i1neal@exnet.iastate.edu (Neal Rauhauser -- ELT Computer Applications Group) (12/18/90)
I have $10k budgeted to build a system to handle 10 - 15 concurrent
users. There will be 8 - 16 serial ports and a 1.5 MHz T1 line. Users
will be running a news reader and downloading information, not much else
will be occuring on the system. I'm not the least bit scared of PC
hardware but my only unix admin experience has been on my 'mighty
but erratic' coherent/286 system and a LMI Lambda - toy unix in both
cases. I need suggestions on the following:
1. processor speed, 386 or 486, cache sizes?
2. system ram, 12 meg? 16 meg? more?
3. disk controller? Adaptec 1740?
4. 16 port serial card?
5. the best worm or MO disk with worm capabilities - necessary for
legal/archive purposes
The most importatn question in my mind is what processor/speed
and how much memory I should use. Mail replies and I'll post a
summary.
Neal i1neal@exnet.iastate.edudavidsen@sixhub.UUCP (Wm E. Davidsen Jr) (12/22/90)
In article <1990Dec18.002013.28881@news.iastate.edu> i1neal@exnet.iastate.edu (Neal Rauhauser -- ELT Computer Applications Group) writes: | | I have $10k budgeted to build a system to handle 10 - 15 concurrent | users. There will be 8 - 16 serial ports and a 1.5 MHz T1 line. Users | will be running a news reader and downloading information, not much else | will be occuring on the system. I waited a few days for other replies on this and didn't see any. The T1 is the kicker. If it is transformed by hardware into ethernet then you have lots of choices. If you want to use some fancy interface you must decide based on what o/s supports it. I would go with Xenix. It's not remotely state of the art, but it is well tested and very reliable. I'm painfully aware that more modern systems, and most versions of V.3.2, are not quite as reliable. It supports multiport serial very well, it supports TCP if you don't need NFS, and is generally a good solution to your particular problem. | 1. processor speed, 386 or 486, cache sizes? With smart serial cards a 25MHz 386 will be fine. | 2. system ram, 12 meg? 16 meg? more? Go with 16, memory is down around $42/MB, choose a system which put it all on the motherboard. Regardless of ads, many versions of 386 unix become strange after 16MB. You don't need that much, don't blaze a trail for others. | 3. disk controller? Adaptec 1740? Garder variety ESDI. Use two drives to improve performance. I will venture to recommend the CompuAdd caching controller and Seagate (Wren) 766MB (670MB formatted) controller and drives. These have worked well in about a dozen systems at work, with varying loads, and I'm confortable recommending them. | 4. 16 port serial card? Two 8's. You might find performance a hair better one way or the other, but redundancy says use two, and there's no drawback I can see. *Smart* cards, please. | 5. the best worm or MO disk with worm capabilities - necessary for | legal/archive purposes Spend the money on a good tape drive and do the backup faithfully. -- bill davidsen - davidsen@sixhub.uucp (uunet!crdgw1!sixhub!davidsen) sysop *IX BBS and Public Access UNIX moderator of comp.binaries.ibm.pc and 80386 mailing list "Stupidity, like virtue, is its own reward" -me
cpcahil@virtech.uucp (Conor P. Cahill) (12/22/90)
In article <2710@sixhub.UUCP> davidsen@sixhub.UUCP (bill davidsen) writes: > I would go with Xenix. It's not remotely state of the art, but it is >well tested and very reliable. I'm painfully aware that more modern >systems, and most versions of V.3.2, are not quite as reliable. It >supports multiport serial very well, it supports TCP if you don't need >NFS, and is generally a good solution to your particular problem. I have found System V R3.2.2 (ie latest releases from the respective vendors) to be very reliable. The only panic's we have had in the past year have been due to memory problems. The only reboots we have had to do have been because of a hung archive tape drive (that's archive's problem not the OS). We bang the hell out of our system and have had it up for months at a time with no problems. >| 3. disk controller? Adaptec 1740? > > Garder variety ESDI. Use two drives to improve performance. I will Two drives does not get you that much performance improvement on ESDI. However it does get you much more with a SCSI system. I would recommend SCSI (perhaps adaptec 1542, or bustek equivalent) especially if the OS you choose is ISC (because ISC has highly optimized SCSI performance). The additional benefit of SCSI is that the tape drive can use the same controller (thereby saving costs of the controller and saving precious slots and interrupts). >| 4. 16 port serial card? > > Two 8's. You might find performance a hair better one way or the >other, but redundancy says use two, and there's no drawback I can see. >*Smart* cards, please. I would recommend a single card. (Slots are precious commodities, don't waste them). The card I like best (and of course this is just personal opinion) is the Megaport-24. -- Conor P. Cahill (703)430-9247 Virtual Technologies, Inc., uunet!virtech!cpcahil 46030 Manekin Plaza, Suite 160 Sterling, VA 22170
bill@astph.UUCP (Bill Dripps) (12/23/90)
In article <2710@sixhub.UUCP> davidsen@sixhub.UUCP (bill davidsen)
recomends 16MB memory and two eight port serial cards.
