Top 50 unstable DVMRP routes

[Bill Fenner <]

Dear MBone and MBone-EU-OP mailing lists,

  Enclosed is the list of this week's worst DVMRP flappers.
Please see the end of this message for more information.

     AS #srcs  #down(%)  #flapping Route               Rank avg.flaps/hour
------- ----- ---------- --------- ------------------- ---- --------------
      1	  496  496(100%)    0(0%)  none                1       0
    721	   58   58(100%)    0(0%)  none                2       0
     -1	  646  645(99%)     0(0%)  none                3       0
      0	   84   84(100%)    0(0%)  none                4       0
   7018	   17   17(100%)    0(0%)  none                5       0
     71	   25   25(100%)    0(0%)  none                6       0
     33	    3    3(100%)    0(0%)  none                7       0
    714	  538  538(100%)    0(0%)  none                8       0
      3	   21   21(100%)    0(0%)  none                9       0
   3360	    1    1(100%)    0(0%)  none                10      0
   6172	   31   31(100%)    0(0%)  none                11      0
   7015	    5    5(100%)    0(0%)  none                12      0
   6401	    7    7(100%)    0(0%)  none                13      0
   2686	   19   19(100%)    0(0%)  none                14      0
    237	  100  100(100%)    0(0%)  none                15      0
     32	  159  159(100%)    0(0%)  none                16      0
    174	  228  228(100%)    0(0%)  none                17      0
   7332	    4    4(100%)    0(0%)  none                18      0
   7377	    3    3(100%)    0(0%)  none                19      0
   6878	    1    1(100%)    0(0%)  none                20      0
   2647	    1    1(100%)    0(0%)  none                21      0
   1680	    8    8(100%)    0(0%)  none                22      0
   3320	    7    7(100%)    0(0%)  none                23      0
      9	   44   44(100%)    0(0%)  none                24      0
     16	   27   27(100%)    0(0%)  none                25      0
     34	    8    8(100%)    0(0%)  none                26      0
     46	    4    4(100%)    0(0%)  none                27      0
     27	   12   12(100%)    0(0%)  none                28      0
    226	   36   36(100%)    0(0%)  none                29      0
     17	   20   20(100%)    0(0%)  none                30      0
    786	  457  457(100%)    0(0%)  none                31      0
     25	  218  218(100%)    0(0%)  none                32      0
    224	   30   30(100%)    0(0%)  none                33      0
    114	   20   20(100%)    0(0%)  none                34      0
   1852	   12   12(100%)    0(0%)  none                35      0
     22	   26   26(100%)    0(0%)  none                36      0
   1785	   97   97(100%)    0(0%)  none                37      0
   2637	   40   40(100%)    0(0%)  none                38      0
     13	   20   20(100%)    0(0%)  none                39      0
   1201	    2    2(100%)    0(0%)  none                40      0
   3354	   21   21(100%)    0(0%)  none                41      0
     26	   12   12(100%)    0(0%)  none                42      0
     55	   11   11(100%)    0(0%)  none                43      0
   1717	  298  298(100%)    0(0%)  none                44      0
     73	   12   12(100%)    0(0%)  none                45      0
     52	   42   34(80%)     0(0%)  none                46      0
    239	    7    7(100%)    0(0%)  none                47      0
    217	   10   10(100%)    0(0%)  none                48      0
     41	    8    8(100%)    0(0%)  none                49      0
   1742	  199  199(100%)    0(0%)  none                50      0


  UCLA and PARC are engaged in MBone monitoring and
measurement.  One of the major problems with the current DVMRP
MBone is route instability.  Most of the instability appears to
be caused by pathologies as opposed to real link or router
instability.  DVMRP does not handle instability well, so it's to
everyone's benefit to eliminate these pathologies.  This
periodic message is meant to point out the routes that appear
unstable in the hopes that the cause of the instability will be
corrected.

  More information on the data collection is available from

	http://ganef.cs.ucla.edu/~masseyd/Route/

  More information on the known bugs which cause instability
is available from

	http://ganef.cs.ucla.edu/~masseyd/Route/bugs.html

  We are still experimenting with the best format in which to
present this information.  We're currently grouping routes
by unicast AS number, and only reporting the worst offender
in a given AS.  Other unstable routes in the same AS can be
found by referring to the "AS Number Search" from the search
menu of the above web page, or by looking at this week's or
this month's current data on the web page.

  The fields in the following table are:
AS:
	The AS number associated with the source in the unicast
	routing table.  This may be the route's source, or may be
	the route's source's unicast provider, who may or may not
	be their multicast provider.
#sources:
	The number of sources in the DVMRP routing table associated
	with this AS.
#down:
	The number of above sources whose DVMRP routes are currently
	considered down.
#flapping:
	The number of above sources whose DVMRP routes are experiencing
	routing instability during this period.
Route:
	The most unstable route in the AS over the measurement period.
Rank:
	The rank (by flaps/hour) of this route in the top AS's.
flaps/hour:
	The average flaps-per-hour of the most unstable route.
	A stable route never flaps; if a link goes down and
	comes back up routes related to it might flap once or
	twice.  A route which flaps, on average, more than once
	per hour over a week's time, is clearly either behind a
	link that goes up and down about once an hour (which
	deserves to be investigated) or is experiencing a
	pathology like those listed on the above-referenced bugs
	page (which deserves to be investigated).

  The table lists the 50 worst AS's, sorted by the IP address of the
most unstable route in the AS, with an additional rank by most unstable
route in the AS.  This data is from a single point of view at
UCLA; the data is of course colored by its collection point.  In
particular, network problems close to the collection point can
raise the perceived instability for a large number of sources.
However, we believe that averaging over the week lessens these
effects greatly, so although there may be occasional blips, in
general if you're in this table, you should try to find out
why.