Interesting analysis of linux kernel threading by IBM

[Malcolm Beattie]

Larry McVoy writes:
> : Or a machine with a lot of users.  (For example, a University unix server)

You rang?

> : If it wants to be the most efficient desktop machine, then it doesn't
> : need it NOW.  However, the average number of programs people are
> : running on their machine are increasing, not decreasing.  
> 
> This is a completely unsubstatiated statement.  OK, everyone, do this:
> 
> $ vmstat
> load free cach swap pgin  pgou  dk0 dk1 dk2 dk3 ipkt opkt  int  ctx  usr sys idl
> 0.00  4.3 96.9  0.4    0     0    0   0   0   0    0    0  104   53    0   3  97

As a concrete example of what you saying: on one of our two
interactive general-purpose servers we currently have:
[~]ermine% uptime
10:33  up 29 days, 15:54,  227 users,  load average: 1.61, 1.84, 1.84
% lusers
 176 telnetd
   9 rlogind
   8 xterm
 217 tcsh
   5 ftpd
 448 imapd-4.1.FINAL
  11 popper
   0 apache
 528 distinct users
 649 sessions
% vmstat
Virtual Memory Statistics: (pagesize = 8192)
  procs    memory         pages                          intr        cpu      
  r  w  u  act  free wire fault cow zero react pin pout  in  sy  cs  us  sy  id
  51307 29   97K 1424  28K 2629M 139M 169M 769K 167M 6522 237  6K 946   6   6  88

and on the other we have

% uptime
10:37  up 59 days,  2:08,  286 users,  load average: 8.96, 7.34, 7.08
% lusers
 263 telnetd
   7 rlogind
   7 xterm
 285 tcsh
   3 ftpd
  88 imapd-4.1.FINAL
  16 popper
   0 apache
 253 pine
 354 distinct users
 377 sessions

They're both old (oh, so old) Alphaservers with 1GB RAM: the first
has 4 x 250MHz CPUs; the second has 2 x 300MHz CPUs (both running
Digital UNIX as it happens, as you can see from the low uptimes
compared to our Linux boxes :-).

> Those people with the load field higher than 10, please tell us what you are
> doing.

The second machine above does actually tend go over a load average of
10 at peak times (to 20 or 30...) but that's when it's even more
riduculously overloaded (~500 concurrent users our of a user
population of 12000). As Larry says, even with many hundreds of people
hammering on an old machine, the load average stays low when the system
is behaving. Now it *is* possible to make the load average go sky high
by running silly or buggy programs but we tend to catch and LART the
users that run them fairly quickly.

--Malcolm

-- 
Malcolm Beattie <mbeattie en sable.ox.ac.uk>
Unix Systems Programmer
Oxford University Computing Services

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo en vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/