DASD Stress Testing.

Anuj Dhawan · Posted: Tue Jul 17, 2012 11:15 am

Hi,

This thread SPOOL VOLUMES ARE FULL' issue during a longtime job talks about DASD stress testing and that got me curious.

I did a bit of homework with google and end-up with Storage Calculations, DASD Utilization Tables -- these are not what I'm looking for. Looks like "DASD stress testing" is yesteryear thing - however, it'd be nice to hear back from who had been involved in this.

It looks like - when DASD were evolving they did test them on their I/O speeds, response time, their block sizes -- but is this all, or there is more to it.

Thanks for stopping by,

dick scherrer · Posted: Wed Jul 18, 2012 5:55 am

Hi Anuj,

In addition to what you mention, we intentionally tried to "break" new devices and device types by putting more of a load on them than would be seen in "real world" situations. Sometimes either a new device of an old model or some new "improved" device failed . . .

We also had some benckmark jobs that we ran on any new device or channel/device configuration to ensure critical systems would not be negatively impacted by the change.

d

Pedro · Posted: Thu Jul 19, 2012 5:50 am

It was fairly early in my career, so I do not recall all that we did.

Similar to Dick, our effort was to try to break new devices. If it did not break in the first week of testing, the new DASD was added to our scratch volumes, then possibly later for the SMS primary storage pools.

I recall having to write something on each byte of the track, on each cylinder. And doing so multiple times.

Sometimes we would find hardware errors with the read/write head.

Sometimes we found surface defects. When it cannot write to a track, an alternate is used. I think there were only 15 alternate tracks. And if you used most of them the first few day, the DASD was reported to the vendor.

expat · Posted: Thu Jul 19, 2012 1:04 pm

That's why there are SMS storage classes defined for defining required response times

David Robinson · Active User Joined: 21 Dec 2011 Posts: 199 Location: UK

I rememer a long time ago when I was in Ops a sysprog "repairing" some damaged DASD. I was impressed at the time, but although it's not something I've ever done I guess it was just using ICKDSF to mark a track as unavailable.

Probably not something we ever do nowadays.

Anuj Dhawan · Posted: Thu Jul 19, 2012 4:20 pm

Thanks for reading through.

expat · Posted: Thu Jul 19, 2012 5:03 pm

Probably the old fixed platter types, 3330, 3350, older 3390, which could be physically removed from the drives

Bill Woodger · Posted: Thu Jul 19, 2012 6:57 pm

3301 Model 1 had removable platters :-) Two seniors swore to me that OPS lost an hour when they got one "stuck" half-in-half-out.

Even after all this time, I'd appreciate knowing if that was possibly true. I still think the little minkies were pulling-my-leg.

expat · Posted: Thu Jul 19, 2012 6:58 pm

Could have been a mechanical failure Bill, it did happen from time to time

Robert Sample · Posted: Thu Jul 19, 2012 7:10 pm

Back in my college days when I worked in the school's computer center, I used to mount 2314 disks, so I can tell you that the disks could get stuck (it was rare, but sometimes the hardware failed). And the noise they made when the power went off ... WOW! They had very strong relays that the disk arm motor pulled against. If the power went out, the relays yanked the arms off the platters before (theoretically) any physical damage could occur. They were quite loud. Once that happened, the IBM CE had to reset the relays before the drives could be powered up again. We kept a disk pack around where there had been a hardware failure. Hold the disk pack up to the light at the right angle, and you could see through the hole that went through all the platters.

dbzTHEdinosauer · Posted: Thu Jul 19, 2012 7:12 pm

back in the early 70's i worked on a lot of ncr hardware.
they had removable discs,
problem was, you removed the dust cover then closed the cabinet.

needless to say, the discs would have read/write errors,
and since it was normally due to the dust
(or cigarette ashes from the operators)
not only was the platter physically scratched, it normally meant the read/write heads were also damaged.
compound that with an unknowing operator,
who would mount the unreadable disc in another unit (in an attempt to read the disc)
....
one night we lost 5 units to this silliness until I made it to the operations room
and flung the disc, frisbee-like, into a cement wall.
nobody said a word to me, we recovered and reran the previous days run, then ran the update for that night.
fortunately we had 20 units (actually, by this time only 15).

removable disc units - talk about planned-obsolescence.

Akatsukami · Posted: Thu Jul 19, 2012 7:40 pm

About 30 or so years ago, I worked at a company that at one time had had a GE-235 computer. When it was finally scrapped, an operator took a disk "platter" (more like a cartwheel, IIRC) and made a tabletop from it.

dick scherrer · Posted: Thu Jul 19, 2012 10:24 pm

Hello,

expat · Posted: Thu Jul 19, 2012 10:38 pm

Dick, I'm pretty sure that the Memorex 3350's were removables

dick scherrer · Posted: Thu Jul 19, 2012 11:30 pm

Hi Expat,

Bill Woodger · Posted: Fri Jul 20, 2012 12:28 pm

Our 3350s were Memorex. Definitely non-removable, although there could have been other models?

Whole place was "plug compatible" apart from the 3800 printers.

Anuj Dhawan · Posted: Fri Jul 20, 2012 2:24 pm

I feel overwhlemed with the responses in this thread. Thanks for stopping by,

Regards,

Anuj Dhawan · Posted: Fri Jul 20, 2012 2:44 pm

Anuj Dhawan · Posted: Fri Jul 20, 2012 4:00 pm

Thanks expat.

expat · Posted: Fri Jul 20, 2012 4:17 pm

Those were the days when storage management would analyse the dataset usage and then position datasets on physical volumes for performance reasons.