[Simh] RT-11 Storage Strategy

[Clem Cole]

On Tue, Feb 16, 2016 at 11:33 PM, Will Senn <will.senn at gmail.com> wrote:

> * Is there a preferred disk controller/device?
>
​Hmm hard to say what prefered would mean.  I can tell you what history was
by what people tended to buy.  For the original Unibus PDP-11
implementations, most that had disks used the RK05s because that was what
was available. The RL series and RK06/07 come later [Note RK06/07 use a
different control - the RK611 and are different technologies in package,
recording etc than the RK03/04/5].​

A large number of people used 3rd party/after-market memory & storage,
particularly for RT-11 because of cost.  There was a thriving business in
those days of 3rd party controllers that emulated the DEC ones and could
talk to 3d party storage such as the CDC.

I do not remember the specifics, but I do remember that RL controllers had
issues with Unix.   I suspect if you look in the driver you may find
comments.

By the time of the QBUS/LSI-11 implementations things started to change.
DEC had more storage options, prices on disks had dropped dramatically and
IIRC SCSI was beginning to show up.  I had stopped doing anything with RT11
by then so I can not tell you what was typical.

That said, if you used RK05's you certainly would be building a simulated
Unibus system that was close to what many people had.




> * Is there a controller that supports more disk devices than another (RL
> vs RK, etc)?
>
​RK04/5/6 and RP04/5/6 controllers certainly support 8 units.
My memory is RL can only support 4 (check the bits in the Unix driver - it
should be pretty easy to see).   Dan Klein and I wrote the original RK611
driver for Unix by hacking on the RP06 driver.   My memory is that it
supported 8 units also, but we only had 2 on our system.




> * Does one device have more capacity than another (either via single disk
> raw capacity or via overall capacity of attached units)?
>
​Yes, RP06's were the largest until the SCSI drives show up.  But I did not
think RT-11 supported them because it's file system would have overflowed.
Unix did it by partitioning the physical disk into logical disks.   Since
Unix could had a uniform namespace, by mounting the logical disk it could
piut the back together​ so the only downside was whatever limit the OS had
for the largest file.  RT-11 could not do that, so if you partitioned the
drive it, you would have DSK C:, DSK D:, DSK E: ...




> * Is one device/controller more reliable in SimH than another?
>
​No idea.   ​




> * Do disks need to be formatted before initializing?
>
Formatted no, initialized yes.​

Formatting is a low level process that sets up the raw drive, regardless of
OS.   Its make the physical disk pack recognizable to the physical
controller (i.e. turning it in a set of blocks addressed with HD, CYL,
SEC).   This was usually performed with a standalone program that was part
of the disk controller diagnostics.


Initialization is putting an OS specific file structure on the drive so the
logical vector of disk sectors are interpreted as a file system.  In older
Unix this is mkfs command, in newer versions, newfs.




> * Are there some known best practice configurations (so many RL
> controller, with so many drives, or so many RL and so many RK, etc.)?
>

​A typical small system install, had 2, maybe 4 RK05 that were
dismountable.   DEC later created a version of the RK05 that was not
dismountable, but could get twice the storage density.  A lot of 11/34A's
were sold with the configuration.​
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.trailing-edge.com/pipermail/simh/attachments/20160217/8a3ae5f4/attachment.html>