Nice and confused about nice

Hangdog42 · 06-02-2003, 03:12 PM

I've got seti@home running in the background and I start it with -nice 19 in the hopes that it doesn't get processor time when I'm doing more important things. However, I've noticed that when I am doing processor intensive work (like compiling or doing some big honking SQL joins) seti still gets about half of the CPU time (according to top).

So my questions are:
1) is this normal?
2) is there a way to really clamp down on seti when a higher priority process comes along?

Tinkster · 06-02-2003, 04:03 PM

I'm not interested in seti (I believe that
the best proof for alien intelligence is the
fact that they don't try to conatct us ;})
so I wouldn't know about seti's "niceness" ...
but in general nice is ranged from -20 to 19,
with 0 being the norm.

If you find that your queries run too slow
with seti active, re-nice your sq or the shell
that does the compilingl to -10, -15, -19 ...

Cheers,
Tink

Hangdog42 · 06-02-2003, 08:27 PM

Yeah, looking for aliens does seem a bit off. I may switch over to folding proteins. At least I KNOW that is doing some good.

Thanks for the suggestion. I'll give it a rip and see if that does it.

cuckoopint · 06-02-2003, 09:19 PM

1. top is not very accurate w/ cpu activity
2. nice does not chunk up your cpu to portions and divide evenly. That's why sometimes its not the right solution. Instead, it prioritizes cpu time (a bit different).
e.g. Whatever you're doing in the foreground may be waiting for a reply - SETI may not be waiting and hence still use those cycles.

I remember someone released a kernel hack that divides the cpu power in the way you describe it - ie. SETI gets 10% cpu, etc. Maybe try a google search.

Hangdog42 · 06-03-2003, 07:17 AM

Quote:

nice does not chunk up your cpu to portions and divide evenly

Yeah, I was starting to guess that from watching some SQL queries. What's bugging me is that the same query runs significantly faster if I kill seti first.

I'm going to mess with priorities a bit, but if that doesn't do it I'll probably just dump seti. I've finally got a nice kernel and I really don't want to mess with it. Thanks for the suggestion though.

0x4B · 06-03-2003, 04:44 PM

I don't suppose you have a dual processor? even though seti@home is nice, the compilation would likely only use one so seti would be free to use the other (thus getting about half of the CPU time)