Maximum Size of Hard Disk Drives for PDP-11 Software
Jerome H. Fine
systems used with PDP-11 hardware and or emulators such
as Ersatz-11 or SIMH, namely RT-11, RSX-11, RSTS/E
What is the largest hard disk drive allowed? By way of some
answers, I have found (and was rather shocked) that under
DOS and FAT16, the maximum size of a FILE is 2 GigaBytes
(less one byte I am told). With Windows 95/Windows 98 SE,
and FAT16, the maximum size is the same and with FAT32
the maximum size of a file is 4 GigaBytes (less one byte).
With the demo/hobby version of Ersatz-11, the maximum
size of a file used as a hard disk image is 32 MBytes. Thus,
that ends the question right there - the limit is arbitrary!
With the Full (Commercial) version of Ersatz-11, I attempted
to create a VERY large file and used PUTR. Very quickly,
I found that the maximum size of a file in PUTR is almost
2 GigaBytes, more specifically 2,147,450,880 bytes. This
is 32768 bytes less than 2 GigaBytes or 64 blocks less than
4,194,304 blocks which works out to 4,194,240 blocks.
It might have been possible to get a bit closer, but this size
was sufficient for the test. Note that I was using Windows 98 SE
with a FAT32 and a 38 GigaByte hard drive with about
30 GigaBytes of empty space. So I presume that it would
have been possible to create at least a 2 GigaByte file (or
exactly 2,147,483,648 bytes) with other software, but that
was not a problem. I promptly created five files called:
BIG1.DSK, Big2.DSK, BIG3.DSK, BIG4.DSK and
BIG5.DSK or almost 10 GigaBytes. Ersatz-11 then allows
MOUNT DU0: BIG1+BIG2+BIG3+BIG4+BIG5
plus I mounted a "small" RL02 image with all of the software
I needed, including DUSTAT - THANK YOU Tim Shoppa -
and proceeded to boot RT-11.
As hoped for, DUSTAT reported that DU0: was 316 RT-11
full partitions in size with the last partition being 320 blocks
smaller than a fill partition or only 65216 blocks. Since RT-11
can't specify (well currently RT-11 can't do that - but if someone
asks me to enhance RT-11 so that it can be done) a partition
number greater than 255 decimal, there was not much point
in testing further under Ersatz-11 with files (or even a group
of files) that are disk images.
Since I don't have a SCSI hard disk drive larger than 2 GigaBytes,
I am unable to do a test with Ersatz-11 with a raw SCSI device -
or at least to test the maximum size for a raw SCSI device. I am
quite confident that the Full version of Ersatz-11 can handle raw
SCSI hard drives of at least 8 GigaBytes. However, does anyone
reading this have a larger SCSI hard disk drive which they have
MOUNT DU0: SCSI0:
successfully under Ersatz-11?
As for real PDP-11 hardware, I understand that Tim Shoppa has
an 8 GigaByte hard drive - (I tend to assume SCSI) - which was
used under RT-11 with all 256 RT-11 partitions. This I would
assume that MSCP host adapters and controllers can handle
a drive that large. Has anyone used a larger drive on a real
PDP-11 system?
Finally, does anyone know of the maximum drive size as far as
the operating systems are concerned? For RT-11, there seems
no reason why a 2 TerraByte (65536 RT-11 partitions) should
not be possible. Does anyone know the maximum drive size
that RSX-11 currently handles?
Sincerely yours,
Jerome Fine
--
To obtain the original e-mail address, please remove
the ten characters which immediately follow the 'at'.
If you attempted to send a reply and the original e-mail
address has been discontinued due a high volume of junk
e-mail, then the semi-permanent e-mail address can be
obtained by replacing the four characters preceding the
'at' with the four digits of the current year.
Bob Supnik
'fseek' call. The largest disk current simulated is the RA92 at
1.5GB, for the VAX and PDP11.
/Bob
Tim Shoppa
> >plus I mounted a "small" RL02 image with all of the software
> >I needed, including DUSTAT - THANK YOU Tim Shoppa -
You're welcome :-). I don't have nearly enough time to play with
RT-11 anymore, but I thought DUSTAT was a pretty neat little tool.
John Wilson has something similar (but not identical) for querying
MSCP controllers.
> >with the last partition being 320 blocks
> >smaller than a fill partition or only 65216 blocks. Since RT-11
> ...
> >As for real PDP-11 hardware, I understand that Tim Shoppa has
> >an 8 GigaByte hard drive - (I tend to assume SCSI)
Actually, a 9Gbyte drive, but only the first 8 Gbytes are addressable
with "normal" (256 partition) RT-11 DU driver access.
Note that later versions of RT-11 have a DU driver that support .SPFUN's
out to terabytes, but not for filesystem type access, only through .SPFUN's.
> > - which was
> >used under RT-11 with all 256 RT-11 partitions. This I would
> >assume that MSCP host adapters and controllers can handle
> >a drive that large. Has anyone used a larger drive on a real
> >PDP-11 system?
All of the 9Gbyte drive is accessible under VMS using the same Q-bus
controller.
(Incidentally, the drive was a Seagate ST410800N. These drives now have
a street price on Ebay of less than $10... I remember just a few years
ago when it was closer to $5000!)
> >Finally, does anyone know of the maximum drive size as far as
> >the operating systems are concerned? For RT-11, there seems
> >no reason why a 2 TerraByte (65536 RT-11 partitions) should
> >not be possible.
Possible, but completely unmanageable :-)
> >Does anyone know the maximum drive size
> >that RSX-11 currently handles?
There was a patch issued by Mentec a couple years ago to fix a problem
with drives larger than 8 Gbytes under 11M+.
Tim.
Bob Schor
"Jerome H. Fine" wrote:
> I would appreciate some help with all 3 of the major operating
> systems used with PDP-11 hardware and or emulators such
> as Ersatz-11 or SIMH, namely RT-11, RSX-11, RSTS/E
>
> What is the largest hard disk drive allowed? By way of some
> answers, I have found (and was rather shocked) that under
> DOS and FAT16, the maximum size of a FILE is 2 GigaBytes
> (less one byte I am told). With Windows 95/Windows 98 SE,
> and FAT16, the maximum size is the same and with FAT32
> the maximum size of a file is 4 GigaBytes (less one byte).
Jerry,
In the PC (Fat16, Fat32), the file size is determined by the number
of clusters. Cluster size is a power of 2, with the smallest being (I
think) 4096. For DOS and Win9x, the largest cluster is 32,768; NT will
allow 65,536-byte clusters (I presume they say the cluster size is 0 =
2**16 mod 2**16, and "understand" this). In Fat16, clusters are
addressed using a 16-bit word, with a few entries (including -1)
reserved as special flags. If we assume only one such flag (my fuzzy
memory says this is wrong, but assume it for now), then you can address
2**16-1 clusters of 2**15, for a maximum disk size of 2,147,250,880.
What Fat32 did was to use four bytes/cluster to allow you to address (in
principle) 2**32 clusters (I think I read that implementations use only
3 bytes), and the default cluster size has been reduced, to 4096 I
believe. With these "small" clusters, you'd have a max size of about
18Tb (assuming all 4 bytes used, whereas if only 3 bytes are used, the
max size with small clusters is only(!) 68Gb, easily reachable today).
Bob Schor
Jerome H. Fine
> > >plus I mounted a "small" RL02 image with all of the software
> > >I needed, including DUSTAT - THANK YOU Tim Shoppa -
> You're welcome :-). I don't have nearly enough time to play with
> RT-11 anymore, but I thought DUSTAT was a pretty neat little tool.
Jerome Fine replies:
You don't take enough credit.
> John Wilson has something similar (but not identical) for querying
> MSCP controllers.
