[net.unix] 11/70 <-> uVax II link

jwp@uwmacc.UUCP (Jeffrey W Percival) (09/25/85)

We have a PDP 11/70 running 2.8BSD.  We are thinking of getting
a MicroVax II, and getting Berkeley Vax Unix to run on that.  Instead
of trying to get networking going on our 11/70, we thought we'd
get a Systems Industries disk and controller that will be shared
by the two machines.  Is this a reasonable approach?  Our
expectation is that file systems on the shared disk will be
accessible to users of both Unixes.  Are the 2.8/4.2 file systems
identical?  Can we just "mount" a shared file system on each of
the systems, so that the directories and files in that file
system appear in the directory paths of users on both machines?

I would appreciate any comments on this.  Thanks!

-- 
	Jeff Percival ...!uwvax!uwmacc!jwp

dan@rna.UUCP (Dan Ts'o) (09/27/85)

In article <1486@uwmacc.UUCP> jwp@uwmacc.UUCP (Jeffrey W Percival) writes:
>We have a PDP 11/70 running 2.8BSD.  We are thinking of getting
>a MicroVax II, and getting Berkeley Vax Unix to run on that.  Instead
>of trying to get networking going on our 11/70, we thought we'd
>get a Systems Industries disk and controller that will be shared
>by the two machines.  Is this a reasonable approach?  Our
>expectation is that file systems on the shared disk will be
>accessible to users of both Unixes.  Are the 2.8/4.2 file systems
>identical?  Can we just "mount" a shared file system on each of
>the systems, so that the directories and files in that file
>system appear in the directory paths of users on both machines?

	In general, not a very feasible idea.

	- Both 4.2BSD and Ultrix use a very different filesystem layout from
older UNIXs, including 2.8BSD. One cannot just mount a 2.8BSD filesystem on
a 4.2BSD UNIX and vice-versa. It might be reasonable, if you really wanted to,
to put another filesystem type within 4.2BSD to allow it to read the 2.8BSD
filesystem. Another alternative is to try to run 4.1BSD on the MicrovaxII,
which has a very similar filesystem structure to 2.8BSD. Even with these two
possibilities, one runs into another problem (at this level): the PDP-11 C
compiler stores longs differently than the VAX C compiler (gotcha!). This
can be fixed by altering the PDP-11 C compiler. I believe Donn Seeley when
through this entire exercise to allow his 2.8BSD PDP-11's to access his 4.1BSD
VAX filesystems and vice-versa. You may ask him if he thinks it is worth it.
Moving 4.1BSD to the MicrovaxII sounds pretty stupid itself. Perhaps you should
consider System V on the MicrovaxII (which DEC may well offer). Well, this is
ridiculous enough, but...

	- You are also screwed because even if they were two identical machines
or filesystems, you could not mount them both read-write (and get away with it).
As soon as one UNIX wrote the filesystem, modifying inodes and the freelist, the
other would get mighty confused. Indeed, even mounting only one read-write is
not a good idea, though probably not fatal. You could devise some kind of
locking scheme to allow multiple read-write access. A lot of work. Just
having both read only would be okay.

	- Since the MicrovaxII is a Qbus machine, you will need a Qbus
controller from System Industries that is format compatible with and can dual
access with your Unibus SI controller. Most Qbus SMD controllers do heavy
interleaving (3:1), more so than Unibus controllers, thus making the two
not format compatible. Perhaps SI does, I don't know.

	All in all, not very feasible in my mind. A more reasonble tack IS to
network the two machines, using either 2.9BSD or Ultrix-11 V3 (which seems to
work quite well), on your PDP-11/70.

					Cheers,
					Dan Ts'o
					Dept. Neurobiology
					Rockefeller Univ.
					1230 York Ave.
					NY, NY 10021
					212-570-7671
					...cmcl2!rna!dan
					rna!dan@cmcl2.arpa

pdg@ihdev.UUCP (P. D. Guthrie) (09/30/85)

In article <440@rna.UUCP> dan@rna.UUCP (Dan Ts'o) writes:
>In article <1486@uwmacc.UUCP> jwp@uwmacc.UUCP (Jeffrey W Percival) writes:
>>We have a PDP 11/70 running 2.8BSD.  We are thinking of getting
>>a MicroVax II, and getting Berkeley Vax Unix to run on that.  Instead
>>of trying to get networking going on our 11/70, we thought we'd
>>get a Systems Industries disk and controller that will be shared
>>by the two machines.  Is this a reasonable approach?  Our
>>expectation is that file systems on the shared disk will be
>>accessible to users of both Unixes.  Are the 2.8/4.2 file systems
>>identical?  Can we just "mount" a shared file system on each of
>>the systems, so that the directories and files in that file
>>system appear in the directory paths of users on both machines?
>
>
>	- You are also screwed because even if they were two identical machines
>or filesystems, you could not mount them both read-write (and get away with it).
>As soon as one UNIX wrote the filesystem, modifying inodes and the freelist, the
>other would get mighty confused. Indeed, even mounting only one read-write is
>not a good idea, though probably not fatal. You could devise some kind of
>locking scheme to allow multiple read-write access. A lot of work. Just
>having both read only would be okay.
>
>	- Since the MicrovaxII is a Qbus machine, you will need a Qbus
>controller from System Industries that is format compatible with and can dual
>access with your Unibus SI controller. Most Qbus SMD controllers do heavy
>interleaving (3:1), more so than Unibus controllers, thus making the two
>not format compatible. Perhaps SI does, I don't know.
>

If I remember correctly from some junk mail from SI, they offer a system
for two Micro-vaxes to share the same disk drive. They must provide some
sort of device driver mods for both Ultrix and Vmess, I would expect, so
the above may be possible. What I would be interested in is if it would
be possible to format one partition VMS, and one Unix, and run each OS
on one machine. You then write a device driver for each OS to access the
partition and format of the other machine, allowing for quick and easy 
Unix <=> Vmess conversion, and of course, the advantages of both OS's
for your data. Any opinions?

					Paul Guthrie.

scooter@genie.UUCP (Scooter Morris) (10/02/85)

> .... Our
> expectation is that file systems on the shared disk will be
> accessible to users of both Unixes.  Are the 2.8/4.2 file systems
> identical?  Can we just "mount" a shared file system on each of
> the systems, so that the directories and files in that file
> system appear in the directory paths of users on both machines?
> 
> -- 
> 	Jeff Percival ...!uwvax!uwmacc!jwp

No!!  The 2.8 and 4.2 filesystems are VERY different.  2.8 had 1K byte
blocks on all devices and other than that was a (sort of) vanilla v7
filesystem.  4.2 had different blocking factors on each partition and
uses clustering to improve performance.  The inodes aren't even the
same format.  Why not just bring up the Seismo/Harvard version of 2.9
and ethernetting your systems together??

						Scooter Morris
						Genentech, Inc.