Steve_Lammert@TRANSARC.COM (12/05/90)
We have several DECstation 2100s with multiple SCSI disks.
These are server machines, typically with a small boot/usr/swap disk
and several larger disks for file storage. An example would be
one Connor 3100 disk (internal) and three WREN VIIs (external).
I'd like to add a fourth WREN VII to these systems, but worry
about the ability of the 2100 to handle this configuration.
Discounting cable length limits (which I know about)
and configuring the kernel properly to handle the disk (also known),
can anyone forsee any problems with adding the fourth external disk
(for a total of five reasonably fast SCSI disks)?
Is anyone successfully pushing a 2100 as a file server with
a similar configuration?
Thanks.
----------------------------------------------------------------------------
Steve Lammert | Gulf Tower | Internet: shl@transarc.com
Facilities Coordinator | 707 Grant Street | Voice: +1 412 338-4443
Transarc Corporation | Pittsburgh PA 15219 | FAX: +1 412 338-4404
----------------------------------------------------------------------------
"No matter where you go, there you are." -- Buckaroo Banzai
----------------------------------------------------------------------------davecb@yunexus.YorkU.CA (David Collier-Brown) (12/05/90)
Steve_Lammert@TRANSARC.COM writes: >I'd like to add a fourth WREN VII to these systems, but worry >about the ability of the 2100 to handle this configuration. I'm mildly doubtfull **if the server is functioning as a workstation at the same time**. Using somewhat slower RZ56's we found a combined station/fileserver was not giving good performance with three disks, bud did give good performance with a pair of Wrens. I didn't measure anything, so I don't know for sure what ``knee'' we hit, but I mildly suspect total i/o's per second, based on some measurements on non-workstation servers with three RZ56s >Is anyone successfully pushing a 2100 as a file server with >a similar configuration? Yes and no: our other configuration is one 2100 to 11 diskless clients, and the performance problems there are strictly bottlenecks in swap i/o (not that I would recommend diskless 2100s to my worst enemy, you understand). There are no visible bottlenecks on the non-swap disks. --dave -- David Collier-Brown, | davecb@Nexus.YorkU.CA, ...!yunexus!davecb or 72 Abitibi Ave., | {toronto area...}lethe!dave or just Willowdale, Ontario, | postmaster@{nexus.}yorku.ca CANADA. 416-223-8968 | work phone (416) 736-5257 x 22075
jg@crl.dec.com (Jim Gettys) (12/05/90)
Should work just fine. And yes, cable length is (as always with SCSI) the major issue. All cable counts, including connectors and internal cables inside of boxes. If you keep within the spec, you should do ok. I've certainly seen DS3100's with five disks (and it is identical to the DS2100 except for clock rate, cheaper processor/FPU, and cheaper cache ram). - Jim
pavlov@canisius.UUCP (Greg Pavlov) (12/06/90)
In article <obKybH=0BwwK8Thd4s@transarc.com>, Steve_Lammert@TRANSARC.COM writes: > > We have several DECstation 2100s with multiple SCSI disks. > These are server machines, typically with a small boot/usr/swap disk... > > I'd like to add a fourth WREN VII to these systems, but worry > about the ability of the 2100 to handle this configuration. > The 2100 will "handle" the additional disk, but unless you are severely disk seek-limited, you are unlikely to get additional i/o output. E.g., if the additional disk is to accomodate additional users and/or applica- tions, I think that you will see degredation in performance per user. greg pavlov pavlov@stewart.fstrf.org
pavlov@canisius.UUCP (Greg Pavlov) (12/06/90)
In article <1990Dec4.213701.10079@crl.dec.com>, jg@crl.dec.com (Jim Gettys) writes: > > I've certainly seen DS3100's with five disks (and it is identical to the > DS2100 except for clock rate, cheaper processor/FPU, and cheaper cache ram). > - Jim ....and the same rip-roarin' SCSI interface ? (has DEC ever considered offering an upgrade on those ? Would make the machines somewhat more useful. And the DS5000 shows that DEC can provide a reasonable SCSI when it wants to) greg pavlov, fstrf, amherst, ny pavlov@stewart.fstrf.org