Saturday, February 03, 2007

Public servant

Cluster Report for Sat, 3 Feb 2007 10:31:59 +0800

CPUs Total:19
Hosts up:18
Hosts down:0

Avg Load (15, 5, 1m):
362%, 362%, 363%
2007-02-03 11:02

(This is not the flag of Japan.)

Here is a preview of CSRC's ganglia cluster load monitor
The problem with public servers... Anyone just gets to do anything no matter how reckless their programming habits are. With 512MB of memory and 10GB of swap, on each node just imagine how much greedy a program/(group of programs) could get to use more than 300% of these resources.

If someone else (aside from myself) will complain that he/she cannot run his/her program, then I will have to kill...

It's ironic that some people are on a high with their problem solving skills but could not interpret their solutions into feasible/efficient programs/code.

Scientific computing environments such as Matlab, Scilab, Mathematica, Maple etc... are good since they shorten the transition from solution to code to results visualization (especially the lazy man's symbolic Mathematica that does most of the thinking for you). However, their ease does not promote efficient programming. I even know someone who worries about his/her very long complicated equation. As if he/she has to evaluate it's value in a single blow. Most formulas involving two or more terms can be broken down into parts.

I guess some will never learn.

No comments: