CIDRizing the Internet

yakov at watson.ibm.com yakov at watson.ibm.com
Thu Apr 21 17:33:36 UTC 1994


fyi
Yakov.
******  The following is a COPY  *****************************
Received: from merit.edu by watson.ibm.com (IBM VM SMTP V2R3) with TCP;
   Thu, 21 Apr 94 12:17:01 EDT
Received: from localhost (jyy at localhost) by merit.edu (8.6.8.1/merit-1.0) with SMTP id MAA14094; Thu, 21 Apr 1994 12:13:50 -0400
Message-Id: <199404211613.MAA14094 at merit.edu>
To: bgpd at merit.edu
cc: jyy at merit.edu
Subject: CIDR, proxy-aggr or Die?
Date: Thu, 21 Apr 1994 12:13:49 -0400
From: Jessica Yu <jyy at merit.edu>

Hi,

Below is the NSF/ANSNET routing table size growth history data which Merit has been
collecting.  It shows that, on average, every two weeks there are ~500 more routes
added to the routing table between 4/93 - 3/94 and the average growth is 4%.  Note,
this is prior to people start to withdraw more specific routes.  It also shows
that during the first half of month 4/94, the routing table increased by 333 routes
instead of 857 during the previous two-week period.  This shows the effect of CIDR.

So what's the point?

1. Since we start to withdraw more specific routes around Apr. 1st.'94, it helped
   reduce the growth of the routing table by more than a half (4.1% vs 1.7%)

2. The routing table grows faster since the beginning of this year than last year,
   the average increase of two-week period since 1/94 - 3/94 is 776 routes or 4.57%.
   Everyone can do a calculation and figure out when his/her routers will run
   out of memory.

3. It still adds 333 more routes during the two-week period when we start to
   withdraw more specific routes.  We need to do more CIDR.  If we withdraw
   more or the same amount of routes added, we win.

   That requires us -

    When advertise new routes, advertise aggregates not aggregatable
    specific routes.
    Do more aggregation on existing specific routes and withdraw them.

   So again:

    ASs who do BGP4 already, please advertise aggregates and remove
    your more specific routes.

    ASs who do not do BGP4 yet, please either run BGP4 and do aggregation
    or delegate your neighbor ASs to proxy aggregate your routes.

It used to be:     CIDR, default or die!
It is now:    CIDR, proxy-aggr or die!

                             --jessica

P.S. The reason this set of data is shown in bi-weekly format is that we want to
see the effect of CIDR since 4/94.  Since 4/15, there are a lot more withdrawn
happened so it should be interesting to see the data at the end of this month.
(Enke:thanks for collecting the data).

# 2 week           # max r_tab   # 2wk growth(%) #2wk rts increase
# ===============================================================
93/04/15            7972         2.64            205
93/04/30            8239         3.35            267
93/05/15            8538         3.63            299
93/05/30            8961         4.95            423
93/06/15            9244         3.16            283
93/06/30            9534         3.14            290
93/07/15            9739         2.15            205
93/07/30           10113         3.84            374
93/08/15           10484         3.67            371
93/08/30           10879         3.77            395
93/09/15           11244         3.36            365
93/09/30           11621         3.35            377
93/10/15           12150         4.55            529
93/10/30           12703         4.55            553
93/11/15           13409         5.56            706
93/11/30           13886         3.56            477
93/12/15           14649         5.49            763
93/12/30           14802         1.04            153
94/01/15           15509         4.78            707
94/01/30           16281         4.98            772
94/02/15           16890         3.74            609
94/02/30           17713         4.87            823
94/03/15           18531         4.62            818
94/03/30           19388         4.62            857

              avg:     4.06%           496

94/04/15           19721         1.72            333






More information about the NANOG mailing list