[net.news] Strange news article path

joel@gould9.UUCP (05/05/86)

[Preface to net.gods: this is not meant as a criticism, merely an
 observation on the deviation between theory and reality.]

There are a large number of USENET sites on the west coast, including
several "backbone" sites.  However, there doesn't seem to be a lot
of north/south traffic on the net, particularly to Southern California.
Instead, news seems to propagate east/west several times before it
reaches here.  This has been noted many times previously by others.

The path of the enclosed article (which may still be on your system) 
is perhaps exaggerated, but not really atypical.   By my calculations, 
it took nearly 6 days, visited the east coast twice (mdivax1 is near 
Vancouver, B.C.) and went through 25 sites before coming to San Diego via 
the "unofficial" decvax->ittatc->dcdwest->sdcsvax link (which, although 
not a "backbone" link, carries about 85% of the traffic into San Diego)

    Path: sdcsvax!dcdwest!ittatc!decvax!bellcore!petrus!sabre!zeta!epsilon!gamma!ulysses!mhuxr!mhuxn!ihnp4!qantel!lll-lcc!lll-crg!seismo!rochester!bullwinkle!uw-beaver!ubc-vision!mdivax1!haft
    From: haft@mdivax1.UUCP (haft)
    Newsgroups: net.wanted,net.decus,net.unix
    Subject: NFS and ULTRIX
    Message-ID: <201@mdivax1.UUCP>
    Date: 25 Apr 86 16:57:06 GMT
    Date-Received: 1 May 86 00:38:25 GMT
    Organization: Mobile Data International Inc., Rich., B.C., Canada
    Lines: 14

Some other paths:
    Path: pyramid!pesnta!hplabs!qantel!lll-lcc!lll-crg!seismo!rochester!bullwinkle!uw-beaver!ubc-vision!mdivax1!haft
    Date-Received: 30 Apr 86 10:24:41 GMT

    Path: bonnie!clyde!watmath!utzoo!utcsri!ubc-vision!mdivax1!haft
    Date-Received: 1 May 86 01:37:28 GMT

I would note that the actual path bares only slight resemblance to
the nominal backbone links.  In particular, 
	uw-beaver!tektronix!hplabs
	uw-beaver!tektronix!decvax
	ubc-vision!alberta!ihnp4
each had 4 days or so and still couldn't beat a more circuitous route.  

(Just for fun, I've shown the west coast sites in upper case.)

                              >>>>>>>>>>>>>>>>>>>>>>>>>>>>(1 site)>>>>>>>V
                              ^            May 1                         V
 cbosgd-----------------------^-----------clyde<<<<<<<<<watmath<<<<<<<<utzoo
   | >>>>>>>>>>>>>>>>>>>>>>>>>^>>>>>>>>>    |                            :
   | ^                        ^        >    |                            :
 ihnp4-------alberta------UBC-VISION   V   burl                          :
   ^ \                Apr 25^ V        V    |  >>>>>(5 sites)>>>>V       :
   ^ cuae2              <<<UW-BEAVER   >>>ulysses-------------bellcore   :
   ^                    V     |                                  V       :
   ^                    V TEKTRONIX---------------------------decvax---linus
   ^                    V     |                                /V|       | 
 QANTEL>>>>>>>>>>>>>>>>>V>>>  | OLIVEB------GLACIER------DECWRL V|       |
   ^                    V   > | /                               V|       |
   ^  kddlab------------V---HPLABS-------SDCRDCF                V|       |
   ^    |          (2 sites)  : Apr 30      |                   V|       |
   ^    |               V     :          SDCSVAX<<<<<<(2 sites)<<|       |
   ^    |               V    hao            | May 1              |       |
   ^    |               V     :          drillsys                |       |
   ^    |               >>>>V :             |                    |       |
   ^  mcvax.................seismo........gatech-----akgua-----mcnc----philabs
   ^                         < \                                         |
LLL-LCC<<<<<LLL-CRG<<<<<<<<<<   \--------------------------------------cmcl2

Sure, this shows that there's enough redundancy in the net that news
will propagate even when several backbone links are out.  But it also
suggests that some of these links are nominal at best.  Also, I suspect
there should be more efficient ways of moving the news along the west 
coast rather that sending it trans-continent 4 times.

