Re: Scalability issues with Kerberos

Thread View: comp.protocols.kerberos

2 messages

[T] Back to Threads [V] Tree View [G] Group Articles

2 total messages Started by Marc Horowitz Sat, 23 Jan 1993 22:42

#939

Author: Marc Horowitz
Date: Sat, 23 Jan 1993 22:42

13 lines
1082 bytes

>> Although the load isn't constant over 24 hours, it doesn't vary as much
>> as you might think:

I see a peak when 9-5'ers get in in the morning, and another peak when
students get out of class.  And the fact that students have no
discernible daily cycle accounts of the 1000+ requests at every hour
of the day.  I doubt intel would have such a distribution.  Most
people would kinit once when they got there and get most of the keys
they need first thing in the morning, leading to a large spike at the
time work starts, and not much elsewhere (except engineers, who also
don't have any discernible daily cycle :-).

		Marc

Re: Scalability issues with Kerberos

#940

Author: John Hascall
Date: Sun, 24 Jan 1993 03:41

49 lines
2560 bytes

I wrote:
> Although the load isn't constant over 24 hours, it doesn't vary as much
> as you might think:

Marc Horowitz <marc@MIT.EDU> writes:
}I see a peak when 9-5'ers get in in the morning, and another peak when
}students get out of class.  And the fact that students have no
}discernible daily cycle accounts of the 1000+ requests at every hour
}of the day.  I doubt intel would have such a distribution.

Well, I did a little more work -- I added code in our Kerberos server to
measure the time from request-arrival to reply-departure.  I'll let
it run Monday for some real data, but here's what the first 6 minutes
of running it this evening gave (yes, another gruesome ascii chart):

     90+             *               Time from Request to Reply
       |             *                 (in 256ths of a second)
     80+             *
       |             *
     70+             *
       |             *
 C   60+             *
 o     |             *
 u   50+             *
 n     |             *
 t   40+             *
       |           * *
     30+           * *
       |           * *
     20+           * * *
       |           * * *
     10+           * * *
       |           * * *           * * *   * * *
      0+-*-*-*-*-*-+-+-+-*-*-*-*-*-+-+-+-*-+-+-+-*-*-*-*-*-*-*-*-*-*-*-*-*-
         0 1 2 3 4 5 6 7 8 9 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 3 3 3
                             0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2

As you can see, it generally takes our server (an 8MB DS2100 [~8 Specmarks]
with a not-very-fast RZ55 disk) about 6/256 (0.023) sec to service a request.
One might hypothesize that using a modern workstation as a Kserver you should
be able to handle more than 100 requests per second.  I suppose if you needed
more oomph, you could rewrite the server to be multi-threaded (*shudder*).

John
-- 
John Hascall                   ``An ill-chosen word is the fool's messenger.''
Systems Software Engineer
Project Vincent
Iowa State University Computation Center  +  Ames, IA  50011  +  515/294-9551

Thread Navigation

This is a paginated view of messages in the thread with full content displayed inline.

Messages are displayed in chronological order, with the original post highlighted in green.

Use pagination controls to navigate through all messages in large threads.

Back to All Threads