Aucbvax.3165
fa.unix-wizards
utzoo!decvax!ucbvax!unix-wizards
Tue Sep 15 19:27:54 1981
Re:  11/70 Loads
>From mike.bmd70@BRL Tue Sep 15 19:21:50 1981
Lt. Breitling -
       We are running an enhanced V6 UNIX on 2 11/70s here, and have
some of the following performance observations:

1)  The NCP code does not seem to use much of the CPU, cycle wise.  The
   user programs are not really much worse.  You do pay a penalty in
   core usage, because the network software does occupy a fair amount
   of core...  3 TELNETs, the NCP, ServerTelnetDaemon, FTPServerDaemon,
   and an active FTP or two takes almost 600 Kb.  (This observation
   made with memory lights & memory utilization analyzer).  This does not
   grow linearly with load, however, thanks to the "pure text" mechanism.

2)  Our primary 11/70 delivers sub-second (0.1 -> 0.2 s) response time for
   most requests (editor commands, WHO, MOUNT, etc) which do not demand
   significant I/O from the moving head disks.  We can deliver this with
   30 users on the system (1/3 doing program development, 2/3 editing and
   TROFF).  However, we have a rather special hardware configuration which
   aids this:
               ROOT            4096 blocks of BULK CORE (RF-11 interface)
               SWAP            4872 blocks of RK-11
               /tmp            4096 blocks of BULK MOS
                *              3 CDC 80 Mb drives, Massbus interface
             -MEMORY-          1.5 Mb

3)  "Number cruncher" processes tend to get little or no CPU time durring
   prime time;  they just sit around and wait for spare cycles.  This
   is due to our "High Performance Scheduler".  Gives great interactive
   response, and prevents "batch" work from interfering with interactive
   terminals.

4)  Although 1.5 Mb of main memory does seem to be sufficient now, when
   we had only 0.5 Mb, we did observe swapping rates of up to 10/sec
   before performance began to suffer too much.  Swapping is to be avoided.

5)  With the addition of another MASSBUS disk interface and more drives,
   we could probably support 48-55 users with the same performance,
   although the CPU would probably start to run out of steam at this point.

6)  I could go on and on about performance issues.  If there is interest,
   we could put together a somewhat longer letter....  What interests
   people?

7)  We plan to put UNIX on the HEP (Heterogeneous Element Processor),
   a multi-CPU machine, where the raw speed of each CPU element is 10 MIPS
   and only costs $250K - $500K (not sure).  Our initial configuration
   will be 40 MIPS.  (From the people who brought you ENIAC, ...HEP!)
   We think it will make a superb interactive facility.

                               Cheers,
                                -Mike

-----------------------------------------------------------------
gopher://quux.org/ conversion by John Goerzen <[email protected]>
of http://communication.ucsd.edu/A-News/


This Usenet Oldnews Archive
article may be copied and distributed freely, provided:

1. There is no money collected for the text(s) of the articles.

2. The following notice remains appended to each copy:

The Usenet Oldnews Archive: Compilation Copyright (C) 1981, 1996
Bruce Jones, Henry Spencer, David Wiseman.