tony@brueer.UUCP (02/13/86)
We are planning to buy a Sun workstation configuration consisting of 12 seats. The options we are considering are: (1) 12 disk-less Sun 3/50 workstations, each with 4Mb of memory and the 68881 chip, and a 3/180S fileserver with 4Mb of memory and 2 380Mb Fujitsu Eagle drives on a single controller. (2) 12 Sun 3/50 workstations, each with 4Mb of memory and the 68881 chip, 6 of which are disk-less and 6 of which have SCSI-connected 71 Mb disks. Each pair of 3/50s would share a 71 Mb disk for "disk serving". Also a 3/180S fileserver with 4Mb of memory but only 1 380Mb Fujitsu Eagle. The second configuration has been put up because of a general "lack of faith" in the disk-less approach, especially when going beyond 12 seats. This is NOT based on experience however and the second configuration has cost penalties. Has anybody experience of large disk-less configurations? (needn't be of 3 series Suns - 2 series would do) Has anybody done comparisons between disk-less and SCSI-disked nodes? Any comments? I don't mind if Sun U.S. answers! Apologies if this has been a recent topic of discussion. Please e-mail any comments to: tony%ee.brunel.ac.uk@ucl-cs.arpa (or whatever) Thanks Tony Begg
bzs@bu-cs.UUCP (Barry Shein) (02/17/86)
Ok folks, I am not ready to make general recommendations yet but I do believe that if you are planning on running SUN/3s as file servers for NFS you want more than 4MB of memory, probably 8MB will do it, especially if people will be logged in and working directly on the server (say, thru dumb terminals.) Other than that my experiences have all been quite positive (actually I'm tickled pink!) Given the price of the added memory compared with the price of a server plus volume discounts where applicable I see no reason to hesitate. Probably less than 10% additional in price (and only that much because the rest of the system is so inexpensive.) Sun has also been reducing their memory prices I believe, ask your salesthing. -Barry Shein, Boston University
pavlov@hscfvax.UUCP (840033@G.Pavlov) (02/17/86)
Likewise, I would appreciate information/thoughts/etc on using a 3/160 or 3/180 "server" as a general-purpose time-sharing machine (eventually part of a network of other machines, including "traditional" Sun workstation configurations). Comparisons to VAXen or multi-user HP9000 (series 500) would have the most meaning for me. Also: anyone have expereince networking Suns and HP9000's ?? thank you kindly, greg pavlov, fstrf, amherst, ny harvard!talcott!wjh12!pavlov
pavlov@hscfvax.UUCP (02/17/86)
sorry, that's harvard!talcott!wjh12!hscfvax!pavlov ^^^^^^^
larry@geowhiz.UUCP (Larry McVoy) (02/18/86)
In article <241@brueer.brueer.UUCP> tony@brueer.UUCP (Tony Begg) writes: >We are planning to buy a Sun workstation configuration consisting of 12 >seats. The options we are considering are: > >Has anybody experience of large disk-less configurations? (needn't be of 3 >series Suns - 2 series would do) Has anybody done comparisons between disk-less >and SCSI-disked nodes? Any comments? I don't mind if Sun U.S. answers! > >Apologies if this has been a recent topic of discussion. Nope - it hasn't been a subject of recent discussion and I think that it is of general enough interest to be a topic of discussion on the net. I know I'd like to hear about it... -- Larry McVoy ----------- Arpa: mcvoy@rsch.wisc.edu Uucp: {seismo, ihnp4}!uwvax!geowhiz!geophiz!larry "Just remember, wherever you go -- there you are." -Buckaroo Banzai
mark@umcp-cs.UUCP (Mark Weiser) (02/19/86)
In article <176@hscfvax.UUCP> pavlov@hscfvax.UUCP writes: > > Likewise, I would appreciate information/thoughts/etc on using a 3/160 or > 3/180 "server" as a general-purpose time-sharing machine (eventually part > of a network of other machines, including "traditional" Sun workstation > configurations). Comparisons to VAXen or multi-user HP9000 (series 500) > would have the most meaning for me. > One problem with Suns as general time-sharing machines is relatively slow terminal i/o. I believe I read a Sun spec which only promised 4800 baud continuous output (9600 baud input). I can confirm that when my Sun is handling 4800 baud the number of interrupts shoots through the roof. Perhaps there are smarter boards on the way, but where our 780 supports 80 terminals, I wouldn't want to be on a sun-3 with 8, even though its cpu is theoretically faster. One the other hand, we routinely use Suns here with 1 or 2 terminals in addition to the hi-res console. This is how faculty (who mostly have the suns) get extra terminals to their students (who mostly don't). -mark -- Spoken: Mark Weiser ARPA: mark@maryland Phone: +1-301-454-7817 CSNet: mark@umcp-cs UUCP: {seismo,allegra}!umcp-cs!mark USPS: Computer Science Dept., University of Maryland, College Park, MD 20742
kwan@smeagol.UUCP (Richard Kwan) (02/24/86)
In article <<3266@umcp-cs.UUCP> Mark Weiser writes: > In article <176@hscfvax.UUCP> pavlov@hscfvax.UUCP writes: > > > > Likewise, I would appreciate information/thoughts/etc on using a 3/160 or > > 3/180 "server" as a general-purpose time-sharing machine (eventually part ... > One problem with Suns as general time-sharing machines is relatively > slow terminal i/o. I believe I read a Sun spec which only promised > 4800 baud continuous output (9600 baud input). I can confirm that > when my Sun is handling 4800 baud the number of interrupts shoots > through the roof. Perhaps there are smarter boards on the way, > but where our 780 supports 80 terminals, I wouldn't want to be on > a sun-3 with 8, even though its cpu is theoretically faster. > > One the other hand, we routinely use Suns here with 1 or 2 terminals > in addition to the hi-res console. ... This sounds like use of the serial I/O ports on the CPU or SCSI board (using zs.o). Anyone using the Systech terminal MUX on a Sun-3/180? For that matter, how does the Systech board on a Sun-2 compare to a DZ (or other MUX board) on a 780? -- Rick Kwan JPL Spacecraft Data Systems sdcrdcf!smeagol!kwan (UUCP) ia-sun2!smeagol!kwan@csvax.caltech.EDU (ARPA) -------------------------------------------------------------------- "...jumpin' into hyperspace ain't like dustin' crops, boy." H. Solo --------------------------------------------------------------------
bzs@bu-cs.UUCP (Barry Shein) (02/27/86)
At Boston University we are using two SUN/3s with a 16 port Systech mux on each. We typically have 12-15 users on each system (almost all on the terminal ports.) I have seen no performance problems, the machine almost always appears lightly loaded at worst. I think those of you familiar with vaxen can think of the systech as being similar in this regard to a dmf32 or able (or, dh) mux. -Barry Shein, Boston University