Network World
Monday, October 13, 2008
DNSstuff.com
Get information about your IP
IP Information
50+ On-demand DNS and network tools

Jeff Doyle on IP Routing

Cisco Subnet

Navigation

Is Your BGP Up to Snuff?

Routing best practices change and evolve as we collectively gain experience, and this is nowhere more true than with the Internet itself. Because of the vast and growing economic and cultural dependency on the Internet, that ultimate “network of networks” is subject to a level of scrutiny and analysis far beyond what any mere enterprise network undergoes. There are a tremendous number of amazingly intelligent, diligent, committed people striving to make the Internet better.

Because of these efforts, BGP best practices change faster than those of other routing protocols. If you aren’t paying attention, you can get left behind.

Take, for example, route flap damping. At the turn of the century it was accepted wisdom that damping was a good thing (RIPE-229). Misbehaving prefixes accrued penalties, and with enough penalties the prefix was suppressed for a certain amount of time. Most people understood that this contributed to BGP stability by preventing update churn caused by perhaps thousands of prefixes being withdrawn and then re-advertised in various parts of the world.

Then in 2002 a new study demonstrated that, surprisingly, this understanding was wrong and in fact damping causes more instability than it prevents: A flap causes upstream routers to select and advertise next-best routes, which are seen as prefix attribute changes, which in turn accrue their own penalties, which can cause the newly advertised paths to be suppressed. The effect can cascade throughout the Internet’s default-free zone. So where route flap damping was once best practice it is now discouraged (RIPE-378).

Although this advised change has been around for more than a year, a great many network operators are unaware of it.

Conducting a best-practice review of your BGP policies every couple of years is a great way to not only insure that you are up to date, but also serves as a “spring cleaning” exercise to tidy up your policy configurations.


How does one keep up?

Useful answer?
0

A point well taken, Jeff. Where does one go to keep on top of these things? Too much noise on NANOG. Hrm, maybe if RIPE offered an RSS feed of those documents...

Thanks,

Sean

RIPE RSS feed

Useful answer?
0

Sean,

I wouldn't hold your breath on that one...............

RSS Feed

Useful answer?
0

That didn't sound too hopeful, so I created my own:

https://ertw.com/~sean/ripe.xml

Updated daily. My contact information is in the feed if there are problems.

Sean

Visited your web url

Useful answer?
0

Hi Sean,

Visited your web url and this is what appeared:

--------------------------------

There is a problem with this website's security certificate.

The security certificate presented by this website was not issued by a trusted certificate authority.

Security certificate problems may indicate an attempt to fool you or intercept any data you send to the server.

We recommend that you close this webpage and do not continue to this website.

--------------------------------

Sincerely,

Brad Reese
http://www.BradReese.Com

URL

Useful answer?
0

Sorry Brad, I should have given the non https URL

http://ertw.com/~sean/ripe.xml

Sean

Check your other posts too

Useful answer?
0

Sean,

You may wish to check your other posts too, as pretty sure ran into the same problem with past "URLs" that you have posted.

Sincerely,

Brad Reese
http://www.BradReese.Com

Thanks jeff, Good reminder

Useful answer?
0

Thanks jeff, Good reminder about BGP. I also learned that in our network, all the PE routers (they both receive and advertise routes) enable the BGP damping function, while all P routers(they only receive routes) disabled by default.

Actually, in my personal opinion, Route Reflection is one of the problem we are facing in BGP network: It consumes too much routers resource. For example,for the redundancy, every RR client normally establishes peers to two RR servers. In this situation, one BGP RR server may get 4 copies of the global routing table. The 4 copies consume too much router memory and CPU resource, although only one copy is activated.

It seems like the nature of BGP RR mechanism. And it's even worse in the IPv6 network.

Re: Thanks jeff, Good reminder

Useful answer?
0

Hi Ryon,

This is apples and oranges, I think. The only purpose of RRs is to keep IBGP adjacencies scalable; they don’t directly help with route tables. If you eliminate the RRs your alternative is a full IBGP mesh, and you would still have the four copies of the BGP table (I assume from what you’re saying these are from PEs with upstream peering), except you would have them everywhere, not just on the RRs. Assuming you can’t limit the table sizes, about the only choice you have is a beefier router.

Feel free to contact me offline if you would like to discuss this in depth.

--Jeff

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

About Jeff Doyle

Jeff Doyle is president of Jeff Doyle and Associates, an IP network consultancy. Jeff is the author of Routing TCP/IP, Volumes I (read an excerpt) and II and of OSPF and IS-IS: Choosing an IGP for Large-Scale Networks. He is a frequent speaker on IPv6, MPLS, and large-scale routing.

Contact him.

RSS feed XML feed

Jeff Doyle archive.

Cisco Subnet

RSS feed Cisco news RSS feed

Advertisement: