Tape Read-After-Write Error Management with Inter Block Gaps

[firebirdta84]

I see that some tape drives were able to locate and overwrite one single block at a time, and I'm curious about how these drives handle Read-After-Write "Block Error Management" (my phrase).

I'm exploring how the Kennedy 6450 managed erroneous blocks on its read-after-write procedure, and I'm a bit baffled.

I am hoping that other more well-known older tape formats managed this a certain way, and I can at least follow that pastern of logic to start.

Here's my question:

what does it do when it finds an error? Does it go back and try to re-write that block before moving on to the next one?

What if the section of the tape where that block was written was physically bad, and no amount of overwriting fixes ths? How does the system indicate that this block should be ignored when read?

More details, context, and pictures at:

http://bit.ly/1NxeSM3

Thanks!
-AJ

 

[Chuck(G)]

My impression is that when a RAW error occurs, the drive backspaces to the IRG and erases a length of tape, then attempts the write again. What you wind up with is a big IRG. Similarly, you can rewrite blocks if a sufficiently large IRG is present between blocks.

(I assume you're talking about 7- or 9-track open-reel 1/2" tape.)

 

[firebirdta84]

Hi, Chuck(G),

I was hoping you would weigh in here, thank you.

What do you mean by "IRG"? I'm guessing you mean "Inter Record Gap"? If so, it must be the same as Inter Block Gap in this context. Although, I would love to understand the difference between a record and a block in this context.

And I'm talking about the strange Kennedy 6450/6455 QIC tape drive format that I link here: http://bit.ly/1NxeSM3

It's actually a 4-track QIC model device, but I am hoping that they designed it it on some precedent set in the past on the open-reel drives from earlier.

 

[Chuck(G)]

Mostly a difference in culture (CDC vs. the world). The old CDC name is IRG (I'm old); CDC differentiated between PRUs (physical record units) and LRUs ("logical records"; i.e. records before blocking.). Besides, I can pronounce "irg", but not "ibg".

I suspect the algorithm in the Kennedy QIC is pretty much the same as 9 track, except that the Kennedy actually tries to fill the gap with something other than a DC erase, since QIC blocks generally have ID bursts pre- and post-pended to data blocks (After all, you have to be able to read backwards). I'd have to check, however. On 9 track drives, though, I've seen a bad tape that's mostly erased gaps. The drive looks like it's really moving along, but in fact, very little is actually written.

Sometimes, this business with ID bursts can actually come in handy. Most 9-track drives allow changing density only at BOT. I ran across a tape that had a file written at 6250, followed by most of the rest of the reel written at 1600. I finally coaxed a drive into treating the 6250 part as erased gap and recovered the entire 1600 part.

 

[firebirdta84]

Thanks, Chuck(G), I appreciate all of this insight.

So, given what I have here, if you had to guess, could we say that, if a block exists, it was deemed to be valid at the time of read-after-write? The alternative is if the block was deemed bad at that time, it would have been erased, and therefore, not exist today.

Do you think my logic is correct here?

 

[Chuck(G)]

If the drive has read-after write heads, i.e. a head assembly that employs a read head that checks what the write head is doing, then yes. While this was pretty much the minimum for half-inch drives, the bulk of QIC drives were not constructed that way. I remember the annoyance of having to run a verify on a backup after writing it. Wasted time.

I don' know if this is the case with your Kennedy QIC unit, but if it's not explicitly mentioned, I'd assume that there was no immediate read-after-write checking.

 

[firebirdta84]

Ah, good points. It is explicitly mentioned, several times:

In the KENNEDY 6450 Cartridge Tape Drive Manual, page 4
"Recording Head: Serpentine, Read-After-Write, with Selective Erase"

And, at the bottom of page 2-30 of the Kennedy 6455 Cartridge Tape Maintenance manual, we see:

2.3.3.3 The Data Transfer Group
....skipping to ....
b. Write (Code - 01101)
Drive writes one record with standard IBGs. Performs read-after-write check.
Write commands from LLP will start 3 inches from LLP (Logical Load Point)

-------------------------------------------

OK, this helps a lot. If the block exists today, then I can have some level of confidence that 35 years ago, when it was written, that it was deemed valid by the read-after-write at the time of writing, and any error I find today is worth the effort to attempt to correct.

Oh, an for what it's worth, it seems that at least the QIC drives that I'm working with, do, in fact, have the read head placed just after the write head, all in the same head assembly. This includes the Archive 5945 and the Wangtek 5099 (that you recommended for me, actually...thanks!).

Thanks!

-AJ
http://MicrotechM1.blogspot.com
http://MightyFrame.com