(PS: I would appreciate any date stamps and paths from other backbones).
-- 
	Joel West	 	(619) 457-9681
	CACI, Inc. Federal, 3344 N. Torrey Pines Ct., La Jolla, CA  92037
	{cbosgd, ihnp4, sdcsvax, ucla-cs} !gould9!joel
	joel%gould9.uucp@NOSC.ARPA

rap@oliveb.UUCP (05/07/86)

I have to agree with this, especially in view of the  recent  talk  of
cutting the costs of net news.  Our Unix Administration group, not too
long ago, pointed out two Eastern sites to each other that were  about
25  miles  or  so  apart and were both polling California for news and
mail.  Maybe someone should take a good look at the distribution paths
of  net  news and recommend a reorganization to the appropriate system
administrators.
-- 
					Robert A. Pease
    {hplabs|zehntel|fortune|ios|tolerant|allegra|tymix}!oliveb!oliven!rap

ekrell@neptune.cs.ucla.edu (Eduardo Krell) (05/07/86)

ucla-cs (in southern CA) and nike (in northern CA) have established a news
link for ca.* newsgroups. ucla-cs has a news feed with sdcrdcf (a backbone)
so I expect this to be a good link between northern and southern CA.
    Eduardo Krell               UCLA Computer Science Department
    ekrell@ucla-locus.arpa      ..!{sdcrdcf,ihnp4,trwspp,ucbvax}!ucla-cs!ekrell

blarson@usc-oberon.UUCP (Bob Larson) (05/09/86)

I've noticed this phenominon also.  There are several things to
remember when optimising news paths, including:  cost is not
proportinal distance (besides phone company rate structure anomilies
(interstate vs. intrastate) there are things such as dedicated lines,
network connections (bitnet, arpanet), discounts to frequently called
areas, and pc-pursuit (sp?)), connection quality is not proportional
to distance, and usenet costs should preferably be shared.

Our news feed here at usc-oberon is bursty.  Often little news will
arive for several days then large batches will appear on a single day.
Once this there was almost nothing for a week, and then the weeks
worth of news came in over about 36 hours.  (Much of it winding up in
"junk" because it was over 2 weeks old.)  Our main news feed is
sdcrdcf, a "backbone" site, but we also exchange news with smegol and
uscvax (besides a bunch of local systems that have us as the only
newsfeed.)  The week long news lack was probably due to problems with
sdcrdcf, since the news that did come in seemed to come either from
smegol or trwrb!uscvax.

One thing I am considering is getting news running on my computer,
then maybe getting pc-pursuit and limited (to the newsgroups I want to
carry) newsfeeds in the cities it serves.  Several people taking this
aproach could vastly improve the connectivity of usenet for the groups
they are interested in.  
-- 
Bob Larson
Arpa: Blarson@Usc-Ecl.Arpa
Uucp: ihnp4!sdcrdcf!usc-oberon!blarson

grr@cbmvax.cbm.UUCP (George Robbins) (05/12/86)

In article <829@oliveb.UUCP> rap@olivej.UUCP (Robert A. Pease) writes:
>I have to agree with this, especially in view of the  recent  talk  of
>cutting the costs of net news.  Our Unix Administration group, not too
>long ago, pointed out two Eastern sites to each other that were  about
>25  miles  or  so  apart and were both polling California for news and
>mail.  Maybe someone should take a good look at the distribution paths
>of  net  news and recommend a reorganization to the appropriate system
>administrators.
>-- 
>					Robert A. Pease
>    {hplabs|zehntel|fortune|ios|tolerant|allegra|tymix}!oliveb!oliven!rap

I guess it wouldn't hurt to review some of the paths, but you should note
that most of the wild coast-to-coast routes represent 'private' paths that
piggy back on intra-corporate leased lines or other networks, where the
distance/cost tradeoffs are irrelevant or swamped by other usage.

