pwolfe@kailand.KAI.COM (12/05/88)
I've recently received information about the EXABYTE EXB-8200 8mm cartridge tape subsystem, with which one may store up to 2 Gigabytes (2,000 Megabytes) on a single tape cartridge. Sounds very attractive for running unattended overnight backups. It includes an integrated SCSI controller and formatter, onboard error recovery, and 256 Kb buffer. The SCSI interface meets the ANSI X3.131-1986 specification. National Peripherals sells the EXB-8200 for about $4,000, including cables and desktop cabinet (it's a 5-1/4" form factor - maybe this fit inside the Sequent system cabinets?). This device is supposed to work with Sun workstations, so I assume device driver software already exists somewhere. Actually, I was wondering if any gurus out there could tell me: - how difficult is device driver software to work on? - can anybody relate any experiance with this particular device? - is Sequent's SCSI (ANSI X3.T9.2) capable of supporting this drive? - do you have any info on connecting SCSI peripherals to Sequents? Patrick Wolfe (pat@kai.com, kailand!pat) System Manager, Kuck and Associates, Inc.
cs002@unocss.UUCP (Stan Wileman) (12/06/88)
In article <2400029@kailand>, pwolfe@kailand.KAI.COM writes: > > I've recently received information about the EXABYTE EXB-8200 8mm cartridge > tape subsystem, with which one may store up to 2 Gigabytes (2,000 Megabytes) > on a single tape cartridge. Sounds very attractive for running unattended > overnight backups. > > - can anybody relate any experiance with this particular device? > - is Sequent's SCSI (ANSI X3.T9.2) capable of supporting this drive? > - do you have any info on connecting SCSI peripherals to Sequents? I'm not certain how useful this will be, but the folks at Perfect Byte tested the EXABYTE on our B8K, connecting it in place of the usual cartridge drive. The DYNIX driver for the cartridge tape worked reasonably well, although the EXABYTE supports some operations not included in the driver. We were successful in reading and writing with no significant changes whatever. -- Stan Wileman uucp: uunet!btni!unocss!cs002 Math/Comp. Sci. Dept. Inet: cs002@unocss.unl.edu U. of Nebraska at Omaha Bitnet: cs002@unoma1.bitnet
ronc@fai.UUCP (Ronald O. Christian) (12/07/88)
In article <2400029@kailand> pwolfe@kailand.KAI.COM writes: >Actually, I was wondering if any gurus out there could tell me: > - how difficult is device driver software to work on? > - can anybody relate any experiance with this particular device? > - is Sequent's SCSI (ANSI X3.T9.2) capable of supporting this drive? > - do you have any info on connecting SCSI peripherals to Sequents? We have an Exabyte on order. We're going to experiment with it on our Symmetry 81. If there's people out there that are interested in the results, I can post our experiences. If memory serves, Sequent is evaluating the Exabyte even as I write. (Source: The last SURF meeting.) As I recall, the issue was not the interface between the Exabyte and the Symmetry, but that the Exabyte had reliability problems. I guess we'll find out. We're presently buying over $9000 worth of 9 track tapes a year for daily incrementals, and *that* has *got* to *stop*! According to the Exabyte rep, the drive looks like an Archive streamer (with a very looooong tape) to the SCSI controller. He said that at least on the Sun, the same driver works for the Archive and th Exabyte. We are betting that the Archive driver on the Symmetry will also talk to the Exabyte. Ron -- Ronald O. Christian (Fujitsu America Inc., San Jose, Calif.) {amdahl, pyramid, sun, unisoft, uunet}!fai!ronc -or- ronc@fai.com
neal@hacgate.scg.hac.com (Neal Thompson) (12/08/88)
In article <1180@fai.UUCP> ronc@fai.fai.com (Ronald O. Christian) writes: ... > >If memory serves, Sequent is evaluating the Exabyte even as I write. >(Source: The last SURF meeting.) As I recall, the issue was not the >interface between the Exabyte and the Symmetry, but that the Exabyte >had reliability problems. I guess we'll find out. We're presently ... At a recent SURF meeting in Newport Beach, I asked a Sequent Tech Support guy about the Exabyte. He said that Sequent's SCSI was so finely tuned that he guaranteed the Exabyte wouldn't work. I'd be happy to hear that Sequent had solved that problem and could connect an Exabyte (or any similar drive) to the SCSI. Neal Thompson neal@hacgate.scg.hac.com
liam@cs.qmc.ac.uk (William Roberts) (12/10/88)