Be aware that some serial cards conflict with 16MB of memory.
Ask before you buy.larry@nstar.rn.com (Larry Snyder) (12/23/90)
cpcahil@virtech.uucp (Conor P. Cahill) writes: >In article <2710@sixhub.UUCP> davidsen@sixhub.UUCP (bill davidsen) writes: >> I would go with Xenix. It's not remotely state of the art, but it is >>well tested and very reliable. I'm painfully aware that more modern >>systems, and most versions of V.3.2, are not quite as reliable. It >>supports multiport serial very well, it supports TCP if you don't need >>NFS, and is generally a good solution to your particular problem. >I have found System V R3.2.2 (ie latest releases from the respective >vendors) to be very reliable. The only panic's we have had in the past >year have been due to memory problems. The only reboots we have had to >do have been because of a hung archive tape drive (that's archive's problem >not the OS). We bang the hell out of our system and have had it up for >months at a time with no problems. I agree with Conor - I can't image anyone with the hardware resources these days wanting to go with Xenix. If he has 10 grand (I believe that was the amount) to work with - he could put together a screamer Unix based system with FAST SCSI drives. The Xenix file system is so very slow - and the chance that continued development for third party applications is grim. >>| 3. disk controller? Adaptec 1740? >> >> Garder variety ESDI. Use two drives to improve performance. I will >Two drives does not get you that much performance improvement on ESDI. However >it does get you much more with a SCSI system. I would recommend SCSI (perhaps >adaptec 1542, or bustek equivalent) especially if the OS you choose is >ISC (because ISC has highly optimized SCSI performance). Again, I agree with Conor - a pair of SCSI drives that support SYNC transfers will scream past all the ESDI systems I've seen (including the 15m variety).. >The additional benefit of SCSI is that the tape drive can use the same >controller (thereby saving costs of the controller and saving precious >slots and interrupts). Plus - DAT and optics - and lots of devices without the need to add another controller.. >>| 4. 16 port serial card? >> >> Two 8's. You might find performance a hair better one way or the >>other, but redundancy says use two, and there's no drawback I can see. >>*Smart* cards, please. >I would recommend a single card. (Slots are precious commodities, don't >waste them). The card I like best (and of course this is just personal >opinion) is the Megaport-24. I don't have enough experience with this to comment.. -- Larry Snyder, Northern Star Communications, Notre Dame, IN USA {larry@nstar.rn.com, uunet!nstar!larry, larry%nstar@iuvax.cs.indiana.edu} backbone usenet newsfeeds available Public Access Unix Site (219) 289-0282 (5 high speed lines)
jay@metran.UUCP (Jay Ts) (12/23/90)
In article <1990Dec22.143706.4493@virtech.uucp>, cpcahil@virtech.uucp (Conor P. Cahill) writes: > In article <2710@sixhub.UUCP> davidsen@sixhub.UUCP (bill davidsen) writes: > > >| 4. 16 port serial card? > > > > Two 8's. You might find performance a hair better one way or the > >other, but redundancy says use two, and there's no drawback I can see. > >*Smart* cards, please. > > I would recommend a single card. (Slots are precious commodities, don't > waste them). The card I like best (and of course this is just personal > opinion) is the Megaport-24. One of my clients suffered a direct lightning strike to their building this summer. The computer manager watched the electricity arc across the ceiling. The only thing that kept them from being able to do useful work with their computer afterwards was a blown multiport I/O card. When the mp card was removed from their system, it booted and ran workably, but of course all of their terminals were useless. I think this sort of situation is what Bill is referring to when he says "redundancy says use two". It's a limitation, of course, if you don't have enough slots, and it's probably more expensive, too. Also, if a 2-card system were hit by lightning, maybe both would fail. Which bridges nicely into my main point. Arnet reports that 70% of their cards were returned due to surges on the serial lines. Upon inspection, the card referred to above had a (physically) blown line driver chip. Later, we found that a few of the Wyse 60 terminals had the same problem (they use the same industry standard chips). I read recently (in an electronics industry mag, I think; sorry I don't have a definitive reference) that even if you have a UPS on the computer, and every terminal and other peripheral