Jump to content
  • entries
    0
  • comments
    0
  • views
    324

DEC And Computer System Hardware Integrity


xxmikexx

128 views

In the Backup thread j flanagan brought up the subject of computers manufactured by DEC a long time ago. I said I would open a new thread for that subject. This is that thread ...

 

xxxxxxxxxxxxxxxxxxxx

 

j flanagan,

 

It wasn't all sweetness and light regarding DEC computer system integrity. While our CPUs were very very reliable, our disk drives were not -- could not be -- comparably reliable. The affordable technology of the time simply did not permit it. The reason is that the hard drives of the time were prone to head crashes, something that is encountered today only very infrequently but was a constant concern during the 70s, which was the time frame of my employment at DEC.

 

The classical head crash happened as follows ...

 

At some point, for whatever reason a particle would become dislodged from the recording medium. If the particle came from a recording track then the parent sector would go bad. But more likely the particle came from the area between tracks so that the flanking sectors would remain good. However ...

 

Now we have a particle flying around inside the drive case because it was flung off the recording medium by the fast rotation of the drive. Usually such a particle would bounce around between the walls of the housing and the spinning recording medium, ending up settling at the bottom of the housing. Unfortunately, once in a while the dislodged particle would strike hard elsewhere on the medium, gouging it, and now we have two particles flying around. Most of the time the two particles will settle, but sometimes instead a cascade will build. When the cascade begins, death of the drive is only minutes away.

 

When a cascade gets going, sooner or later a particle is going to encounter one of the flying heads, trying to fit itself in between the head and the recording medium. If the size of the particle is roughly that of the head flying height the particle is quite likely to induce head flutter, and this flutter inevitably resulted in a literal head crash, with the head impacting the recording medium just like an airplane hitting the ground, with equally catastrophic results.

 

xxxxxxxxxxxxxxxxxxxx

 

So the wise customer had data on duplicate hard drives, and the hard drives themselves would be backed up to magtape. Trouble is, with magtape the heads wear and without constant finicky maintenance there can be no guarantee that an tape years or even months earlier will be readable on the drive that created it. Thus multiple tape drives were a good idea, and of course tapes had to go offsite in duplicate if one was to minimize risk of data loss. The head wear issue encouraged the three-generations of tapes backup philosophy of the day.

 

xxxxxxxxxxxxxxxxxxxx

 

Let the record show that webmaster Nels Anderson later worked at Prime Computer, who addressed the problem of system reliability by having a pair of computers for a single system, sharing memory, with one computer "shadowing" the other. The result of their very carefully designed system architecutre was that in general no data would be lost even if one of the machines went down.

 

But of course Nels cut his programming teeth on a high school computer donated by DEC, as so many other later industry professionals did. :)

Edited by xxmikexx

2 Comments


Recommended Comments

Hey Mike. First of all, feel free to call me John. We have several Johns where I work, so I get in the habit of signing my posts j flanagan to single me out.

 

Looks like some things never change. The mechanical components - disk drives, tape drives, etc - are always going to be the Achilles heel of reliability. Granted, things have gotten light years better even in the six years I've been in the industry.

 

I've heard some very good stories about hardware from the era. One of our procurement managers is a retired Marine Colonel who at one point in his career was responsible for the IT infrastructure for a west coast division. One of the systems that he was responsible for was a mainframe on one of the command and control ships (I believe it may have been the Blue Ridge, but don't quote me on that). The ship was just coming out of the dry dock and was going out to see for a few days of shake downs. On the list of tasks was a shock test to verify hull integrity. My co-worker was on board to baby sit his mainframe and see how well it would survive the shock. The time of the test comes and there is a thunderous boom and the ship shakes violently. Every console goes out. Unfortunately, when the mainframe was installed, nobody took any effort to shock isolate the machine. Every drive in the machine had been hit with a head crash. Fortunately, his supervisor at the time had different ideas of a successful test. The mainframe was able to power on, so I guess it had technically survived.

 

It's very interesting that you bring up the work at Prime Computer. Virtualization is another task that is on my plate. The technology pendulum just keeps swinging back and forth. In the mainframe days, you had these monster computers who ran many virtualized systems within the mainframe. In the 1990's, with the cheap Intel x86's finally becoming powerful enough to handle the work load of the data center, everyone wanted to assign a stand alone physical server for everything. This trend has progressed until the last few years. Now we have all these extremely powerful and specialized machines that spend most of their time, essentially, idling and waiting for something to do. VMware comes along, and "invented" virtualization to run several virtual servers on a powerful server to maximize your hardware usage and lower your hardware costs. So, here we are, 20 years later...falling back to very large, powerful, and expensive systems...running many virtualized systems and services.

 

Mike, I know you like to ask people questions to get them talking. Now is my turn to get YOU talking! :) You mentioned Nels cutting his teeth on a machine donated by DEC. What sparked your interest in the field and pulled you to it? Let me know to start another blog post if you feel that this is straying too far from the original topic.

Link to comment

Flargan/John,

 

I always start reply posts with the forum name of the person I'm replying to so that readers will never get confused by the use of first names. However, I don't mean to be unfriendly. So from now on you are "Flargan/John", okay?

 

xxxxxxxxxxxxxxxxxxxx

 

Yes, we need a new thread for the matter of how I got into computing. That new thread is going to spawn other threads, and so it will go, as this conversation group gets larger and larger.

Link to comment
×
×
  • Create New...