deboni@fernando.llnl.gov (05/30/90)
The following is a tale of mistakes I made trying to provide file service for a bunch of Macintioshes. I believe my experiences indicate the superiority of certain ways of doing things, over others, and of certain commercial products over others. Take it for what it's worth. Once upon a time, I bought a CDC Wren VI SCSI drive, intending to it hook to a Sun 3/280 file server. This site has an Ethernet backbone with a lot of Apple- talk zones bridged to it by Kintetics Fastpaths. My group has quite a few Macs, a Fastpath, and the Sun, and we needed a Mac file server for the group. We bought TOPS for the Macs and the Sun. Aftern fooling around with the software, we thought we understood it, and that the hardware would be easy to deal with. Our first paroblems occurred when we tried to hook the Wren to the Sun and get it working. The documentation was subminimal, and we had a bad time dealing with the issues of termination and SCSI unit number (address). We're reasonably knowledgeable, but Sun does SCSI stuff differently from Apple. We put the drive on a chain originating in the Sun, which had a built-in cartridge drive. The cartridge drive had long since proven inadequate, so we'd previously bought a nice little Exabyte 8mm cassette drive. It worked fine, with a switch on the back labeled "Termination", in the OFF position. This violated my expectations, but there's no sense arguing with good. In the next few weeks, I hooked the Wren to the SCSI chain in a variety of ways, but failed to get satisfactory behavior out of it. It would often appear to be working for a while, and then seem to freeze in the middle of a file transfer, locking up the participating Mac. I spent a lot of time, trying to decipher the photocopied manual and fiddling with jumpers. I tried it in the middle and at the end of the chain, terminated and unterminated. I tried changing the motor start option, and the source of the termination power. I even tried changing the state of the Termination switch on the tape drive. Eventually, with luck and the help of some kind folks who answered my posting to one of the SUN newsgroups, I managed to get the two drives to work on the chain together. The Wren is a 766 MB drive, but it came preformatted, with a usable Unix volume structure, which reduced its capacity to something over 650 MB. After my Unix wizard installed a file system on it, it had only a bit over 600 MB of usable space. This seemed like a large amount of overhead to me, but the wizard said that such overheads are normal, and I decided not to worry about it. The Wren was accessible from the Mac side of the net via TOPS, and could be backed up to the Exabyte. We used the Wren as a dedicated volume for Mac files, linked into our home directories on the SUN. We worked this way for several months. The only trouble was that it was very slow. TOPS for the Mac allows one to mount Sun directories on a Mac desktop that show up as disk volume icons. These are then manipulable the same as any other such volumes/icons. However, under normal daily loading, TOPS would alert and beep my Mac every thirty seconds or so during file movements (and sometimes even when the Mac was idle) and tell me that it was having trouble connecting with my Sun. If I left it alone, the alert would disappear in a few seconds, and then reappear in another half minute or so. The file movements tended to be very slow. For a while I thought the drive was at fault. The dearler assured me it was not. Then I thought TOPS was at fault. Their tech support folks were sympathetic but unhelpful. Then I thought maybe the BSD Unix-based Sun OS was somehow not scheduling the TOPS process often enough. My local Sun high priestess and wizard quickly refuted this sinful and unwholesome speculation. Finally, I got tired of it all, got up on my hind legs and demanded some resources for an Appleshare file server. I finagled my way into commandeering a spare (?!) Mac Plus and buying the Appleshare server software for it. I backed up the Wren drive to tape, removed it from the Sun, and put it on the Mac Plus. Then I tried to use La Cie's SilverLining to format the disk, install a driver for the Mac, and partition the drive into several volumes. However, the process was still not yet Mac-simple. First, I had to find a cable. The Wren has a pair of 3-row DB-50 sockets for SCSI connections. Macs have 2-row DB-25's, and their drives usually have Centronics-style 50-pin connectors. I posted another article to another newsgroup about it, and got several friendly responses, some suggesting sources for a custom cable. The one that solved the problem for me suggested I open the Wren drive box, unhook the drive itself from the cable connecting to the DB-50, and substitute a new internal cable with a Centronics connector on it. All SCSI drives use a standard 50-pin ribbon connector on their controller cards. I scrounged an internal ribbon cable from a dead Jasmine drive and installed it. Then I went through the termination and addressing battle again. But this time I got smart and started out by removing all the jumpers from the drive controller board. Setting the address at zero seemed reasonable. The motor start option did not seem likely to cause me trouble, so I let it default to whatever no jumper means. However, termination continued to be a problem. I tried pulling both the termination power jumper and the resistor pack, as advised by the drive's alleged documentation. Then I put the reistors back and tried jumpering termination power from the cable, that seeming like a reasonable way to do it. Then I made a startling discovery: The Wren docs showed a picture of a row of seven jumper pin pairs, but there were actually eight pairs on the drive controller card. I'd probably been off by one, in one direction or another, in everything I'd tried to do so far. After careful examination, I managed to convince myself the one of the jumpers at one end was not connected to anything, and proceeded on that basis. However, I still was having troubles. I was using SCSI Tools and SCSI Probe on the Mac to test whether the drive was visible on the Mac's SCSI chain. The problem was, it never was. I discovered that adding a DataFrame 40 onto the end of the chain made both drives visible, but I didn't want to try formatting the Wren with another drive running on the system. (Just overcautious, I guess.) Finally, in desperation, I just went ahead and launched SilverLining, in spite of the Wren's invisibility to SCSI Tools/Probe, and to my surprise, It was able to see the drive and recognize its type and size. At this point, I went with the flow and formatted the drive, installed a driver, and initialized the drive (does anyone out there know what the difference is between "Formatting" and "Initializing" a SCSI drive?). I was puzzled that the Wren still wound up with only around 650 MB of usable space, but I was encouraged that it was working. I broke the drive up into four volumes, and installed the Appleshare Server software on one of them. To this date, some weeks later, the server is still working fine. I was most gratified when, early on in the process, I ran a quick and dirty preformance test, and found that the Appleshare server could provide a file to my Mac II in less than one fourth the time required by the Sun/TOPS Ether/Fastpath/Appletalk combo, and around one half the time required to get it from another Mac Plus running TOPS. The permanent solution to the cable problem is going to be achieved by getting a two-foot length of 50-conductor ribbon cable, a press-fit SCSI controller connector, and a couple of press-fit Centronics connectors, and manufacturing it in-house. I'm told that all these parts can be bought at electronics supply stores. But I have techs that know about such stuff. I'm just glad it all finally works well enough to depend on!
