two questions - SWAG answers entertained
one: AS hop count for average e2e packet flow, eg. from origin to destination, how many ASN's will a packet traverse? two: number/location of IX that monitor/forbid transit across exchange fabric? --bill (doing grunt work for a study on Landauer Entropy)
On Oct 4, 2007, at 11:58 AM, bmanning@vacation.karoshi.com wrote:
one: AS hop count for average e2e packet flow, eg. from origin to destination, how many ASN's will a packet traverse?
You might look at Figure 8c of http://www.multicasttech.com/status/ If you assume that destinations IP addresses are selected at random, that tells you the PDF of the AS hop counts from here. The mode is 3 hops. Here is the data for Noon EDT, in case you want to do more. # hops = Number of transit AS hops required to reach the source from AS 16517 ## hop_histogram Unicast AS Hop Histogram ## hop_histogram Number of Hops | Number of ASN | Number of Prefixes | Address Space Used ## hop_histogram # hop_histogram 1 1 968 2.78565e+07 # hop_histogram 2 1587 32780 7.74648e+08 # hop_histogram 3 12829 122320 1.06436e+09 # hop_histogram 4 8754 56473 2.98075e+08 # hop_histogram 5 2549 14877 4.19192e+07 # hop_histogram 6 432 1932 1.18231e+07 # hop_histogram 7 39 215 872786 # hop_histogram 8 6 32 50112 # hop_histogram 9 0 0 0 # hop_histogram 10 0 0 0 # hop_histogram 11 0 0 0 # hop_histogram 12 0 0 0 # hop_histogram 13 0 0 0 # hop_histogram 14 1 1 4094 This of course is polluted somewhat by as path prepending, which could be stripped off if desired. Regards Marshall
two: number/location of IX that monitor/forbid transit across exchange fabric?
--bill (doing grunt work for a study on Landauer Entropy)
On Thu, 04 Oct 2007 13:01:19 EDT, Marshall Eubanks said:
## hop_histogram # hop_histogram 1 1 968 2.78565e+07 # hop_histogram 2 1587 32780 7.74648e+08 # hop_histogram 3 12829 122320 1.06436e+09 # hop_histogram 4 8754 56473 2.98075e+08 # hop_histogram 5 2549 14877 4.19192e+07 # hop_histogram 6 432 1932 1.18231e+07 # hop_histogram 7 39 215 872786 # hop_histogram 8 6 32 50112 # hop_histogram 9 0 0 0 # hop_histogram 10 0 0 0 # hop_histogram 11 0 0 0 # hop_histogram 12 0 0 0 # hop_histogram 13 0 0 0 # hop_histogram 14 1 1 4094
Are you able to share the identity/AS path of that outlier at 14, or at least explain how somebody got a path length of 14 without any paths of lengths 9 through 13 showing up?
On Oct 5, 2007, at 11:10 AM, Valdis.Kletnieks@vt.edu wrote:
On Thu, 04 Oct 2007 13:01:19 EDT, Marshall Eubanks said:
## hop_histogram # hop_histogram 1 1 968 2.78565e+07 # hop_histogram 2 1587 32780 7.74648e+08 # hop_histogram 3 12829 122320 1.06436e+09 # hop_histogram 4 8754 56473 2.98075e+08 # hop_histogram 5 2549 14877 4.19192e+07 # hop_histogram 6 432 1932 1.18231e+07 # hop_histogram 7 39 215 872786 # hop_histogram 8 6 32 50112 # hop_histogram 9 0 0 0 # hop_histogram 10 0 0 0 # hop_histogram 11 0 0 0 # hop_histogram 12 0 0 0 # hop_histogram 13 0 0 0 # hop_histogram 14 1 1 4094
Are you able to share the identity/AS path of that outlier at 14, or at least explain how somebody got a path length of 14 without any paths of lengths 9 through 13 showing up?
