LinuxSA Mailing list archives

Index: [thread] [date] [subject] [author] [stats]
  From: Alan Kennington <akenning@dog.topology.org>
  To  : Glen Turner <glen.turner@aarnet.edu.au>
  Date: Wed, 31 Jan 2001 13:00:56 +1030

Re: uh oh! -- BIND's no good again - warning, warning

On Wed, Jan 31, 2001 at 12:32:41PM +1030, Glen Turner wrote:
> Mark Newton wrote:
> > 
> >  > Large is no great problem in general.
> >  > But if the old named runs in one process of 2 MBytes at start-up,
> >  > and the new one runs in 5 processes of 2 MBytes, then I don't count
> >  > this as an improvement. If the software had been written by MS,
> >  > people would call this bloatware.
> >  > But since linux is without sin....
> > 
> > Are you sure you're counting correctly?
> > 
> > named v9 is multithreaded.  Are you sure you're not seeing 5 threads
> > sharing the same 2 Mbytes?  I don't think the situation is anywhere
> > near as bad as you're making it sound (indeed, it's probably about 20%
> > as bad :-)
> 
> I have Mark Crispin, one of the ten BIND authors, looking
> over my shoulder as I write.  He says that the processes
> all share memory.  He also says that BIND9 is like BIND8
> in that the database is held in a red-black tree in memory.

Glen,

I'm very sceptical about this.
I saw my free swap space go down by the same amount as the
BIND memory went up.
I checked the three relevant columns - SIZE, RSS and SHARE in top.
The documentation was clear about the threading and lighweight process
aspects etc. So I was expecting to see less real memory usage.

However, I suspect that with my 2.0.36 kernel on RedHat 5.2,
the memory sharing was probably just not happening.
In fact, clearly it was not.
As soon as I wiped 9.1 and installed 8.2.3, the virtual memory
usage went down by 8 MBytes.
I.e. those 5 processes at 2 MBytes each were real processes, not just
threads, and the shared memory was rather small.
My principal reason for dumping 9.1 was the RAM usage.
I just simply don't know what type and amount of RAM to put into
my old Digital Venturis 466.
I'll go out and buy more RAM if someone can tell me what
kind I need.

Just for a little moment, I've just re-run the old 9.1 version.
"Top" gave me this:

---------------------------------------------------------------------
 12:52pm  up 18 days,  1:58, 14 users,  load average: 0.48, 0.14, 0.04
98 processes: 97 sleeping, 1 running, 0 zombie, 0 stopped
CPU states:  9.0% user, 10.6% system,  0.0% nice, 80.5% idle
Mem:   14652K av,  14192K used,    460K free,   6308K shrd,    256K buff
Swap:  33228K av,  23776K used,   9452K free                  3076K cached 

  PID USER     PRI  NI  SIZE  RSS SHARE STAT  LIB %CPU %MEM   TIME COMMAND  
// This is version 9.1.0.
26821 root       0   0  2596 2060  1060 S       0  0.0 14.0   0:00 named
26820 root       0   0  2596 2060  1060 S       0  0.0 14.0   0:00 named
26822 root       1   0  2596 2060  1060 S       0  2.2 14.0   0:02 named
26823 root       0   0  2596 2060  1060 S       0  0.0 14.0   0:00 named
26824 root       0   0  2596 2060  1060 S       0  0.0 14.0   0:00 named

// This is version 8.2.3.
25566 root       0   0  1452  712   204 S       0  0.0  4.8   0:34 named 
---------------------------------------------------------------------
[The 18-day up time is due to electricity failure.
The machine only crashes when the elctricity is stopped - or
sold interstate at a higher price.]

The amount of memory in the SHARE column is much greater than I
remember seeing when it was really running.
Obviously there's a little port re-use clash when 9.1.0 runs on
top of 8.2.3:

---------------------------------------------------------------
Jan 31 12:52:32 dog ./named[26822]: listening on IPv4 interface lo, 127.0.0.1#53
Jan 31 12:52:32 dog ./named[26822]: binding TCP socket: address in use
Jan 31 12:52:32 dog ./named[26822]: listening on IPv4 interface eth0, 203.38.148.51#53
Jan 31 12:52:32 dog ./named[26822]: binding TCP socket: address in use
Jan 31 12:52:32 dog ./named[26822]: listening on IPv4 interface ppp0, 139.130.140.14#53
Jan 31 12:52:32 dog ./named[26822]: binding TCP socket: address in use          
---------------------------------------------------------------

Cheers,
Alan Kennington.

-- 
LinuxSA WWW: http://www.linuxsa.org.au/  IRC: #linuxsa on irc.linux.org.au
To unsubscribe from the LinuxSA list:
  mail linuxsa-request@linuxsa.org.au with "unsubscribe" as the subject


Index: [thread] [date] [subject] [author] [stats]
Return to the LinuxSA Mailing List Information Page