jskuskin@eleazar.dartmouth.edu (Jeffrey Kuskin) (05/30/90)
[Huge description of SCSI woes by deboni@fernando.llnl.gov deleted] Be aware when connecting non-Mac SCSI drives to a Mac that the Mac does NOT supply termination power. If the device in question expects to get its termination power from the SCSI connector, you will have many potential problems. One way to test for this setup is to stick a Mac-specific peripheral at the end of the SCSI chain. Since it should be terminated properly (without relying on the connector's termination power), everything will work fine. If you place the non-Mac device at the end of the chain and things stop working, you might be experiencing the aformentioned termination power problem. -- Jeff Kuskin, Dartmouth College
dan@hpnmdla.HP.COM (Dan Pleasant) (05/30/90)
>[Huge description of SCSI woes by deboni@fernando.llnl.gov deleted] > >Be aware when connecting non-Mac SCSI drives to a Mac that the >Mac does NOT supply termination power. If the device in question >expects to get its termination power from the SCSI connector, >you will have many potential problems. One way to test for this >setup is to stick a Mac-specific peripheral at the end of the >SCSI chain. Since it should be terminated properly (without >relying on the connector's termination power), everything will >work fine. If you place the non-Mac device at the end of the chain >and things stop working, you might be experiencing the aformentioned >termination power problem. > >-- Jeff Kuskin, Dartmouth College ---------- Having had recent experience with just this problem, I can assure you that the Mac IIfx *does* supply termination power, and the Mac Portable *does not*. I can't speak about the other models with certainty, but I think they do supply termination power. Dan Pleasant
dplatt@coherent.com (Dave Platt) (05/31/90)
In article <22391@dartvax.Dartmouth.EDU> jskuskin@eleazar.dartmouth.edu (Jeffrey Kuskin) writes: > Be aware when connecting non-Mac SCSI drives to a Mac that the > Mac does NOT supply termination power. That is true of the Mac Plus, and [I believe] of the Mac Portable. It is not true of the SE, SE/30, and all members of the Mac II family... these machines do provide termination power (unless you inadvertently short out the terminator-power line and blow the little surface-mount fuse on the motherboard).
alex@grian.cps.altadena.ca.us (Alex Pournelle) (06/28/90)
RANDO@applelink.apple.com (Damon Rando) writes: >Both the Portable and the Mac Plus do not supply TERMPWR from the CPU. >All other Macs do suppy TERMPWR. Most drives also supply TERMPWR for the >external terminator. It is best to use a drive that supplies TERMPWR as >the last drive in a SCSI chain due to the transmission line >characteristics of the SCSI chain. Drive manufactures are now increasing >the capacitance at the TERMPWR supply to decrease the probability of >glitches on the line (ie. it's important that the drive supply its own >TERMPWR). The goal in SCSI termination is to keep the TERMPWR line at +5V >with as little noise as possible. A few, like Everex in several of its guises--most notably EMAC--have new spiffy boxes with a switch marked "Plus/Other", which optionally supplies external termination power. The older SuperMAC 20-110 meg drives have OMTI interface cards that can also be switched. But I still believe his problem this time is Unit Attention, which afflicts especially Pluses. Even brand-spankin' new drives need this line cut. (I made a previous post on this earlier today.) >This is the >reason for the new black terminator in the new IIfx box. Note (**1 is >better 2 is bad** NEVER USE MORE THAN ONE BLACK TERMINATOR) And boy, it sure makes a difference in SCSI performance. My dad's IIfx was dead slow with the grey one, but the black one makes it really scary. Good job, guys! Alex "It is said by many that Pournelles never offer praise. This is not true--it is merely that praise is only offered where it is truly deserved." --Surviving in a World of Type A Personalities, Volume One, Page 943 (also discussed in Volume Eleven) -- Alex Pournelle, freelance thinker Also: Workman & Associates, Data recovery for PCs, Macs, others ...elroy!grian!alex; BIX: alex; voice: (818) 791-7979 fax: (818) 794-2297 bbs: 791-1013; 8N1 24/12/3