I seem to be confused. DUSTAT is for MSCP controllers
under RT-11 as far as I understand. Since I not aware of the
one by John Wilson, I just wanted to be sure everyone else
understood that both programs are for the MSCP interface.
As for the one by John Wilson, are you referring to the E11 command:
SHOW DU0:
which provides some of that information directly under E11 or
a program which runs under RT-11 in the same manner as
DUSTAT? And is John's version available on his web site?
If so, any idea where?
> > >As for real PDP-11 hardware, I understand that Tim Shoppa has
> > >an 8 GigaByte hard drive - (I tend to assume SCSI)
> Actually, a 9Gbyte drive, but only the first 8 Gbytes are addressable
> with "normal" (256 partition) RT-11 DU driver access.
I tend to change the word "normal" to be "DEC distributed"
since there seems to be no reason at all for restricting RT-11
to just 256 partitions - well, "no longer" any reason. When
V5.03 of RT-11 was released in 1985, I can't remember any
SCSI drives at all being available. And even in 1992 with the
release of V5.06 of RT-11, I can't seem to remember that
a drive as large as 8 GBytes was available.
> Note that later versions of RT-11 have a DU driver that support .SPFUN's
> out to terabytes, but not for filesystem type access, only through .SPFUN's.
Which seems to be confirmed via DUSTAT when I used
E11 to determine the number of blocks in an image
file to be almost 310 RT-11 partitions.
> > > - which was
> > >used under RT-11 with all 256 RT-11 partitions. This I would
> > >assume that MSCP host adapters and controllers can handle
> > >a drive that large. Has anyone used a larger drive on a real
> > >PDP-11 system?
> All of the 9Gbyte drive is accessible under VMS using the same Q-bus
> controller.
> (Incidentally, the drive was a Seagate ST410800N. These drives now have
> a street price on Ebay of less than $10... I remember just a few years
> ago when it was closer to $5000!)
Also, the other 3 1/2" drives were even more available. For
smaller drives like the ST32550N models, I picked up quite a
few to use with RT-11. These are only 1/2" high with a
50 pin header. I probably have a few green container boxes
around - the drives were from an Alpha RAID.
> > >Finally, does anyone know of the maximum drive size as far as
> > >the operating systems are concerned? For RT-11, there seems
> > >no reason why a 2 TerraByte (65536 RT-11 partitions) should
> > >not be possible.
> Possible, but completely unmanageable :-)
Hey, I can't see how you can keep track of even 256 RT-11
partitions. But, I did modify DIR.SAV so that I can:
DIRECTORY DU0:/VOLUME[[[:iiiiii]:jjjjjj]...:rrrrrr]
where iiiiii = Starting Block Number of a DSK file
jjjjjj = Starting Block Number of a nested DSK file
rrrrrr = Starting Block Number of innermost nested DSK file
or
DIRECTORY DU0:/VOLUME:tttttt
where tttttt = iiiiii + jjjjjj + rrrrrr
This allows me to look at directories of [nested] DSK files
without needing to MOUNT them as a logical - plus if
there is no extended device support with a nesting that is
beyond 8 for the distributed monitor, I can still look at
the directory for a nested DSK file.
Thus, if someone has all 256 RT-11 partitions and makes
a copy of just all the primary directories (68 blocks or less)
to a drive partition which is always available, then it is
very easy to determine exactly where a file is located.
If that MSCP device driver is NOT resident (i.e. is
NOT supporting the SY: device), then it is possible to
bring the associated RT-11 partition online with the
UNLOAD, SET and LOAD commands. Otherwise,
RT-11 only needs to be booted after a SET command.
As for using an MSCP drive more than 256 RT-11 partitions,
I once looked at the MSCP code in RT-11. It would seem
to be rather straightforward to modify the use of the mapping
table in RT-11 so that the physical UNIT number word (which
is restricted to a maximum of 255 in any case) occupies the
byte currently used by the partition number byte and visa versa.
That would permit an RT-11 partition number of 65535.
