a while.
> --snip--
> Machines are all 1GB or better, 2GHz or better, on a 0.1GB network
> backbone. 

With 15 guests per host, could you be running yourself out of physical 
memory?  If one guest got memory-intensive it could set all of them 
thrashing, until the memory-intensive task finished, causing a major 
slowdown for everyone.  Use "top" or "ps" on the host to look at the 
resident set size of the guests and the memory and swap summaries, and 
vmstat on the host and the guests to see who was swapping how much in and 
out.

If you gave each guest skimpy memory and a generous swap file, only the 
memory-intensive guest would be slowed down, except when other tasks needed 
to use the disc.

I've had good results with openvpn, but wouldn't have thought of it for 
connecting from a student's workstation to a UML guest.

James F. Carter          Voice 310 825 2897    FAX 310 206 6673
UCLA-Mathnet;  6115 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA  90095-1555
Email: ji...@ma...    http://www.math.ucla.edu/~jimc (q.v. for PGP key)