Dear Managers,
Special thanks to
Dr. Tom Blinn
Daniel S. Riley
Paul Casteels
Their answers are at the end of this message.
To sum up, the algorithm is on firmware (I suspected of that) and is the BEB on our cards.
It was suggested to move to full-duplex or install an Ethernet switching. It was also suggested to change the Ethernet cards to some the with Intel or AMD LANCE chipsets.
Thank you.
Best regards
Antonio Urbano
--------------------------------
Tel.: +351 - 22 995 30 00 (ext. 1131)
Fax: +351 - 22 995 50 62
e-mail: a.urbano_at_mailcity.com
--------------------------------
**** MESSAGES ****
-------------------
-- Dr. Tom Blinn -- (2 messages)
-------------------
** 1st message
What leads you to believe that in a real Ethernet network interface this
is implemented anywhere above the level of the firmware embedded in the
adapter card itself? What leads you to believe that you can change it?
Tom
Dr. Thomas P. Blinn + UNIX Software Group + Compaq Computer Corporation
110 Spit Brook Road, MS ZKO3-2/W17 Nashua, New Hampshire 03062-2698
Technology Partnership Engineering Phone: (603) 884-0646
Internet: tpb_at_zk3.dec.com Compaq's Easynet: alpha::tpb
ACM Member: tpblinn_at_acm.org PC_at_Home: tom_at_felines.mv.net
Worry kills more people than work because more people worry than work.
Keep your stick on the ice. -- Steve Smith ("Red Green")
My favorite palindrome is: Satan, oscillate my metallic sonatas.
-- Phil Agre, pagre_at_alpha.oac.ucla.edu
Yesterday it worked / Today it is not working / UNIX is like that
-- apologies to Margaret Segall
Opinions expressed herein are my own, and do not necessarily represent
those of my employer or anyone else, living or dead, real or imagined.
** 2nd message
My colleague who knows this stuff quite well (he maintains the "tu"
driver and has written the drivers for other Ethernet interfaces for
Tru64 UNIX) comments:
------- Forwarded Message
I wonder where KK is these days...
Oddly enough, the tulip offered another backoff algorithm, but never
implemented it (it only supported the standard "blue-book" algorithm for
collision avoidance - Binary Exponential Backoff). That explained why,
when we enabled the capture effect, nothing different happened.
I don't know what the Intel Ethernet chipset supports in this regards, but
our concentration has been on full duplex connections instead of half
duplex (there are no collisions in full duplex).
As medias get faster, you don't want to deal with collisions.
In real life, what's really important is to ignore the IPG (Inter Packet
Gap) time while in half duplex. You want more time between packets (not
less). You'll actually see fewer collisions if you can't honor the exact
IPG time. In full duplex, of course, the IPG doesn't matter.
The tulip honors (precisely) the IPG, and will cause far more collisions
than the Intel or AMD LANCE chips.
------- End of Forwarded Message
So, it IS in the chip (that is, chip firmware) and you don't have much
control over it. What you get is what you get, unless you want to make
your own chip or write your own driver.
Tom
Dr. Thomas P. Blinn + UNIX Software Group + Compaq Computer Corporation
110 Spit Brook Road, MS ZKO3-2/W17 Nashua, New Hampshire 03062-2698
Technology Partnership Engineering Phone: (603) 884-0646
Internet: tpb_at_zk3.dec.com Compaq's Easynet: alpha::tpb
ACM Member: tpblinn_at_acm.org PC_at_Home: tom_at_felines.mv.net
Worry kills more people than work because more people worry than work.
Keep your stick on the ice. -- Steve Smith ("Red Green")
My favorite palindrome is: Satan, oscillate my metallic sonatas.
-- Phil Agre, pagre_at_alpha.oac.ucla.edu
Yesterday it worked / Today it is not working / UNIX is like that
-- apologies to Margaret Segall
Opinions expressed herein are my own, and do not necessarily represent
those of my employer or anyone else, living or dead, real or imagined.
---------------------
-- Daniel S. Riley --
---------------------
> From this we have several questions:
I'm not an ethernet guru, so prepend an "as far as I know" to all
of these answers.
> 1. What is the algorithm running on the systems, BEB or CABEB ?
Almost certainly BEB. I'm unaware of CABEB every being implemented in
hardware. Another solution, Binary Logarithmic Arbitration Method,
had an active 802.3 working group for a while (IEEE 802.3w) but was
abandoned due to lack of vendor interest. You can find an interesting
summary at [1], details at [2]. Bottom line is that vendors saw EC as
a minor problem, for which full duplex was an adequate solution in the
rare cases where it mattered.
> 3. How can I change the algorithm ? Is there any patch available ?
Arbitration is implemented in the interface hardware, so you'd need
a NIC with hardware support for CABEB or BLAM. AFAIK, no such beast
exists.
[1]
http://www.ece.wpi.edu/courses/ee535/hwk98/hwk3cd98/garrison/garrison.html
[2]
http://www.cs.ucr.edu/blam/report/tableofconts.html
--
Dan Riley dsr_at_mail.lns.cornell.edu
Wilson Lab, Cornell University <URL:http://www.lns.cornell.edu/~dsr/>
"History teaches us that days like this are best spent in bed"
-------------------
-- Paul Casteels --
-------------------
Hello,
The algorithm for collision resolution is implemented in the hardware of
the Ethernet controller. I think your only option is to check the docs of
your controller and replace it if it does not suit you.
Another option for avoiding collisions is of course installing an Ethernet
switch. They are not so expensive anymore.
Best regards,
Paul Casteels casteels_at_uia.ua.ac.be Tel: +32.3.8202455
Fax: +32.3.8202470
University of Antwerp Dpt.Physics
Universiteitsplein 1
B-2610 Wilrijk
Belgium
Send FREE Greetings for Father's Day--or any day!
Click here: http://www.whowhere.lycos.com/redirects/fathers_day.rdct
Received on Thu Jul 06 2000 - 09:03:17 NZST