has a surge protector, not only is the system still vulnerable to this problem, it is MORE vulnerable! Problem is, the MOV surge protectors shunt the electrical surge to ground, which sets up massively different electrical ground voltage potentials between equipment plugged in at different places in the building's wiring. Say, in different rooms. The result of this is that a surge is created in the serial line, which suddenly becomes the electrical connection between equipment running off grounds that are several thousand volts apart. So here's my recommendation: Arnet now includes port surge protection as standard equipment in their multiport cards. Try the Smartport-16s. Jay Ts, Director Metran Technology uunet!pdn!tscs!metran!jay
larry@nstar.rn.com (Larry Snyder) (12/23/90)
jay@metran.UUCP (Jay Ts) writes: >So here's my recommendation: Arnet now includes port surge protection as >standard equipment in their multiport cards. Try the Smartport-16s. The last thing I heard about arnet boards is that they had problems supporting bidirectional communications for a bank of modems locked at 38,400 bps using only hardware flow control under Interactive Unix. -- Larry Snyder, Northern Star Communications, Notre Dame, IN USA {larry@nstar.rn.com, uunet!nstar!larry, larry%nstar@iuvax.cs.indiana.edu} backbone usenet newsfeeds available Public Access Unix Site (219) 289-0282 (5 high speed lines)
cpcahil@virtech.uucp (Conor P. Cahill) (12/24/90)
In article <358@metran.UUCP> jay@metran.UUCP (Jay Ts) writes: >One of my clients suffered a direct lightning strike to their building this From my experience this is a very rare occurence (not lightning striking, but lightning strikes causing damage to serial port cards). >Arnet reports that 70% of their cards were returned due to surges on the >serial lines. > >Arnet now includes port surge protection as >standard equipment in their multiport cards. This sounds real fishy to me. Imagine me saying something like "70% of the software developed without CPC standards ends up panicing the system", so you should buy my software which happens to be built with CPC standards. The other way to look at it is that maybe the arnet boards are soo bad that the blow apart form the smallest voltage spike or short. Don't take me wrong. I'm not saying how good or bad the arnet cards are. I'm only saying that lightning is not a major concern for day to day operations of a "normal" system (of course if the system must have 100% uptime even in the worst set of conditions AND the client is willing to pay the price, you can use the extra money to provide lightning arrestors). -- Conor P. Cahill (703)430-9247 Virtual Technologies, Inc., uunet!virtech!cpcahil 46030 Manekin Plaza, Suite 160 Sterling, VA 22170
bill@astph.UUCP (Bill Dripps) (12/24/90)
In article <358@metran.UUCP> jay@metran.UUCP (Jay Ts) writes: > >One of my clients suffered a direct lightning strike to their building this >summer. The computer manager watched the electricity arc across the ceiling. >The only thing that kept them from being able to do useful work with their >computer afterwards was a blown multiport I/O card. When the mp card was >removed from their system, it booted and ran workably, but of course all of >their terminals were useless. I think this sort of situation is what Bill >is referring to when he says "redundancy says use two". > >It's a limitation, of course, if you don't have enough slots, and it's >probably more expensive, too. Also, if a 2-card system were hit by lightning, >maybe both would fail. Which bridges nicely into my main point. > >Arnet reports that 70% of their cards were returned due to surges on the >serial lines. Upon inspection, the card referred to above had a (physically) >blown line driver chip. Later, we found that a few of the Wyse 60 terminals >had the same problem (they use the same industry standard chips). > >So here's my recommendation: Arnet now includes port surge protection as >standard equipment in their multiport cards. Try the Smartport-16s. > > Jay Ts, Director > Metran Technology > uunet!pdn!tscs!metran!jay I must reply to this. We have two Smartport-16's. They are both on the shelf. They may be superior electrically - I wouldn't know - but their ISC driver has bugs: - Modem control is not correct especially at high speed. Expect lots of hung modems and other problems. - Transparent print does not work if ioctl's are issued to the terminal. For example it will not work with vi or ksh. The terminal will hang until the print finishes. - SL/IP will not work at all - not supported. We spent weeks and months with Arnet's