Yes - Here it is : AS 3010 SPRINT-MRN-BLOCK | 1 prefixes | 1 prefixes & 1 ASN supported | 4094 Addresses | 14 hops | as path 174 3491 18734 28509 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 here is the only address block : *> 201.160.16.0/20 38.101.161.116 4001 0 174 3491 18734 28509 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 i I assume that this is some odd internal Sprint thing - maybe a lab in Mexico as it is going through 28509 ? 28509 MX-TCSC23-LACNIC [RUM] {62270 - cuernavaca - MO, MX} TV Cable S.A. de C.V. None of the other 3001-3009 appear directly. Could this be some new form of AS path prepending ? I really don't know. Regards Marshall
How vague can the data be? Path length distribution: Hops in Path: Count 1: 104395 2: 2923484 3: 13778037 4: 17280238 5: 9353555 6: 3795923 7: 1792676 8: 822142 9: 472420 10: 258358 11: 160412 12: 104390 13: 93137 14: 66313 15: 48547 16: 22362 17: 7044 18: 5048 19: 9684 20: 5030 21: 1469 22: 852 23: 497 24: 132 25: 132 26: 229 27: 1288 28: 12080 29: 1609 30: 819 31: 12 32: 1 34: 30 35: 55 And just for kicks, prepends: Prepend distribution: Prepends in Path: Count 2: 3514603 3: 1754138 4: 1209997 5: 562524 6: 336495 7: 263039 8: 64786 9: 85553 10: 90446 11: 89088 12: 2722 13: 18072 14: 494 15: 6 16: 423 17: 69 20: 111 21: 34 22: 80 25: 260
I've had a few questions about the data I posted. That was a total of all ribs I have access to, not a summary of uniques. I thought maybe a report of OIX would be a better example. Here's the full rib report: Total prefixes seen in RIB: 238997 Total origin ASes seen in RIB: 26514 Total 32 bit ASes seen in RIB: 0 Total peer pairs seen in RIB: 55947 RIB Peers not found in all other tables: 4 RIB Prefixes not exactly matched in all other tables: 123 RIB more specific prefix count: 123 Number of unique prefixes in RIB compared against all data: 123 Number of unique origin ASes in RIB compared against all data: 4 Number of ASes announcing only one prefix: 11229 Number of Transit Only ASes: 125 RIB average path length: 4.09 All tables Total IP Coverage: 2549455474 59.4% All tables Total IP Coverage including BOGON: 3220536059 75.0% RIB Total IP Coverage: 1790591076 41.7% RIB Total IP Coverage including BOGON: 3201119844 74.5% RIB IP Coverage by Region RIB IP Coverage for ARIN 1038310514 46.2% RIB IP Coverage for APNIC 399476161 17.8% RIB IP Coverage for RIPENCC 312598027 13.9% RIB IP Coverage for RIPE 197332093 8.8% RIB IP Coverage for JPIRR 110040248 4.9% RIB IP Coverage for UNKNOWN 109583355 4.9% RIB IP Coverage for LACNIC 68517416 3.0% RIB IP Coverage for AFRINIC 12606792 0.6% RIB Total AS Coverage: 26514 RIB AS Coverage by Region RIB AS Coverage for ARIN: 12525 47.2% RIB AS Coverage for RIPE: 10418 39.3% RIB AS Coverage for APNIC: 2648 10.0% RIB AS Coverage for LACNIC: 377 1.4% RIB AS Coverage for JPNIC: 150 0.6% RIB AS Coverage for KRNIC: 131 0.5% RIB AS Coverage for AFRINIC: 84 0.3% RIB AS Coverage for ID-NIC: 77 0.3% RIB AS Coverage for CNNIC: 52 0.2% RIB AS Coverage for PRIVATE: 21 0.1% RIB AS Coverage for TWNIC: 17 0.1% RIB AS Coverage for VNNIC: 13 0.0% RIB AS Coverage for IANA: 1 0.0% Prefix length distribution /8: 20 /9: 9 /10: 16 /11: 37 /12: 136 /13: 271 /14: 483 /15: 946 /16: 9891 /17: 4240 /18: 6843 /19: 14700 /20: 17674 /21: 15748 /22: 20131 /23: 21032 /24: 126563 /25: 619 /26: 638 /27: 353 /28: 138 /29: 102 /30: 122 /32: 50 Path length distribution: Hops in Path: Count 1: 9247 2: 579155 3: 2947058 4: 2928712 5: 1317423 6: 521409 7: 264027 8: 121191 9: 73516 10: 37090 11: 23273 12: 14804 13: 13515 14: 8818 15: 7767 16: 3740 17: 885 18: 888 19: 1130 20: 386 21: 80 22: 103 23: 46 24: 7 25: 14 26: 59 27: 30 28: 1164 29: 7 Prepend length distribution: # Prepends in Path: Count 2: 531052 3: 283552 4: 200956 5: 90091 6: 50456 7: 45784 8: 8428 9: 15447 10: 8736 11: 15066 12: 459 13: 1489 14: 46 16: 67 17: 22 20: 14 21: 2 22: 8 25: 4 Total AS Sets: 1874
--On 8 October 2007 10:50:49 -0400 Jason Lewis <jlewis@packetnexus.com> wrote:
RIB IP Coverage for RIPENCC 312598027 13.9% RIB IP Coverage for RIPE 197332093 8.8%
Just out of interest, what's the difference between "RIPENCC" (the Regional Registry?) and "RIPE" (a vaguely defined group of people who contribute to the RIPE working group mailing lists and/or attend RIPE meetings and/or have some interest in IP networking in and around Europe and/or ... ?) here? James
The nature of registry data leads to bugs like this, they should all be RIPENCC. jas James Aldridge wrote:
--On 8 October 2007 10:50:49 -0400 Jason Lewis <jlewis@packetnexus.com> wrote:
RIB IP Coverage for RIPENCC 312598027 13.9% RIB IP Coverage for RIPE 197332093 8.8%
Just out of interest, what's the difference between "RIPENCC" (the Regional Registry?) and "RIPE" (a vaguely defined group of people who contribute to the RIPE working group mailing lists and/or attend RIPE meetings and/or have some interest in IP networking in and around Europe and/or ... ?) here?
James
On Fri, Oct 05, 2007 at 08:36:16AM +0900, Randy Bush wrote:
one: AS hop count for average e2e packet flow, eg. from origin to destination, how many ASN's will a packet traverse?
wrong forum. there have been good formal measurement papers on this. do a lit search.
randy
ere, this might be the ideal forum. citseer is very helpful, some formal analysis from a few data points. my request for SWAG data from operations folks was/is a way to sanity check simulation work that forms the basis for much of the formal papers. anyway... thanks to the kind folks who have (mercifully) responded off list. we now return you to the usual NANOG postings. --bill
participants (6)
-
bmanning＠vacation.karoshi.com -
James Aldridge -
Jason Lewis -
Marshall Eubanks -
Randy Bush -
Valdis.Kletnieks＠vt.edu