While some of the code in the DU(X).SYS device driver
would need to be changed along with the ability of the
SHOW DEVICE:DU
command (I would put that code into DUX.SYS in any case -
SET DU SHOW
the required changes don't seem all that difficult). MUCH MORE
important would be to find a couple of bits in the queue element
so that 256 RT-11 partitions could at the same time.
Does anyone remember the queue element layout well enough
to determine if one or even better 2 more bits could be found
so that 128 or better still 256 RT-11 partitions could be used
at the same time directly under RT-11? If so, I am sure the
problem of Rad50 names for that many partitions could easily
be resolved as well.
> > >Does anyone know the maximum drive size
> > >that RSX-11 currently handles?
> There was a patch issued by Mentec a couple years ago to fix a problem
> with drives larger than 8 Gbytes under 11M+.
Interesting. Does the patch require any particular version
of RSX-11M+ or can the patch be applied to most versions?
Is the patch available anywhere or must Mentec provide it?
Zane H. Healy
> Under SIMH, maximum disk size is 2GB, limited by the semantics of the
> 'fseek' call. The largest disk current simulated is the RA92 at
> 1.5GB, for the VAX and PDP11.
I find a RA82 to be the preferable disk size for PDP-11 emulation (VAX would
be the same if I was useing it). A RA82 image at 608MB fits nicely on a
CD-R, which means that you can then easily access the data on a real PDP-11.
On real PDP-11's I'd been limiting myself to 100-200MB SCSI drives until
this year, in part as I'd gotten a nice stash of drives that size that had
come out of old IBM PC's. Now I'm starting to use 2GB 7200RPM drives as
they now seem to be considered 'trash' and hold more data than I need them
to. I really can't see the use for 4-9GB drives on a Hobbyist system, for
one thing that's the size drives I'm currently using on my VMS server!
Zane
Douglas A. Gwyn
> (Incidentally, the drive was a Seagate ST410800N.
> These drives now have a street price on Ebay of less
> than $10... I remember just a few years
> ago when it was closer to $5000!)
Really? I got one soon after it came out, for my
homebrew PC, and was apparently the first to try
installing Solaris x86 on a partition of it; the
result was a drive that couldn't even be re-fdisked
using a DOS boot floppy (DOS would hang on booting
and never reach the command input). I finally
fixed it by carefully timing unplugging and
plugging the SCSI cable. With Sun's help we found
a way to get Solaris installed, but it took a whole
day on the phone. Anyway, I don't think it cost me
$5000, probably less than $2000.
Tim Shoppa
List price of the ST410800N at introduction (early 94)
was $3450, but due to high demand street price was
up to $1000 higher :-). Then add on some spare change
for a SCSI enclosure and it gets close to $5000. By
1996 or so you could get them for $2000.
I see that on E-bay last night several went unsold for
$2.99 - a good lesson for those who view computers as an
appreciating investment :-)
Tim.
Zane H. Healy
> List price of the ST410800N at introduction (early 94)
> was $3450, but due to high demand street price was
> up to $1000 higher :-). Then add on some spare change
> for a SCSI enclosure and it gets close to $5000. By
> 1996 or so you could get them for $2000.
> I see that on E-bay last night several went unsold for
> $2.99 - a good lesson for those who view computers as an
> appreciating investment :-)
I can't imagine why people are having a hard time selling them :^) After
all, who in this day and age wants a 5.25" full height SCSI drive (unless
it's totally free). The only thing I'm still using that size drive for is
my PDP-11/44 (but then those are physically small compared to the RL02's).
I like nice small, low power, quiet drives for my PDP-11/73! Switching the
/73 to small SCSI drives from 5.25" full height MFM and ESDI drives really
helped quiet it down!
Still, I remember how happy I was to get a pair of 2GB 5.25" full height
SCSI drives, for free, for my PowerMac 8500/180 back in '97/98 despite the
cost of the SCSI enclosure and the software I needed to format them.
Zane