On Sat, 16 Mar 2002 09:06:06 -0800 (PST)
Dianne Marie Montesa <[EMAIL PROTECTED]> wrote:

> hi list
> 
> it just so happen that i am currently researching
> about threads-max when i saw this thread on my mails.
> the internet sources say the computation for
> thread-max is 
> 
> ((RAM * 1024 * 1024) /8192)/2
> 
> check out:
> 
> http://www.linuxjournal.com/article.php?sid=3814
> 
> however, on actual experience, i have a web/java
> server with 2 GB RAM and the default thread-max set is
> only 32768. if i should use the computation above it
> would be way more than 32768. 

This holds with the one thing I did find on the net.  Apperently this is new to the 
2.4 kernel and has a Theoretical limit of 32768.  I would guess that this means it 
hasn't been tested above this limit.  And thanks loads for the link and info.

James

> 
> it seems that it is assumed (by the kernel developers)
> that 512 MB RAM is the max RAM which is why even if i
> have more than 512 MB, my thread-max was set to 32768.
> or maybe they think that 32768 is the max that is
> still workable for the linux OS. i really dont have
> any idea about that.
> 
> im not sure how its gonna affect the system by setting
> the max thread to the computed value or at least
> double the current default. anyway, i am gonna make
> some tests to see if its gonna make the java apps
> perform better or if its gonna kill it. 
> 
> hope the info above sheds light on your queries about
> the computation.
> 
> cheers,
> dianne
> 
> --- damian <[EMAIL PROTECTED]> wrote:
> > El vie, 15-03-2002 a las 16:52, James escribi_:
> > > On Fri, 15 Mar 2002 09:47:14 -0500
> > > Bill Davidson <[EMAIL PROTECTED]> wrote:
> > > 
> > > > Now I'm curious. How did you find that out?
> > > 
> > > I looked into /proc/sys/kernel/threads-max the
> > number was 3063 (I hadn't set anything here and was
> > curious so this is a stock number)  Then since my
> > ram was larger and the stock number was larger I
> > divided 3063 by 384 and came up with (rounded) 7.98 
> > when I did 1024 divided by 128 I got 8 so I assumed
> > (lousy idea but I did it anyway) that the ratio is
> > around 8 x ram = number in
> > /proc/sys/kernel/threads-max.
> > > 
> > > 
> > > James
> > > 
> > > > 
> > > > Bill
> > > > 
> > > > On Friday 15 March 2002 03:49 am, James wrote:
> > > > > Raffaele,
> > > > >   Your letter is one of those that peeks
> > curiosity.  I looked at my box
> > > > > (running 384 megs) and found that the number
> > was 3063  you said you had
> > > > > 128 megs and 1024 ... The ratio seems to be
> > about 8xram in megs ....
> > > > > don't know what this means or why it's set to
> > this but maybe this will
> > > > > help.
> > > > >
> > > > > James
> > > > >
> > > > >
> > > > > On Fri, 15 Mar 2002 08:52:14 +0100
> > > > >
> > > > > Raffaele Belardi <[EMAIL PROTECTED]>
> > wrote:
> > > > > > I had problems running a fairly big applet
> > in the Forte for Java 3.0
> > > > > > with JSDK 1.3.1: the IDE started ok, but
> > when I tried to execute the
> > > > > > applet I got an "out of memory" error (on a
> > 128Mbyte machine).
> > > > > >
> > > > > > I found out that for some reason the
> > threads-max kernel tunable was
> > > > > > set to 1024, which turns out to be too small
> > for the IDE. [is this a
> > > > > > default for MDK 8.1, or was it changed when
> > I played with the security
> > > > > > things?]
> > > > > >
> > > > > > Changing to 32K with
> > > > > > # echo 32768 > /proc/sys/kernel/threads-max
> > > > > > fixed the issue.
> > > > > >
> > > > > > bye,
> > > > > >
> > > > > > raffaele - italy
> > > > 
> > > > 
> > 
> > 
> > well i'm using 192 MB ram and my threads.max is
> > 3071...?
> > 
> > Damian
> > 
> > 
> > 
> > > Want to buy your Pack or Services from
> MandrakeSoft?
> > 
> > Go to http://www.mandrakestore.com
> > 
> 
> 
> __________________________________________________
> Do You Yahoo!?
> Yahoo! Sports - live college hoops coverage
> http://sports.yahoo.com/
> 
> 

Want to buy your Pack or Services from MandrakeSoft? 
Go to http://www.mandrakestore.com

Reply via email to