Searching \ for '[EE]: network technologies (was "multiple serial d' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/io/serials.htm?key=serial
Search entire site for: 'network technologies (was "multiple serial d'.

Exact match. Not showing close matches.
PICList Thread
'[EE]: network technologies (was "multiple serial d'
2000\10\03@121128 by Bill Westfield

face picon face
   If you look at collision detection network bandwidth charts, they go
   exponentially down when you get over 40-50% utilization.

Theoretically true (so the scientists say) for gausian distributions of
packets sizes (which do not occur in real life), but long since disproven
in almost all practical cases (Boggs, Jacobson, etc.)


   We switched from ARCNET to Ethernet on our DEC system, and during busy
   times of the day, network speeds were terrible. At least with ARCNET's
   polled (token ring) protocol, bandwidth was predictable.

I think there was something else wrong with your ethernet.  In the "early
days", ethernets were plagued by a number of performance-destroying
problems; primarilly "broadcast storms"; but they weren't due to the nature
of ethernet itself.  In a way, it's sort of sad to see the token ring
protocols (4Mb, 16Mb, FDDI) all dying out as ethernet proves to be a far
cheaper solution.  (Of course, these days noone in their right mind runs
the 1000 node ethernets that were originally spec'ed, either...)

BillW
cisco

--
http://www.piclist.com hint: To leave the PICList
spam_OUTpiclist-unsubscribe-requestTakeThisOuTspammitvma.mit.edu


2000\10\03@123538 by Sean H. Breheny

face picon face
At 09:09 AM 10/3/00 -0700, William Chops Westfield wrote:
>     If you look at collision detection network bandwidth charts, they go
>     exponentially down when you get over 40-50% utilization.
>
>Theoretically true (so the scientists say) for gausian distributions of
>packets sizes (which do not occur in real life), but long since disproven
>in almost all practical cases (Boggs, Jacobson, etc.)

Aren't many theoreticians now looking at alternative distributions for
packet sizes? In my probability class, there were several example problems
using Paredo distributions, and the prof. said that Paredo is an old
distribution which didn't find much use until very recently when it was
found to be a better approximation (than the classical ones) to
distributions typically found in networks.

Sean

--
http://www.piclist.com hint: To leave the PICList
.....piclist-unsubscribe-requestKILLspamspam@spam@mitvma.mit.edu


More... (looser matching)
- Last day of these posts
- In 2000 , 2001 only
- Today
- New search...