[mod.computers.vax] clusters - update on system disk requirements

LEICHTER-JERRY@YALE.ARPA (01/22/87)

In a recent note asking about how much disk space was needed for the system
disk in a Local Area VAXCluster, I said:

    An RD52 is probably just large enough for your system disk, but I'd really
    recommend an RD53.  (Actually, I'm not even sure an RD52 is supported as
    the system disk on LAV boot node.)

I've been informed (by a correspondent who must remain anonymous) that the
situation from the point of view of SUPPORT is as follows:

    A Microvax II serving as a Local Area VAXcluster boot node must have an
    RD54, RA60, RA80, or RA81 as a system disk.  An RD53 system disk (while
    it may work) is not supported.  VAXstation II systems are not supported
    as boot nodes.

He then goes on to say:

    Of course, you and I both know that an RD53 will work.  :-)  As a matter
    of fact, I recently constructed a 4-node Local Area VAXcluster using an
    RD53 on a MicroVAX II boot node; each satellite node had a page and swap
    file on their local disk.  The system disk had lots of layered products
    installed, so it was pretty much maxed out.  Fortunately, we were able to
    switch boot nodes to one with two RA81 disks!

As always, understand what "an RD53 will work" means before deciding to try
that route:  If it works, fine; if it doesn't, or if it works now but breaks
tomorrow, just when you need it; or if it "works" but with unacceptable
performance - you are on your own.  "You break it, you bought it."  (If it
were up to me, I'd buy an RD54 - they are not THAT expensive, compared to your
total system cost, and you can very quickly chew up the difference in the time
you spend trying to get a non-supported configuration to work.  And then you'd
find yourself short of disk space....  Of course, an RA disk would be even
nicer, but then we're talking REAL money.)

BTW, VAXStation II's as boot nodes would almost certainly fall into the "un-
acceptable performance" category in almost any configuration - driving that
screen COSTS.
							-- Jerry
-------

sasaki@HARVARD.HARVARD.EDU (Marty Sasaki) (01/23/87)

If you are interested in performance with uVAXen, especially in
clusters, I would suggest using another manufacturers controller and
disk.

Our uVAX cluster's boot disk is a Fujitsu Super Eagle using Emulex's
QD-32 (?) controller. Throughput to the disk is much faster than to
either RA or RD disks. The CPU overhead on the machine acting as the
disk server is higher, but in our application, hardly noticeable.

The QD-32 is a much better controller, it is cheaper, it is smaller,
it is faster, and it uses less power than the KDA-50. Dialog and
System's Industries make controllers with similar performance and
cost, and I am sure that there are others.

The only real disadvantages is that these disks aren't MSCP, and you
are tied to the manufacturer for driver software.
----------------
  Marty Sasaki				uucp:   harvard!sasaki
  Ziff Davis Technical Information Co.	arpa:	sasaki@harvard.harvard.edu
  80 Blanchard Road			bitnet: sasaki@harvunxh
  Burlington, MA 01803			phone:	617-273-5500

cetron%utah-ced@UTAH-CS.ARPA (Ed Cetron) (01/23/87)

also note that at least two 3rd party manufacturers (including SI)
build eagle/super eagle's which are ra series compatible (they
plug right in to the uda-50, kda-50, kdb-50...)

also: of about 8 super eagles here at the U of U, most, if not all have
come DOA or died with 1-3 months....eagles have been perfect.... Fujitsu has
been extending warranties here up to 3 years, hoping to have better luck by
repeated replacements until you get the 'odd' good one.

-ed cetron