Hey, this looks pretty good. Let's hope the trend continues (if, in fact, there *is* a trend).

...

Date Prefixes 231196 40687 241196 40908 251196 40844 261196 41043 271196 41074 281196 39803 291196 39735

I suspect that this is more due to the holiday than any real change. ;-(

Someone could do a paper on periodic influences on the number of global prefixes. I'd really love to know when triple-witching day is... ;-)

Tony

It may only be a trend as seen from the xara.net boxen. This is what I see from this neck of the woods: sandbox>sh ip bgp sum BGP table version is 311040, main routing table version 311040 41429 network entries (126264/162202 paths) using 9935800 bytes of memory 10181 BGP path attribute entries using 1129596 bytes of memory 0 BGP route-map cache entries using 0 bytes of memory 5355 BGP filter-list cache entries using 85680 bytes of memory Dampening enabled. 780 history paths, 63 dampened paths -- --bill

...

This is what I see from this neck of the woods:

which is not very useful with no historical data.

True enough. Perhaps tony would be willing to share his methods so the results may be comparable. I've got about two months worth of data off sandbax. -- --bill

This is not to pick on CERFNET but just to highlight a problem of actually tracking the size of the routing table in general. This CERFnet case seems to be this way becuase it is a direct peer of Bill's box even though I can see no reason why the more specifics are needed.

Looking a little more it seems a large amount of more specifics are being announced to Bill's boxe which aren't being announced to the xara.net router. Perhaps ISPs are taking more care at places like MAE-East with their outbound filters than they are at Bill's peering point even though Bill only has 6 active EBGP neighbors and the xara.net router has 39 ;-(.

Apparently true. The specifics leak from one of our CIDR blocks was only being made to Bill's peering point ( otherwise known as MAE-LA or MELEE when first started ;-) ). Being fixed shortly. Thanks On the larger question, it would be difficult to find two places on the net with a consistent view of the global routing table although for trend analysis both MAE East and the Sprint NAP should have a more of a representative view than MAE-LA. --pushpendra Pushpendra Mohta pushp@cerf.net +1 619 455 3908

Bill et al,

...

It may only be a trend as seen from the xara.net boxen.

is basically AGIS full table union Sprint full table union UUnet full table union

For reasons two complicated to go into I've got entirely different full routing on a box in London, and yes, it's a different size (smaller actually last time I looked).

Alex Bligh Xara Networks

Does this call into question the idea that there is a single "default-free" view of the Internet? My box is attempting to collect a view of the routing system as seen at a particular point in the topology. While this may not be interesting to Randy, others might find it useful. I would be interested in tracking the views as seen at a number of different places in the topology. The deltas between the growth in the number of widely dispersed routes vs the local varience would be an interesting number to track. --bill

Tony Li <tli@jnx.com> writes: * * Hey, this looks pretty good. Let's hope the trend continues (if, in fact * , * there *is* a trend). * * >Date Prefixes * >231196 40687 * >241196 40908 * >251196 40844 * >261196 41043 * >271196 41074 * >281196 39803 * >291196 39735 * * I suspect that this is more due to the holiday than any real change. ;-( Could be but the most interesting thing is it does look like some fairly large sets of prefixes went away [from my vantage point ;-)] around the 28th. I just added to my delta checker both classful and classles delta checks to help show the change. Here's what the trend looked like today. Date Prefixes 241196 40908 251196 40844 261196 41043 271196 41074 281196 39803 291196 39735 301196 39676 011296 39610 Here are the Deltas (Note AS174 and AS4200, if your are interested in knowing what went away to cause these deltas mail me directly) This a snapshot of changes in classful routes being withdrawn and added. Top 20 Withdrawn Classful Routes from 24Nov96 to 01Dec96 -319 AS174 Performance Systems International -133 AS3804 Bell Blobal Solutions -98 AS4200 AGIS (Apex Global Information Ser -75 AS681 KAWAIHIKO-1 -62 AS4454 OIR Telecommunications, State of -59 AS2048 State of Louisiana/Office of Tele -51 AS3749 Tennessee Board of Regents -48 AS568 JIS (Joint Interconnection Servic -47 AS560 BBN Planet, New England Region (N -45 AS4983 Intel Corporation Autonomous Syst -34 AS2044 AUSNet, Inc. USA -27 AS721 DLA-ASNBLOCK-AS -23 AS3343 RUNNet -21 AS71 Hewlett-Packard -19 AS3904 ASTHOUGHTPRT -17 AS6474 FLASHNET-AS -14 AS2386 INS-AS -13 AS4239 Coast to Coast Telecommunications -12 AS3249 Estpak Data Ltd. -11 AS4175 CONNECT-NCM Top 20 Added Classful Routes from 24Nov96 to 01Dec96 87 AS297 NASA Internet 24 AS4565 HLC Internet 19 AS6347 DIAMOND 17 AS3320 Deutsche Telekom AG 14 AS1717 RENATER 13 AS813 UUNET Canada (ASN-UUNETCA-AS1) 11 AS4270 RIU 8 AS6463 Rogers Network Services 6 AS3563 Pilot Network Services, Inc. 5 AS5769 Le Groupe Videotron Ltee 4 AS5553 TSTU 3 AS3951 ICONNET 2 AS3354 THENET-AS-1 1 AS816 UUNETCA-AS4 This a snapshot of changes in classles routes being withdrawn and added. Top 20 Withdrawn Classles Routes from 24Nov96 to 01Dec96 -228 AS4200 AGIS (Apex Global Information Ser -43 AS5672 PBI (Pacific Bell Internet) -24 AS4648 APNIC-AS-BLOCK -23 AS1239 SprintLink Backbone -18 AS1 BBN Planet backbone -10 AS6912 Hayes Computer Systems, Inc. -6 AS4983 Intel Corporation Autonomous Syst -5 AS3561 MCI -4 AS4492 RTD -3 AS6971 WCO -2 AS5696 Primary AS for GoodNet -1 AS3742 SEMAPHORE-1 Top 20 Added Classles Routes from 24Nov96 to 01Dec96 10 AS701 Alternet 8 AS4565 HLC Internet 6 AS2548 DIGEX-AS 4 AS1717 RENATER 3 AS3559 KORNET-3559 Autonomous-system 2 AS4750 APNIC-AS-BLOCK 1 AS6299 Telalink Corporation --Tony