The Great IGP Debate
I was delighted to see this email last week posted to the IETF discussion list by my friend Ole Jacobsen (Editor & Publisher, The Internet Protocol Journal), indicating that ConneXions reports are now available online. Previously only available in scarce printed formatted, “CONNEXIONS–The Interoperability Report” includes 10 Volumes dating from 1987-1996. There’s a great wealth of information contained in these reports, from both a historical and current technical perspective.
One of my favorite articles, from Volume 5, was the oft-referenced The Great IGP Debate. Part One of the article, IS-IS and Integrated Routing, is provided by Radia Perlman and Ross Callon. Part Two, The Open Shortest Path First (OSPF) Routing Protocol is provided by Milo Medin. The is the first comparative document on the two interior gateway routing protocols, and was quite a useful resource for network operators struggling with which IGP to use during initial commercial Internet backbone deployments. A more recent corollary was provided by Dave Katz’s at NANOG, in his OSPF and IS-IS – A Comparative Anatomy talk.
Several years ago I conducted a survey of which large network operators run which IGP. The results basically (IIRC) indicated that 90% of the top 20 largest ISPs used IS-IS, while only ~30% of the top 100 ISPs used IS-IS. The reasons for this varied, from actual requirements for integrated routing, to open source availability, and, primarily, implementation stability, scalability and documentation. From a feature set perspective, it’s pretty much a wash today between OSPF and IS-IS, and implementation support by the most popular router vendors is pretty even. The IETF’s OSPF WG and IS-IS WG tend to maintain parity from a protocol extensions perspective, and, interestingly, just last week, many of the basic IS-S protocol specifications were progressed (well, IETF “last called”) on the IETF Standards Track — rather than continuing as informative Informational (non-Internet Standard) RFCs — finally!
I do still hear of folks transitioning from one IGP to another, for example, as AOL did several years ago, and although this pretty much seems like “make busy” work these days, given the parity in features between the two protocols, there are legitimate reasons why one IGP may be more preferable in your environment than another.
Finally, while there exists today much more information regarding these two protocols than 15-20 years ago, the CONNEXIONS articles, coupled with two of the more definitive books on these protocols:
- OSPF: Anatomy of an Internet Routing Protocol, John Moy
- Interconnections, Second Edition, Radia Perlman
should continue to be an incredibly useful resource for any IP network engineer. Having used both OSPF and IS-IS in large ISP operational environments, and contributed to standards and feature capabilities of both protocols, I’ll happily admit that the CONNEXIONS articles helped me considerably early-on.
Thanks to the Ole for pointing this out, and to the folks at Charles Babbage Institute at the University of Minnesota for making CONNEXIONS, and many other publications, widely available.