lws@comm.wang.com (Lyle Seaman) (12/21/90)
I was reading the documentation on Intel's new Ethernet chip, the 82596. It says that the chip supports a mode in which a splitting algorithm is used to resolve collisions instead of simple random wait (yeah, with binary exponential backoff...). My question is: Is anybody using this? Does anybody out there think that this might be fun to do? I guess that it would provide minimal performance improvements for standard LANs, but it might help a lot for LANs which cover large distances. I was intrigued to note that a vendor had actually implemented it, I thought it was sort of an idle academic speculation. -- Lyle Wang lws@capybara.comm.wang.com 508 967 2322 Lowell, MA, USA Source code: the _ultimate_ documentation.
jwb@monu6.cc.monash.edu.au (Jim Breen) (01/04/91)
In article <1990Dec20.183227.19729@comm.wang.com>, lws@comm.wang.com (Lyle Seaman) writes: > I was reading the documentation on Intel's new Ethernet > chip, the 82596. It says that the chip supports a mode > in which a splitting algorithm is used to resolve collisions > instead of simple random wait (yeah, with binary exponential > backoff...). My question is: > > Is anybody using this? Does anybody out there think that > this might be fun to do? I guess that it would provide > minimal performance improvements for standard LANs, but > it might help a lot for LANs which cover large distances. > > I was intrigued to note that a vendor had actually implemented > it, I thought it was sort of an idle academic speculation. I have read one article on a splitting technique, which was being called "802.3D" (D for deterministic). According to the article, it had been developed for/by/in conjunction with/etc the French Navy, which wanted to use ethernet for onboard communications but was not happy about the non-deterministic nature of ethernet's response times. Perhaps this explains this mode in the 82596. (The mind boggles at the thought of an Exocet coming over the horizon while the packets from fire-control to the Phalanxes are going collide, backoff, collide,.....) As I recall, 802.4 also has a splitting technique to sort out collisions arising from >1 responses to a "solicit successor" frame. -- Jim Breen AARNet:jwb@monu6.cc.monash.edu.au Department of Robotics & Digital Technology. Monash University. PO Box 197 Caulfield East VIC 3145 Australia (ph) +61 3 573 2552 (fax) +61 3 573 2745 JIS:$B%8%`!!%V%j!<%s(J