tech support trying to get these bugs fixed. We were shipped several updated drivers. None fixed these bugs. We were told among other things: - We caused the modem problems with improper setup. (Why do our modems work with other serial boards? We allowed Arnet to dial into our system and set it up their way - no improvement) - The transparent print problem is a feature! They designed the driver to freeze the terminal if the program running on the terminal issues an ioctl until the print job finishes. This can take a very long time. - They had no intention of ever making SL\IP work because ISC refused them proper documentation. (Needless to say, ISC was appalled at this!) We finally asked Arnet to honor their "Rock Solid" guarentee and give us our money back. They refused. Is this the kind of relationship you want with your vendor? If all you need is a few 9600 bps terminals Arnet is fine. If you need more .... My advice with all serial boards is try before you buy, but no need to try Arnet! Any rebuttal from Arnet? I will gladly eat this article and more, if you fix the problems, but I will settle for my money. Bill Dripps
bill@unixland.uucp (Bill Heiser) (12/24/90)
In article <1990Dec23.113905.4278@nstar.rn.com> larry@nstar.rn.com (Larry Snyder) writes: > >The last thing I heard about arnet boards is that they had >problems supporting bidirectional communications for a bank >of modems locked at 38,400 bps using only hardware flow control >under Interactive Unix. > Speaking of hardware flow control ... My new AST 4-port board has an option for setting enabling hardware flow control. When initasy is run (on Esix), that's one of the questions that is asked. There's also something about a "flow control register." Neither of these is documented well (if at all) in the AST board documentation. What's your recommendation on this? bill -- home: ...!{uunet,bloom-beacon,esegue}!world!unixland!bill bill@unixland.uucp Public Access Unix - Esix SYSVR3 508-655-3848(12/24) 508-651-8723(12/24/96-HST) 508-651-8733(12/24/96-PEP-V32) other: heiser@world.std.com
larry@nstar.rn.com (Larry Snyder) (12/24/90)
bill@astph.UUCP (Bill Dripps) writes: >My advice with all serial boards is try before you buy, but no need to >try Arnet! Any rebuttal from Arnet? I will gladly eat this article and >more, if you fix the problems, but I will settle for my money. Excellent advise. We picked up the Computone from Tech Data on net 30 - so in case it did't work, we could return it. Likewise, we tried a Digiboard and returned it under their 30 day trial. In all fairness to Digiboard, I have heard from others that they have fixed the problems in their drivers. -- Larry Snyder, Northern Star Communications, Notre Dame, IN USA {larry@nstar.rn.com, uunet!nstar!larry, larry%nstar@iuvax.cs.indiana.edu} backbone usenet newsfeeds available Public Access Unix Site (219) 289-0282 (5 high speed lines)
bill@bilver.uucp (Bill Vermillion) (12/24/90)
In article <1990Dec23.162033.3287@virtech.uucp> cpcahil@virtech.UUCP (Conor P. Cahill) writes: >In article <358@metran.UUCP> jay@metran.UUCP (Jay Ts) writes: >>One of my clients suffered a direct lightning strike to their building this >From my experience this is a very rare occurence (not lightning striking, but >lightning strikes causing damage to serial port cards). >>Arnet reports that 70% of their cards were returned due to surges on the >>serial lines. Gawd - I hope they meant 70% of the returned cards were due to surges ... That I can understand, but having 70% of the cards returned ... ? >The other way to look at it is that maybe the arnet boards are soo bad >that the blow apart form the smallest voltage spike or short. >Don't take me wrong. I'm not saying how good or bad the arnet cards are. >I'm only saying that lightning is not a major concern for day to day >operations of a "normal" system (of course if the system must have 100% >uptime even in the worst set of conditions AND the client is willing >to pay the price, you can use the extra money to provide lightning >arrestors). Lighting does strange things at times. I had a system that appeared to take some sort of surge about 3 months ago. One of the surge surpressors on a terminal even showed evidence of smoke at the ac connection. Three Wyse 60's had some problem, one of which would receive but not transmit. We had an Anvil Onboard, and the diagnostics showed that I 3 ports out of 8 that were not testing properly. Some had handshake problems, others showed data line problems. Anvil said send out the board. They returned it with NO PROBLEMS FOUND. Tested again - same problem. This time they found asked to return the connector panel. All it has is two cables, pin connectors on each cable, and a pwa going to 12 db 25s. This panel was the FIRST panel failure they ever had. While all the serial ports themselves were okay. Lightning is fickle. It is non-predictable, and never appears to do the same things twice (except hit twice in the same place!). bill -- Bill Vermillion - UUCP: uunet!tarpit!bilver!bill : bill@bilver.UUCP
