Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!wuarchive!udel!haven.umd.edu!socrates.umd.edu!mike
From: mike@socrates.umd.edu (mike santangelo (UNIX/VMS Sys Staff))
Newsgroups: comp.sys.hp
Subject: Re: Upgrade or Snakes ?
Message-ID: <1991May03.205924.7068@socrates.umd.edu>
Date: 3 May 91 20:59:24 GMT
References: <1295@theseas.ntua.gr>
Organization: University of Maryland, University College
Lines: 94

zvr@ntua.gr (Alexios Zavras) writes:

>Hi everyone!

Howdy.

>    I would like to get some opinions on the decision we are facing:

>    We have a HP9000 850S, which has a number of terminals (say,
>around 20), and is also on a net with a variety of otehr machines
>(mostly suns).  The idea is to upgrade the whole system, i.e. to
>give better performance to a number of users.
>    One way to go is to upgrade the 850S to 865S (I think I have
>the right numbers), which involves adding a processor, 16MB RAM
>and two 26/64MB controllers.  The problems is that the price HP
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  What do you mean by this?

>quotes for this is, to say the least, *very* high (prices in
>Greece are a lot different than the States, but we're talking
>$136,000 here: I suppose the equivalent in US should be about
 ^^^^^^^^^^^^^
  Sounds about right.

  Specs on 865S: 53 (integer) MIPS
                 11.4 (double-precision) MFLOPS
                 768K Cache (256K I + 512K D)
                 8K TLB (2KB pages)

  The 865S was the fastest HP-PA 1.0 processor HP had until
HP-PA 1.1 stuff came out (e.g. Snakes).  Model 870/100 was
same SPU but had capability to be upgraded in-cabinet to
two of these processors, 870/200.  Both 870 models had 256K
additional I cache.

>the half (?)).
>    As you can probably guess, we thought of better ways to invest
>that kind of money: get some snakes (the 9000/700 machines).

  Terminals...  You using HP asynch MUXs for this or TELNET terminal
servers?  You will need to use TELNET servers to do any serious
asynch user-interaction on the snakes due to the user-license
restriction (2 or 8 user), but if they come in over TELNET,
no problem (thank goodness), they all cound as 1 user, so...

>Possibilities include a 750 and a few 720, or just a bunch of 720s.
>    Now, keep in mind that the 850 already has 2GB disk, so it

  What kinda drives are these?  How are they connected, HPIB or HPFL?
Your likely to have a problem with either on the snakes, especially
HPFL, although there is an EISA HPIB controller.  Does anyone at HP
want to comment on whether the CS/80 and SS/80 stuff is going to be/is
supported driver-wise on HP-UX 8.X on the snakes?  Can you, using
this card, attach 793X disk drives and 7974/7978/7980 tape drives?

>would be a sort of file server to the new machines.  Would it
>be able to handle the load, or would it crash from network requests ?

  The Model 850S has a combined (I & D) 128KB cache, the Snakes have
at LEAST double this, TLB is 4K on 850S, ? on Snakes (can't find,
but page size in HP-PA has doubled from 2K to 4K and the TLB
has some interesting optimization that allow mapping multiple
contiguous virtual pages).  

>    So, another possibility is getting an I/O card for the 850S and
>some 700s, to help it cope with the traffic from the faster snakes.

  I/O card for the 850S?  Huh?  What kind of I/O card?

>    What do you all think ?

  It seems to be a hot topic on the list right now just how well
the Snakes do in a multiuser environment, apparently no one has had
the ability to find out yet. 
  What would be neat to see is a real benchmark along these lines,
the TPC-A would do quite nicely.  Anybody (esp. at HP) done any
preliminary studies along these lines with some popular database
like Informix or something?  Would help gauge multiuser performance.

>Any opinion would be much appreciated,
>-- zvr --
>    +---------------------------+	Alexios Zavras (-zvr-)
>    | H eytyxia den exei enoxes |	zvr@ntua.gr
>    +-----------------------zvr-+	zvr@theseas.ntua.gr
>Somebody said it couldn't be done  /  But he with a chuckle replied,
>Maybe it couldn't, but he would be one  /  Who never said no till he tried.
>So he buckled right in  /  With a brace and a grin
>                                       And if he was worried, he hid it,
>He started to sing  /  As he tackled the thing
>                                       That couldn't be done
>                                       And he did it.
-- 
                                  Mike Santangelo (mike@socrates.umd.edu)
                                  UNIX / VMS Systems Manager