Here's an idea that just popped into my head after NANOG. Lets assume for a moment that the majority of route instabilty is coming from inexperienced providers. With that assumption in hand I propose the following incentive: Finanical compensation for bgp dampening policies. Deep pockets = very little flap penalty Shallow pockets = very heavy flap penalty Also, make the assumption that inexperienced providers arn't going to be able to pay for light flap penalties. So their only choice is to be stable or die. I havn't thought about how or where such a policy would be applied, but the underlying idea would help fix the problem of routing instabilty. I know this one is out there, but I just thought I'd mention it. Scott -- smace@neosoft.com - KC5NUA - Scott Mace - Network Engineer - Neosoft Inc. Any opinions expressed are mine.
Here's an idea that just popped into my head after NANOG.
Lets assume for a moment that the majority of route instabilty is coming from inexperienced providers. With that assumption in hand I propose the following incentive:
Finanical compensation for bgp dampening policies.
Humm.. Given your assumption and your conclusion where would the $ go? How about, under the same assumptions, that we all just wind up the dampening timers faster to say 24 hours or so. A couple of days of being off the net tends to be a much more direct $ problem. That said, I like the idea of communcations and/or training, perhaps in a form of a document for "new" ISPs. In my vision this tome could take its place in the collection of literature for new ISP's connecting to a nap.. Possibly the cost of connecting to a given nap could offset this. In theory, we are all in this together. No amount of punishment will really help solve the problem in the long run (although it is sometimes fun to think about)
Lets assume for a moment that the majority of route instabilty is coming from inexperienced providers. With that assumption in hand
I doubt that this is the case. Poor software design seems to be the major cause. As Brian showed, there are lots and lots of needless updates. For example, delaying a withdraw might allow one to realize that it is no longer needed. There seems to be too much emphasis on short term patches vs study and long term solutions.
On Sat, 26 Oct 1996, Doug Davis wrote:
That said, I like the idea of communcations and/or training, perhaps in a form of a document for "new" ISPs.
There already are some documents like this for ISP's at http://www.amazing.com/internet http://www.mtiweb.com/isp http://www.ra.net/isp.html Obviously the collection could be improved with more info on NAP's and BGP and multihoming and similar stuff. If any of you belong to national ISP mailing lists or ISP mailing lists in other languages please pass these URL's on to those lists. These URL's are especially helpful for helping beginning ISP's to learn the tricks of the trade. Michael Dillon - ISP & Internet Consulting Memra Software Inc. - Fax: +1-604-546-3049 http://www.memra.com - E-mail: michael@memra.com
In message <199610260527.AAA08813@crash.ops.neosoft.com>, Scott Mace writes:
Here's an idea that just popped into my head after NANOG.
Lets assume for a moment that the majority of route instabilty is coming from inexperienced providers. With that assumption in hand I propose the following incentive:
Finanical compensation for bgp dampening policies.
Deep pockets = very little flap penalty Shallow pockets = very heavy flap penalty
Also, make the assumption that inexperienced providers arn't going to be able to pay for light flap penalties. So their only choice is to be stable or die.
I havn't thought about how or where such a policy would be applied, but the underlying idea would help fix the problem of routing instabilty.
I know this one is out there, but I just thought I'd mention it.
Fix it in the peering agreements. Settlements based on flap! :) (only half joking) --- Jeremy Porter, Freeside Communications, Inc. jerry@fc.net PO BOX 80315 Austin, Tx 78708 | 1-800-968-8750 | 512-458-9810 http://www.fc.net
Finanical compensation for bgp dampening policies.
Deep pockets = very little flap penalty Shallow pockets = very heavy flap penalty
The main risk (IMHO) when accepting a new peer is how stable the peer is. Without confidence in the design of the peer's network or trust in its engineers to prevent/solve problems, connecting to a new peer with a significant number of routes can be scary. A peer's having deep pockets usually implies that there's more to lose if they mess up and that they _might_ want to hire good engineers and pre-plan their network, but whether they actually do is another story. :^( Perhaps ISPs/NSPs could build "trial periods" into their peering policies where a peer's learned routes are dampened with a high flap penalty. If a provider meets the pre-determined "quality of service level", the dampening is reduced or even removed at the end of the trial period. If they don't, the agreement is broken and you route the peer's traffic through their transit provider instead (which hopefully have some dampening in place). If a good peer starts having service-affecting problems, put them back on "trial period" (aka probation) and notify them. If you don't live up to your own standards, the idea of having a trial period is moot. -- Eric Ziegast Looking in from the edge
participants (6)
-
dougd＠airmail.net -
Eric Ziegast -
Jeremy Porter -
jon＠branch.net -
Michael Dillon -
Scott Mace