NetNews Usenet Archive 1992 #19/NN_1992_19.iso/spool/vmsnet/internal/1210

Internet Message Format | 1992-08-29 | 2.7 KB

Path: sparky!uunet!haven.umd.edu!darwin.sura.net!jvnc.net!rutgers!network.ucsd.edu!mvb.saic.com!macro32
From: [e-mail]
Newsgroups: vmsnet.internals
Subject: re: RE: VMS tuning for a VAXcluster (LAVC)
Message-ID: <20600630_00057EE8.0095FC77E2287DE0$[e-mail]>
Date: 28 Aug 92 15:32:18 GMT
Organization: Macro32<==>Vmsnet.Internals Gateway
Lines: 46
X-Gateway-Source-Info: Mailing List

Harvey Brydon writes:

...

> My claim is that paging through [an unloaded] ethernet is faster than to a
> local disk.  Of course if a lot of systems page this way, you will have a
> network contention problem which will slow things down.  [By the way, this
> assumes paging to another VMS system.  Paging to an Infoserver, as per the
> VXT2000 workstation is essentially the speed of the disk on the infoserver,
> plus you are generating ethernet traffic.  There is no caching on an
> Infoserver.]
>
> Any disagreements with the above?  Anybody do any quantitative testing on
> this?

An interesting viewpoint, and this may  be true if you have low-end
satellites on a high-end boot server with state-of-the-art disks (or
disk arrays) on an HSC. It certainly is NOT true if your boot node is simply
the most powerful VAXstation in a cluster of ten or twenty others, all with
SCSI disks.

The real weakness in the argumemt, I think, is the idea of an unloaded Ethernet.
10 Mbps was close to infinite when Ethernet was invented, but at most sites
today backbone traffic is saturating at peak times. Using a local disk for
paging (and for copying frequently used images to) reduces load on the
Ethernet, which is good all round. We really need something faster than
Ethernet: I can only hope that the prices of FDDI kit will fall rapidly in
the next few years (and that copper alternatives to fibre will be offered).

What would be really nice would be a cheapish box that took data to-and-from
a boot node at high speed (like FDDI) but in some simple-to-engineer way
(like bit-parallel words down a ribbon cable - maybe use standard SCSI bus?)
This box would in turn have (say) eight ordinary Ethernet ports emerging
from it. Internally it would function like a very fast multiport bridge.
The driver for this box would have an interface just like an Ethernet port
driver, but the data throughput would be a lot higher. I'm a physicist
not an electronics wiz, but I would have thought that a box like
this could be made a lot cheaper than an FDDI repeater and interface, and
by distributing satellites on the ethernet strings connected to it one
could increase the bandwidth between them and the server considerably.
Any entrepreneurs out there?

        Nigel Arnot

                [e-mail]   (internet)
                [e-mail]        (bitnet)