Re: 2 GB myth

wizofoz2k_at_yahoo.com.au (Noons) wrote in message news:<73e20c6c.0411292238.6810e874_at_posting.google.com>...
>
> And an upper limit is very easy to hit directly or indirectly:
> everything in Linux or Unix is represented as a file unit
> including shared memory segments, network connections,
> real files, directories, etcetc. It doesn't take much for a
> server process to hit a few hundred open files.

[Lightbulb dimly flickering above head] Oh yeah, I had forgotten about running into this as O attempted to attach to shared memory on a new system. How might I find the process with the most network connections?

>
> So it may indeed be a good idea to restrict the number of files
> (make them bigger) for very large databases. 50Gb seems to be
> a size that most large capacity/high speed tape drives can cope
> with in a single volume so maybe that is the next "rule of thumb"
> boundary? It's certainly consistent with the practice I'm
> seeing of people running Tb-class dbs.

A Tb here, a Tb there, pretty soon you're talking about some _real_ data.

>
> Dunno about Windoze?

http://terraserver-usa.com/about.aspx?n=AboutTechStor2

jg

--
@home.com is bogus.
"If you had agoraphobia, you'd be home by now." - Bumper sticker