Unfortunatly, most of these links are unofficial, and it would be fatal
to advertise or commit them to outside use.  One just has to accept that
their presence speeds transport across the net, and improves redundancy.

Neither does this reduce the importance of the offical backbone.  While
they could probably reduce their long-haul calling without major effect,
they still serve an important role as regional transfer and distribution
sites - check how many articles do go through one or more backbone sites.

What may people are really looking for is a simple and equitable way to
separate the techie stuff from the humanities stuff from the drivel.
That allows a site to decide what areas they can justify paying for, and
communicate their selection to other sites on the net.
-- 
George Robbins - now working with,	uucp: {ihnp4|seismo|caip}!cbmvax!grr
but no way officially representing	arpa: cbmvax!grr@seismo.css.GOV
Commodore, Engineering Department	fone: 215-431-9255 (only by moonlite)

glenn@sdcrdcf.UUCP (05/13/86)

  UUCP on the machine "sdcrdcf" has had many problems during the last 5 or 6
months.  We apologise for the inconvenience that may have resulted due to slow
news or mail.

  About six months ago we moved the entire computer facility from one building
to another (brand new) building.  Moving that machine ('780) was like moving an
old piano -- it's never quite the same again.  We've also had no end of trouble
with telephone lines and electrical power.  More recently we've had problems
with the sdcrdcf <--> sdcsvax link but, as of the last few days, things are
looking much better.

  We're under a LOT of pressure to reduce our phone bills.  So we've reduced
sites and make calls during the night as much as possible.  From what I hear
They're trying to keep from making many long-distance calls, which naturally
increases the burden of sdcrdcf to move news and mail.

  We're sticking with it at least until there is some reorganisation of USENET
to see if we can still afford to provide the service we are now.  Hopefully
there will be a more diverse set of backbone hosts and that will relieve the
current backbone from moving *all* of the traffic.

Glenn

--------------------------------------------------------------------------------
Glenn C. Scott
System Development Corporation, 2525 Colorado Ave, Santa Monica, CA. 90406 
213-820-4111 X5653	UUCP: glenn@sdcrdcf.UUCP   ARPA: Scott@DOCKMASTER.ARPA

joel@gould9.UUCP (05/13/86)

In article <526@gould9.UUCP>, I wrote:
> The path of the enclosed article (which may still be on your system) 
> is perhaps exaggerated, but not really atypical.
> 
>     Path: sdcsvax!dcdwest!ittatc!decvax!bellcore!petrus!sabre!zeta!epsilon!gamma!ulysses!mhuxr!mhuxn!ihnp4!qantel!lll-lcc!lll-crg!seismo!rochester!bullwinkle!uw-beaver!ubc-vision!mdivax1!haft
...
>     Message-ID: <201@mdivax1.UUCP>

It appears that both sdcrdcf and cbosgd have been having problems recently
(I understand the latter is fixed).  And anyone who's seen hplabs!Unknown
will know that hplabs has been having problems.

On the positive side, I'd just like to note the following article from
the same site:

    Path: gould9!ncr-sd!sdcsvax!sdcrdcf!hplabs!tektronix!uw-beaver!ubc-vision!mdivax1!haft
    From: haft@mdivax1.UUCP (haft)
    Newsgroups: net.periphs,net.graphics,net.wanted
    Subject: Terminal Graphics
    Message-ID: <208@mdivax1.UUCP>
    Date: 9 May 86 23:14:25 GMT
    Date-Received: 12 May 86 11:25:56 GMT

After my previous observation, this article followed *EXACTLY* the 
prescribed backbone path.  It never even left the west coast.

Does the man have power, or what? :-)
-- 
	Joel West	 	(619) 457-9681
	CACI, Inc. Federal, 3344 N. Torrey Pines Ct., La Jolla, CA  92037
	{cbosgd, ihnp4, sdcsvax, ucla-cs} !gould9!joel
	joel%gould9.uucp@NOSC.ARPA