In the UK, Compass sell the Exabyte tape streamer device with a Pertec interface. This is good news because the Balance tape device driver has a Pertec interface, and so we can use the DYNIX we've got without modification. The cost of the Exabyte unit is #4944, the cost of the tape controller is #1683. Maintenance about #50 per month. Tapes are standard 8mm video tapes (prices as at July 1988). We have had this device on order for a long time now (months), but we are getting a curious extra: it will have a switch on the front which connects the drive to an alternative controller that in our case will be fitted inside our VAX 11/750 so that that can write streamer tapes as well. -- William Roberts ARPA: liam@cs.qmc.ac.uk (gw: cs.ucl.edu) Queen Mary College UUCP: liam@qmc-cs.UUCP LONDON, UK Tel: 01-975 5250
csg@pyramid.pyramid.com (Carl S. Gutekunst) (12/10/88)
Followups directed to comp.periphs. [Doesn't anyone at Sequent read their own newsgroup? Sheesh.] In article <1180@fai.UUCP> ronc@fai.fai.com (Ronald O. Christian) writes: >As I recall, the issue was not the interface between the Exabyte and the >Symmetry, but that the Exabyte had reliability problems. When I first saw the 8mm drives, I thought they could be a real godsend for machines with disk farms. (2 GB on a tape smaller than an audio cassette!) And for certain environments, they may well be. But the problems I observed were: - The 8mm drives are slow. You save a lot of storage space compared to 9-track reels (something like 13 9-track tapes to 1 8mm cartridge), but backups take about twice as long. Figure about two and a half hours per tape. Of course, you'll have to change tapes 13 times less often.... - The actual transport mechanism is made by Sony, no matter whose name is on the box. These are consumer quality, designed for use in Camcorders, with a low duty cycle. (Say, a few hours a week.) Run continuously, ours died in less than six weeks. Less use would give longer life, of course. - Media integrity is *highly* dubious. These tapes are again consumer quality, where all you have to do is get enough bits right to create a viewable TV picture. Data storage is a whole 'nother story. I would not expect an 8mm cartridge to hold usable data for more than a year or so. So you'll still need something like 9-track for monthly dumps and such. - Sony has been privately telling potential customers to wait for computer- grade DAT drives. These have about half the storage of 8mm, run at about the same speed. Supposedly both the media and the drives are more reliable than 8mm. One oddity, too. Pyramid and Sequent aren't selling to PC users; anyone who is spending >> $100K on a computer is buying something to set in the data center. It has to look impressive on the balance sheet. And there's this image problem of telling the Data Center staff that they should be buying blank backup tapes at K-Mart or Tower Records.... (Sony actually will sell you "data quality" tapes at about four times the price you can buy them from K-Mart. Same thing, as far as anyone can tell.) The natural technology for production use would seem to be the good ol' 1/4" cartridge. The state-of-the-art here seems to be QIC-150 drives from Cipher, Tandberg, and others: 150MB on a 600 foot tape. This is essentially the same amount of storage as a 6250bpi 9-track reel, with comparable recording times. The interface is usually SCSI or Cipher. This is a long ways away from 2GB per tape. Given the relatively controlled nature of the transport in a 1/4" cartridge, you would think that more modern encoding techniques* would yield incredible amounts of storage, perhaps 10GB, with high speeds and high reliability. But for some reason no one is willing to pursue this; they're mucking about with toy tape drives instead. * If you want to get rich, figure out a cheap way to do helical scan on a 1/4" cartridge tape. <csg>
ronc@fai.UUCP (Ronald O. Christian) (12/14/88)
In article <50451@pyramid.pyramid.com> csg@pyramid.pyramid.com (Carl S. Gutekunst) writes: >Followups directed to comp.periphs. > >[Doesn't anyone at Sequent read their own newsgroup? Sheesh.] What he said. >In article <1180@fai.UUCP> ronc@fai.fai.com (Ronald O. Christian) writes: >- The 8mm drives are slow. You save a lot of storage space compared to 9-track > reels (something like 13 9-track tapes to 1 8mm cartridge), but backups take > about twice as long. Figure about two and a half hours per tape. Of course, > you'll have to change tapes 13 times less often.... This last part is the key. There's this great big space between about 11:00 P.M. and 5:00 A.M. where our machines are very lightly used. This is an ideal time to do backups, but the problem is, I can't afford to pay someone overtime to sit and watch the wheels turn. The big plus about getting 2 Gbytes on one tape is that you can arrange your backups so that tape change is unnecessary. Ergo, backups can be unattended. > [...] These are consumer quality, designed for use in Camcorders, with a > low duty cycle. (Say, a few hours a week.) Run continuously, ours died in > less than six weeks. Less use would give longer life, of course. This makes sense. But the issue with us is monthly level 0 backups. They