nft copy from risc 4.2

nft copy from risc 4.2

Post by Michael A. Crowl » Tue, 04 Aug 1992 23:24:00



I'd like to know if someone can replicate the following problem
with Pathwork's DOS "nft copy ...." command.

It seems that using nft to copy a file from a risc machine running
ultrix 4.2 or 4.2a, the transfer speed is about 2KByte/sec.
Copying the same file from other machines is about 20KB/sec.
The other machines include: VMS, VAX-Ultrix 4.2, Risc Ultrix 4.0.

These results occur with 3Com microchannel boards and old DEPCA
boards.  The results are identical also using both 4.0 and 4.1 client
software.  

All other DECNet copys between machines do not show this slowdown; it
is only nft copy from the risc 4.2[a ] machines.  Nft copying *to*
those machines is not different from any other machine; it is only
copying *from* the risc machine to the PC.

Other kinds of copy from the risc 4.2[a ] machines are not a problem.
When used as file servers, using a DOS copy from those machines is
as fast as the other machines.  DECNet copy from a Macintosh client
does not differ significantly from machine to machine.  Nft copy
from the VAX-Ultrix machine with its disks NFS mounted from the
risc 4.2a machine is fast, while copying directly from the risc 4.2a
machine is slow.

This problem seems to be a specific interaction between nft copy commands
and risc 4.2 Ultrix.  The 4.2 machines have DEC patches installed
for a variety of things including the packet filter patch.  

I would appreciate knowing if anyone can replicate this problem, or
has run into it and solved it.


 
 
 

nft copy from risc 4.2

Post by Jim Dumoul » Sun, 09 Aug 1992 01:00:34



> I'd like to know if someone can replicate the following problem
> with Pathwork's DOS "nft copy ...." command.

> It seems that using nft to copy a file from a risc machine running
> ultrix 4.2 or 4.2a, the transfer speed is about 2KByte/sec.
> Copying the same file from other machines is about 20KB/sec.
> The other machines include: VMS, VAX-Ultrix 4.2, Risc Ultrix 4.0.

> These results occur with 3Com microchannel boards and old DEPCA
> boards.  The results are identical also using both 4.0 and 4.1 client
> software.  

> All other DECNet copys between machines do not show this slowdown; it
> is only nft copy from the risc 4.2[a ] machines.  Nft copying *to*
> those machines is not different from any other machine; it is only
> copying *from* the risc machine to the PC.

> Other kinds of copy from the risc 4.2[a ] machines are not a problem.
> When used as file servers, using a DOS copy from those machines is
> as fast as the other machines.  DECNet copy from a Macintosh client
> does not differ significantly from machine to machine.  Nft copy
> from the VAX-Ultrix machine with its disks NFS mounted from the
> risc 4.2a machine is fast, while copying directly from the risc 4.2a
> machine is slow.

> This problem seems to be a specific interaction between nft copy commands
> and risc 4.2 Ultrix.  The 4.2 machines have DEC patches installed
> for a variety of things including the packet filter patch.  

> I would appreciate knowing if anyone can replicate this problem, or
> has run into it and solved it.



   Check your pipeline quota and try setting it to a lower number.  Your
  loss of thruput is probably being caused by an excessive number of
  retransmits when the higher speed RISC machine overflows the memory
  on the PC's.  The pipeline quota (on the VAX) identifies the number
  of outstanding bytes you can "have in the pipe" before you check to
  see if you've sent too many.   If you're running the TCP/IP version
  of Pathworks you should also check the TCP/IP window size on the
  PC side.

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

   NASA / Payload Operations      SPAN/HEPnet: KSCP00::DUMOULIN
   Kennedy Space Center
   Florida, USA  32899

 
 
 

1. need Non-Interactive dump for RISC Ultrix 4.2

The Ultrix 4.2 dump seems to insist on asking inane
interactive questions like

  `` Is la-de-da disk mounted ("yes" or "no") ''

Have I got this right?  

If so, I understand why DEC thought exabyte drives weren't all
that advantageuous.

I'd be most grateful to know how to use the Ultrix dump
in batch mode, or if it really is braindead, for a
pointer to a non-interactive dump command like the
one on SGIs and Suns.  

Thanks in advance.
--
George Planansky
Department of Biological Chemistry & Molecular Pharmacology
Harvard University Medical School, 240 Longwood Ave., Boston, MA 02115

phone: (617) 432-3919
fax:   (617) 432-4360

2. RasDial() under plain vanilla Win95

3. Non-Interactive dump for RISC Ultrix 4.2 [it is]

4. A javascript preload problem

5. ATI Pro turbo and Premierre 4.2

6. Oasis bug #327965

7. ??? PL/1 & UNIX 4.2 bsd ??? - (nf)

8. Accidently deleted file

9. ACE+TAO 1.3 for SUNSpro 4.2 compiler

10. GNU-Emacs + Continuus/CM 4.2 Integration

11. StarTeam 4.2 SP3 losing of code

12. annc: GUIFFY 4.2 compare/merge update