Stephen, is the passage following my signature below the event which you are referring to? This ATT WorldNet mishap took place last week.
I've forwarded this to one of my Internet Society folks and I've requested an explanation without the fluff and feathering. If I get it, I'll post it.
In general though, there are approximately 67,000 independent routes on the Internet that must be maintained in routing tables now. This is up smartly from just a few thousand routes just a couple of years ago, and there is no end in sight. Managing these tables is a major problem due merely to their sizes at this point.
Keeping the number of additional routes contained through route consolidations among disparate ISPs (and therefore a cause for route and other resource sharing, and the potential for collateral mishaps) is a key goal throughout the net by its operators at this time, and will continue to be this way until routers that are more capable come along. I.e., routers which can handle much larger tables while still being able to forward and process them at increasing speeds. This is not a trivial issue, by the way.
Also, routing tables are refreshed as a result of ongoing "route announcements" which are sent. If an adjacent service provider forwards, or blocks, the wrong route announcements in violation or in contradiction to established peering relationships, then the sort of problem that you cited in your post can occur. I don't mean to suggest, by any means, that this actually took place in the case in point, rather I am merely providing one example of how some ISP routes can become shared, and how at least one potential mishap might occur. If I get further information, I'll post.
In the meantime, if anyone wants to lend correction or further comment to the above, please do so.
Frank
-------from ATT:
Additional information from the AT&T web site about Wednesday's Worldnet
problem.
>An Apology... And an Explanation.
>
>On December 1, 1999, there was an outage that affected AT&T WorldNet Service
>members. First I would like to apologize for any inconvenience this may have
>caused, and second, I'd like to explain what happened.
>
>This problem was not caused by AT&T's network. However, it did affect our
>system, and our members. Another Internet Service Provider made a network
>change that caused the public Internet to have incorrect information on how to
>reach AT&T WorldNet servers. As a result AT&T WorldNet customers could not
>browse the Internet, and mail from other Internet Service Providers to AT&T
>WorldNet members was delayed. AT&T engineers worked to isolate the problem and
>provided a work around to restore service by 4:30 p.m. In addition, AT&T
>notified the other Internet Service Provider, who in turn corrected their
>configuration. The smooth operation of the entire Internet depends on accurate
>information and cooperation from all Internet Service Providers. This reflects
>the distributed nature of the Internet and its design characteristics.
>
>AT&T WorldNet Service continually works to improve its service. In fact, based
>on a recent study by Solomon Wolff Associates, AT&T WorldNet Service was rated
>as having the highest level of customer satisfaction among the five largest
>Internet Service Providers.
>
>Again, you have our sincerest apologies for yesterday's outage. Please be
>assured that we are working to prevent this kind of an interruption in the
>future.
>
>Thank you for being an AT&T WorldNet Service customer.
>
>Ed Chatlos
>AT&T WorldNet® Service Vice President & General Manager
Earlier, someone on NANOG had written:
On Wed, 01 December 1999, <so and so> wrote:
> C: Could this be related?
> infoworld.com I
> don't think so, but it's an interesting note nonetheless. I don't
> see how DNS could be attached to the route flaps, but again I don't
> have a crystal ball into AT&T's network.
I don't have a crystal ball either, I don't even have a chair.
Here is what AT&T posted on their web page about yesterday's event.
>At approximately 8:15 AM [EST] on Wednesday, December 1, AT&T began
>experiencing a series of service interruptions on its dial Internet Protocol
>[IP] platform. The impact was nationwide in scope and affected AT&T's Worldnet
>Service, Business IP Dial Service, and portions of its Virtual Private Network
>service. AT&T 's IP backbone was not affected by this incident, nor were there
>any service disruptions to customers using AT&T's dedicated Internet access
>service.
>Reports indicate that incorrect routing by another Internet Service Provider
>[ISP] generated this problem. That routing error has been corrected, the
>network has been stabilized, and service has been restored.
>We apologize for any inconvenience these interruptions may have caused. As
>always, you can report problems with AT&T WorldNet Service by using the Contact
>Us link on the left side of this page.
>Thank you for using AT&T WorldNet Service. |