take 24 nine-track tapes at present and require an operator to sit around waiting to change them. Assuming we do level 0 on the 8mm drive, we can fit an entire dump on two tapes. Surely two tapes a month isn't too high a duty cycle? My 8mm deck at home gets more use than that. Our cost in 9 track tapes for daily incrementals is $9K per year. For that price we can buy an 8mm drive every year and still come out ahead! >- Media integrity is *highly* dubious. These tapes are again consumer quality, > where all you have to do is get enough bits right to create a viewable TV > picture. Data storage is a whole 'nother story. It's true that data storage is another story. The data storage is digital with error correction, whereas the camcorder signal is analog. One would think that digital recording techniques would be more robust. In rec.video awhile back, someone asked if 9 track tape could be used as video tape. The consensus was that 9 track tape is poorer quality than consumer video tape, not the other way around. >I would not expect an 8mm > cartridge to hold usable data for more than a year or so. Whyever for? What's intrinsic about the 8mm tape that data would only last a year? Is this all speculation, or do you have hard data to back it up? Ron -- Ronald O. Christian (Fujitsu America Inc., San Jose, Calif.) {amdahl, pyramid, sun, unisoft, uunet}!fai!ronc -or- ronc@fai.com
csg@pyramid.pyramid.com (Carl S. Gutekunst) (12/18/88)
See George Goble's articles on Exabyte, which I have added to the references line. The drives Pyramid evaluated were made prior to June 1988, i.e., before Purdue and Exabyte ironed out some major problems. Much credit is due Purdue for working closely on Exabyte to solve their problems, as opposed to stone- throwers like me who just sit back and bitch. :-) Note that even the present drives do require the user to make a number of con- cessions. None of these would be a problem given the environment that Ron Christian described in his previous posting: periodic, overnight, unattended Level 0 backups. In article <1261@fai.UUCP> ronc@fai.fai.com (Ronald O. Christian) writes: [A lot of good reasons why the limitations of an 8mm drive don't cause any problems for his environment.] >> These are consumer quality, designed for use in Camcorders, with a low >> duty cycle. > >This makes sense. But the issue with us is monthly level 0 backups.... >Surely two tapes a month isn't too high a duty cycle? Of course not. But then you still need something else for incrementals. My point was simply that an 8mm drive alone doesn't necessarilly solve all your backup needs. But, as you observed, one can consider the drives disposable, relative to the very high cost of 9-track media. >>- Media integrity is *highly* dubious. > >It's true that data storage is another story. The data storage is digital >with error correction, whereas the camcorder signal is analog. One would >think that digital recording techniques would be more robust. All that digital means is that when the tape starts to deteriorate, you won't notice it for a while; the error correction will hide the deterioration. When the bit error count gets too high, whole 128Kbyte data groups will evaporate. >... The consensus was that 9 track tape is poorer quality than consumer video >tape, not the other way around. It's an apples and oranges comparison. 9-track tape needs to handle a bit den- sity of 112Kbits per square inch. 1/4" cartride using QIC-150 format, 600Kbits per square inch. For 8mm, we're talking 35 MILLION bits per square inch, and with DATs, a mind numbing 114Mbits. (For reference purposes, VHS digital is 13MBits.) The question of whether a 9-track tape can be used at high densities is not a quality issue, but a technology issue. There are manufacturers who are building 9-track, 1/4", and DAT drives and media to commercial/industrial specifications. No one is doing this for 8mm. And at 300 times the bit density, quality control is very important. >Is this all speculation, or do you have hard data to back it up? We have not done the kind of wear testing that the Purdue folks have, so what I know first-hand about tape wear is speculative. (What I saw about drive wear *is* first hand, but based on the Purdue results, somewhat dated.) We do have some hard data on tape wear characteristics, although the source is *ahem* somewhat biased: 3M Corp. :-) Please note that I am not flat-out rejecting the technology of helical scan 8mm (and smaller) tapes; I am asserting that what I have seen so far suggests that the technology is still immature. Much credit is due Purdue and Exabyte for their latest developments. But when Sony and Hewlett-Packard start to ship industrial-grade DAT drives early next year, I would guess that 8mm will prove to have been merely a transition technology. <csg>