BGP Economics ?

Mailman List Mirror (Read Only)
1

Dear Geoff;

   I just noticed the remarkable flattening in the recent growth of the BGP table
in your BGP graphs at :
http://www.telstra.net/ops/bgp/index.html

It didn't even seem this striking at Minneapolis.

Is this cessation of growth real ? Is this a sign of the recent economic slow-down,
or is there some more technical explanation (such as more aggressive aggregation at Telstra)?

                                   Regards
                                   Marshall Eubanks

   Multicast Technologies, Inc.
   10301 Democracy Lane, Suite 410
   Fairfax, Virginia 22030
   Phone : 703-293-9624 Fax : 703-293-9609
   e-mail : tme@on-the-i.com http://www.on-the-i.com

Test your network for multicast : http://www.multicasttech.com/mt/

2

   I just noticed the remarkable flattening in the recent growth of the BGP table

    > in your BGP graphs at :
    > http://www.telstra.net/ops/bgp/index.html
    > Is this cessation of growth real ? Is this a sign of the recent economic slow-down,

There's been quite a bit of discussion of this in the metrics
community. The flattening out appears to be largely in the /21-and-longer
area, such that folks who are filtering haven't seen as much of a
flattening as those who aren't. The link to the economic downturn is one
that Phil Smith suggested. That's a difficult causal relationship to
validate, though, regardless of how much sense it makes.

                                -Bill

3

Is this cessation of growth real ? Is this a sign of the recent economic slow-down,
or is there some more technical explanation (such as more aggressive aggregation at Telstra)?

If you look at the view from AS286, KPNQwest, (http://www.mcvax.org/~jhma/routing/bgp-hist.html)
(also I've just started up looking as AS286 remotely at http://www.telstra.net/ops/bgp/as286) you see nothing of this slow down that AS1221 sees. INdeed the AS286 view suggests that the underlying drivers are HIGHER than the sull FIB tale - compare the green and blue curves to the red curve on James Aldridge's AS286 graphs are you reach the conclusion that the table is growing _DESPITE_ recent reffects to improve aggregation modulo provider policies.

So whats going on? Inside AS1221 there is a fair number of local routes (about 22,000 of them). Over the past three months AS1221 been removing noise components from the external view of AS1221 (such as removing asymmetric satellite services using BGP routing), and the view on these web pages reflects the fact that AS1221 is seeing a total route size value (round 103,000) which, over the past three months converging on the KPNQwest value (99,000). Internally to AS1221 the number of routes remains at 120,000 (I've started looking at this internal noise level at http://www.telstra.net/ops/bgp/pc3)

My personal take on a bottom line: every view of the BGP table is relative, and changing local circumstances as well as changing global circumstances generate changes in the local perspective of the BGP table. Its sometimes a little harder to work out if the local changes reflect some global trend, or are just local changes. That's why multiple views in multiple locations help _a lot_ in working out the difference between global and local trends. In this case it really does appear to be just local issues.

So, if there is a belief that BGP table growth has slowed down in the first three months of this year due to social pressure, I do not support such a view even though the AS1221 data appears to indicate this. Its just local issues. The AS286 view supports the view that the underlying growth drivers are as strong as ever and the various efforts of nag mail of network operators has been largely (and predictably) ineffectual.

4

"There are three kinds of lies; lies, damned lies and
  statistics"
  
      - Benjami Disraeli
  
  Because I was curious, I merged 2, then 3 of these
  bgp graphs. This is done using a graphics tool and is not
  at all alleged to be accurate, but it should be illustrative.
  
    http://www.routingloop.com/share/2bgps.gif
    - geoff's and tbates'
  
    http://www.routingloop.com/share/3bgps.gif
    - geoff's, tbates', and jhma's
  
    http://www.routingloop.com/share/overlay.gif
    - same as 3bgps.gif with 2 extrapolated trends
  
  Conclusions drawn (none of them revelatory):

  Certainly the data is dissimilar, w/ geoff's generally
  higher, undoubtedly due to the reasoning below:

So whats going on? Inside AS1221 there is a fair number of local routes
(about 22,000 of them). Over the past three months AS1221 been removing
noise components from the external view of AS1221 (such as removing
asymmetric satellite services using BGP routing), and the view on these web
....

   The data from different sources is quite different. My experiences
  at a large ISP subsidiary of a large telco, upon considering
  merging w/ another large telco, caused me to do tremendous
  analysis of BGP information, such as comparing 'size' in
  several metrices, such as routes, traffic, address_space, etc..

  What I found (in this past life) was that correlating the data from
  different sources was particulary difficult, if not impossible,
  and that most all views into the global routing table were
  indeed different, as Geoff states below:

My personal take on a bottom line: every view of the BGP table is relative,
and changing local circumstances as well as changing global circumstances
generate changes in the local perspective of the BGP table. Its sometimes a

  It does seem that we've seen a bit of a slow down in routing table
   growth. Someone with more time should take a look at breaking down the
        implied curves over given periods. It sure looks like we were on a
  slow exponential curve from ~1998 to 3Q2000. 3Q2000->Now looks like
  more of a linear growth.

  I suspect if one asked these 3 folks for the tabular data, a wiz with
   something like mathematica could do some really nifty analysis.

  -alan

  legend -- cyan = ~jhma's work
     red = geoff's work
    grey = tbates' work
    purple = extrapolated trends

5

So, if there is a belief that BGP table growth has slowed down in the first three months of this year due to social pressure, I do not support such a view even though the AS1221 data appears to indicate this. Its just local issues. The AS286 view supports the view that the underlying growth drivers are as strong as ever and the various efforts of nag mail of network operators has been largely (and predictably) ineffectual.

I decided to track how well my emails are doing. I have contacted about 25 ASNs so far based on Tony's report over ther past 3 months:

AS Nets then Nets now Net gain/loss
AS1221 1652 1594 -58
AS701 1581 1469 -112
AS11371 324 74 -250
AS4151 277 242 -35
AS3549 432 149 -283
AS271 280 73 -207
AS7545 196 138 -58
AS9269 162 113 -49
AS8006 146 19 -127
AS6429 218 212 -6
AS6595 163 163 0
AS13999 109 87 -22
AS4293 384 372 -12
AS8013 330 565 +235
AS4755 213 204 -9
AS1942 136 55 -81
AS1727 176 14 -162
AS9498 87 80 -7
AS6499 170 39 -131
AS5106 101 101 0
AS11170 64 28 -36
AS16758 63 63 0
AS3464 153 123 -30
AS3749 120 121 +1
AS6413 67 67 0

Withdrawn nets: 1675, added nets: 236, net reduction: 1409 nets. Now I guess some of these would have reduced their routing announcements in any event. Is this a scientific study? No. Does it show that perhaps the routing table growth can be flattened via contacting ASN admins? I believe the answer is yes.

-Hank

PS I have contacted AS8013 (PSA Canada) on March 6 (swip@PSI.CA, noc@psi.ca). No response and only a large growth over the past month. If anyone has better contact info than what is listed in ARIN and RADB, please feel free to let me know.