| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 877.1 | | NETRIX::thomas | Moderator from Mars | Fri Feb 26 1993 10:10 | 1 |
| Fast well-designed card + very good protocol implementation + lots of CPU
|
| 877.2 | Not just the DEFTA | JEDI::CAUDILL | Kelly - NaC Tech Support - 264-3320 | Sat Feb 27 1993 15:48 | 12 |
| I haven't seen any of them, but there have been talks given about the
design of the DEMFA, DEMNA, and DEFTA and probably other cards. The
point being that these cards are are designed to be able to go full
speed. As compared to the DEUNA and others which could not go full
speed.
The number you quote - 96Mbps - is from OSF/1 AXP. Those are Matt's
other two points: lots of CPU and good protocols. OSF/1 has
implemented the Window Scaling option of TCP. This allows larger
windows and therefore less time and packets wasted on acknowlegements.
The work to do this went into OSF/1 last fall. Rumor has it that UCX
has similar features and therefore performance these days.
|
| 877.3 | | KONING::KONING | Paul Koning, A-13683 | Mon Mar 01 1993 11:13 | 7 |
| Not really "good protocols" -- but rather "good protocol implementations".
There's nothing special about TCP that makes it so fast. If anything, with
TCP it requires extra work (window scaling extension). You can get the same
performance with other transports if you use a comparably high quality
implementation. (Well, perhaps not with SNA... :-) )
paul
|
| 877.4 | DEFTA Documentation | WARNUT::FORSHAWJ | Jon Forshaw | Fri Apr 15 1994 05:35 | 13 |
| I have a customer using a DEFTA-FA in an Alpha AXP 3000 Model 600 and
he has asked for some technical documentation regarding the card, and
how to exploit it to its maximum potential, including tuning the
system.
Can you point me in the direction of any documentation that describes
this? I suggested to him that there might be nothing he can do to the
card, which he agreed might be so, but said he would still like to
learn more about what it is doing, even if he can't change it.
Regards,
Jon.
|
| 877.5 | Give him the whitepapers | SCHOOL::WASHABAUGH | Born to be Mild | Tue Apr 19 1994 12:04 | 11 |
| In the whitepapers directory on ftp.digital.com there are two pertenent
whitepapers, fddi-nics and something on TCP-IP and FDDI.
The fddi-nics.ps file is slanted towards the EISA NIC, and the TCP-IP/FDDI
whitepaper is slanted towards IP and UDP on OSF/1.
There is nothing that can be set on the card (including burst size) that will
improve the performance. However, there are possibly things in the configuration
of the system that can make a difference.
doug
|
| 877.6 | Technical Specification | WARNUT::FORSHAWJ | Jon Forshaw | Wed Apr 20 1994 07:46 | 11 |
| Thanks for the information. I will pass the whitepapers on. I have
told the customer that nothing can be altered on the card and
mentioned the performance talked about in this topic. He has asked
for simply some technical documentation/description/specification type
information regarding the card. Is there any such information that
someone could point me to?
Regards,
Jon.
|
| 877.7 | there may be a few more papers | MUDDY::WATERS | | Wed Apr 20 1994 09:13 | 10 |
| Digital should make the DEFTA/DEFEA/DEFpcimumble programmer's spec
available to customers, to answer questions like this.
But in reality, papers of a more tutorial nature may help the customer
more than detailed specs. I suppose there's no article on
DEFTA/DEFEA/DEFpcimumble in the Digital Technical Journal; there is an
article for DEMFA, but that's altogether different.
Try sending mail to K. K. Ramakrishnan, erlang::Rama, who has co-authored
some papers on FDDI adapters. He would know where to find public papers
that describe DEFTA. Please post here what papers you find.
|