jay@metran.UUCP (Jay Ts) (12/29/90)
In article <1990Dec24.152633.5195@bilver.uucp>, bill@bilver.uucp (Bill Vermillion) writes: > In article <1990Dec23.162033.3287@virtech.uucp> cpcahil@virtech.UUCP (Conor P. Cahill) writes: > >In article <358@metran.UUCP> jay@metran.UUCP (Jay Ts) writes: > >>One of my clients suffered a direct lightning strike to their building this > > >From my experience this is a very rare occurence (not lightning striking, but > >lightning strikes causing damage to serial port cards). > > >>Arnet reports that 70% of their cards were returned due to surges on the > >>serial lines. > > Gawd - I hope they meant 70% of the returned cards were due to surges ... > That I can understand, but having 70% of the cards returned ... ? Erps! I guess I must have wrote that wrong! To restate for clarity, 70% of the returned cards failed due to surges on the serial lines. > >The other way to look at it is that maybe the arnet boards are soo bad > >that the blow apart form the smallest voltage spike or short. The problem is with the industry standard line driver chips I wrote about earlier. These chips, 1488 and 1489 line drivers and line receivers, are prone to failure due to surges. They are used not only by multiport cards, but also Wyse 60 and other terminals and computing equipment. See below. > Three Wyse 60's had some problem, one of which would receive but not > transmit. I'll bet anything that you can find a blown 1488 chip in them. I've seen this problem both in the system that was hit by lightning and another client's system (which was not). BTW, on the terminals we opened up, we found that some of the 1488/9's are socketed. Wonder why... > >Don't take me wrong. I'm not saying how good or bad the arnet cards are. > >I'm only saying that lightning is not a major concern for day to day > >operations of a "normal" system (of course if the system must have 100% > >uptime even in the worst set of conditions AND the client is willing > >to pay the price, you can use the extra money to provide lightning > >arrestors). > > Lightning is fickle. It is non-predictable, and never appears to do the > same things twice (except hit twice in the same place!). Here in Florida, it *does* do that! I guess I should have pointed out earlier that I am working in southern Florida, which is the lightning capital of the U.S. Hence I am very conservative in my opinions regarding surge protection. I may seem to be worrying too much, but think about how much it will cost in downtime alone if a system is damaged by a single lightning strike or other power surge. Now I'll leave you with some time-honored advice: "An ounce of prevention is worth a pound of cure." "A stich in time saves nine." Jay Ts, Director Metran Technology uunet!pdn!tscs!metran!jay
larry@nstar.rn.com (Larry Snyder) (01/03/91)
bill@unixland.uucp (Bill Heiser) writes: >In article <1990Dec23.113905.4278@nstar.rn.com> larry@nstar.rn.com (Larry Snyder) writes: >> >>The last thing I heard about arnet boards is that they had >>problems supporting bidirectional communications for a bank >>of modems locked at 38,400 bps using only hardware flow control >>under Interactive Unix. >> >Speaking of hardware flow control ... My new AST 4-port board has an >option for setting enabling hardware flow control. When initasy is >run (on Esix), that's one of the questions that is asked. There's >also something about a "flow control register." Neither of these >is documented well (if at all) in the AST board documentation. What's >your recommendation on this? Use hardware flow control (if it works). Try turning it on in the modem, and in the driver (FAS?). Then try downloading then uploading a file using zmodem at both a high speed and a 2400 baud connection - with the DTE locked at either 19200 or 38400. If you get a flawless transfer (no errors) then you should be all set - however - if flow control isn't working - you will sure enough see errors.. -- Larry Snyder, Northern Star Communications, Notre Dame, IN USA {..!uunet!mailrus!iuvax!ndcheg!nstar!larry, larry%nstar@ndcheg.cheg.nd.edu} backbone usenet newsfeeds available Public Access Unix Site (219) 289-0282 (5 high speed lines)