joshua@pogo.gnu.ai.mit.edu (Joshua Geller) (03/28/91)
Archive-name: internet/talk/irc-backbone/1991-03-27
Archive: slopoke.mlb.semi.harris.com:/pub/incoming/irc.bb.* [132.158.82.36]
Original-posting-by: joshua@pogo.gnu.ai.mit.edu (Joshua Geller)
Original-subject: new irc backbone plan
Reposted-by: emv@msen.com (Edward Vielmetti, MSEN)
john and pjg and myself have come up with a new backbone plan: here it is.
Our basic philosophy has been to minimize active links between hubs (in other
words minimize the numberof 6667's that hubs have for each other), minimize
hops and delays and prevent doubling on nss links. so that when the net burps
the daemons reconfigure the net in one of a limited number of ways, so that NSF
links are used more efficiently.
noting that: classes will be used to favor preferred connections.
e.g.
The following irc backbone plan has been constructed using the following
constraints:
1) A desire to minimize the number of hubs and their possible interconnections.
2) The need to have irc hub traffic follow NSFnet interconnects.
3) The need to prevent multiple irc connections over the same nss link.
4) Making some effort to co-locate areas of large mutual traffic.
5) The need for simple reconnect choices to minimize manual squits.
The significant difference between this plan and the current one is the
minimization of active connections at hubs. The most common case is two
with some at three. In each case connections are along nss-nss links and
multiple nss hops are rare.
We have suggested a set of connection preferences as well as links but
naturally the "steady state" (or most stable configuration) will have
to be determined over time and the configurations are subject to adjustment
within the constraints provided above.
Some feedback please?
Documentation (including ps maps) are available on anonymous ftp at
slopoke.mlb.semi.harris.com /pub/incoming/(ircbb.ps, irc.bb.fig, ircbb.ascii,
irc.bb.netlist).
This NSF map is included for convenience.
NSFNET NSS connectivity 1991
NSS# Site connects to these NSS#'s
5 Pittsburgh PA 8 10 12 18
6 San Diego CA 11 13 14
7 Boulder CO 11 15 16
8 Princeton NJ 5 9 17 CA*net
9 College Park MD 8 10 11
10 Ithaca NY 5 9 17 CA*net CERN
11 Houston TX 6 7 9 18
12 Champaign IL 5 14 16
13 Palo Alto CA 6 14 15
14 Seattle WA 6 12 13 CA*net
15 Salt Lake City UT 7 13 17
16 Lincoln NE 7 12
17 Ann Arbor MI 8 10 15
18 Atlanta GA 5 11
CA*net CA*net
| | \
_Seattle________________________ | \
/ / \ ____Ithaca___\___...CERN
| | \ / | | \
| | ___________________\____Ann_Arbor__|___|__Princeton
| | / \ | | / /
| | / _Lincoln__ \ __Pittsburgh__|__/ /
| | / / \ \ / | | /
| | ___Salt_Lake____Boulder Champaign | College_Park
| | / | | /
| | / | ____________|_______/
| Palo_Alto | / |
| \ | / |
\_San_Diego | / Atlanta
\_____________________Huston_____________/
First name is hub other names are 6667 lines at that hub listed in priority.
Plussed items are nss hops to achieve balance.
algol helium bucsd
arrakis bucsd nyser
betwixt tavi sfsu
*boulder arrakis tavi
bucsd arrakis fairhope lyman algol
byron sfsu betwixt oddjob
+fairhope tavi bucsd
helium algol tavi lyman
lyman bucsd betwixt
nyser fairhope bucsd
oddjob arrakis fairhope
+tavi fairhope betwixt helium
sfsu betwixt byron
Servers at nss points are given as examples using existing hubs.
+-o-----------------------------------
| |se \ O================O
| | \ /|\ it |ca
| | \ / | ---O----------\|
| | \ / | | ------o
| | \ / | | / /|pr
| | bo --/----O----O---+ / |
| | *--------------------- ch |pi / |
| O \ | / |
| |pa \ | / |
| | \ -------=-----------+ / |
| | \ / / ---o
| | \ / o-- / |lyman
| | \ / / cp / |
| | \ / / / |
| | ---------------------------o----------------------o------ |
| |/ ho at |
+-O----------------------------------------------------------------+
sd
se - Seattle Byron
pa - Palo Alto SFSU
sd - San Diego Betwixt
ho - Houston Sesquinet
ch - Champaign Oddjob
it - Ithaca Nysernet
pi - Pittsburgh Fairhope
cp - College Park SURA
at - Atlanta SURA
pr - Princeton NEAR/JVNC
ca - Cambridge NEAR
pjg@acsu.buffalo.edu
john@mlb.semi.harris.com
joshua@gnu.ai.mit.edu
note: since I posted this to operlist this morning, I have gotten feedback
from jarkko, saying that he didn't know how well the connection classes
would work, and from helen re exploiting the T3's more. Thank you very
much! Comments are appreciated.
josh
*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*
@ Joshua Geller | joshua@gnu.ai.mit.edu | USMail: Seattle, WA 98102-1752 @
*-----------------------------------------------------------------------------*
@ Geller's Law: | "If I had a steel Q-tip, I would stuff it in @
* "Never postulate conspiracy | my ear, run 10,000 volts and channel god." *
@ when stupidity will explain |_______________________________________Spamgod @
* the known facts."|"This wallpaper is killing me. One of us has got to go." *
@<><><><><><><><><><><><><><><><><><><><><><><><><><>Oscar Wilde's last words @
*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*@*