Re: TWG questions


Rick Watson (rick@ngp.utexas.edu)
23 Jul 87 20:03:32 GMT


> Dbridge in Version 2 of the code is abysmal. I don't know if it is
> fixed in 3.0 since I now have a real IP path between the two sites.
> It would be nice if Woolongong would take the effort of reversing this
> product (allowing DECNET to flow through their IP paths) but it probalby
> isn't an easy modification.
>
> -Ron

*Flame on*
Saying that "Dbridge [...] is abysmal" is not really very informative.
What specifically did you have problems with?
*Flame off*

We have been using DBRIDGE under v2.x (and now 3.0) with good performance
under light to moderate usage (say 1-4 "active [currently transferring data]"
circuits in/out of each node). Performance will (of course) be dependent
on DECnet circuit speed and loading. Our topology looks like:

192.16.73---ether --+--128.83-----ether-+---------
       | |... | | |...
     +---+ +---+ +---+
     | | | | UTADNX | |--- 10 (arpanet)
     +---+ +---+ +---+
       | |
     ---DECnet 192.16.72-----
       | | | |
     +---+ +---+ +---+ +---+
     | | | | | | | |
     +---+ +---+ +---+ +---+

with DBRIDGE/DECnet making the connections for the network 192.16.72
over serveral leased phone lines that run at 56KB, 230.4KB and ~760KB.
The CPU overhead for the DBRIDGE process on UTADNX can get to be high
if it is having to route too many packets on to 128.83 and beyond. It
really needs to be a dedicated uVax. It would be nice if the DBRIDGE
interface could live in the WIN/TCP kernel (like ELINK does now),
instead of a user-mode process, but I have no idea if that is possible.

Rick Watson
University of Texas Computation Center
 arpa: ccaw001@utadnx.cc.utexas.edu rick@ngp.utexas.edu
 uucp: ...seismo!ut-sally!ut-ngp!rick
 bitnet: bitnet: ccaw001@utadnx
 phone: 512/471-3241



This archive was generated by hypermail 2.0b3 on Thu Mar 09 2000 - 14:38:47 GMT