Now that I'm done with my MBA, it's back to business travel. Since a lot of my trips are from East coast to West coast, that gives me a lot of time to read on the flights. Last week, on my flight out to California, I got around to reading an old issue of Cisco's Internet Protocol Journal (IPJ). IPJ is a very good professional journal with topics that you're not going to read about anywhere else. This is scholarly work by networking and Internet professionals from around the world.
The article that caught my attention was "Extending Router Lifetime with Virtual Aggregation". Nearly 3 years ago, after attending and presenting at FutureNet, I wrote a blog about the Internet Melting Down. Essentially, the IPv4 address space was being subnetted into smaller and smaller chunks which would, eventually, explode the Internet routing tables. This explosion of route entries would not fit into existing router forwarding tables which rely on fix-sized TCAMs.
Point of reference. That blog 33 months ago showed the Internet global routing table with 247,868 routes. Today, my Internet routers show 334,689 routes. That's a 35% increase over 33 months or about 1% growth a month.
This Virtual Aggregation concept promoted the idea of FIB suppression. Essentially, all Internet routes could be held in a router's routing table (control plane) since a router has, ostensibly, unlimited RAM to hold the routes. This allows routers to still advertise all routes and operate in the Internet Default-Free Zone (DFZ) and send all routes to neighbors (iBGP or eBGP). But, via FIB suppression, some routes from the Internet routing table are not installed in the router's forwarding table (data plane) since the forwarding table has a strict, limited TCAM size. For routes that do not exist in the FIB, routers using FIB suppression turn to my oldest network design friend - the "I can fix anything with it" tool - a tunnel.
When packets arrive at a router that doesn't have the route in the FIB, instead of dropping the packet, the FIB suppressed router tunnels the packet to a router who does have the route in its FIB and can route it appropriately. Designed correctly, this can extend the life of older routers should the Internet routing table grow beyond their TCAM size. This is a very interesting concept and a good article to read. This is the type of research that will show up as a feature in IOS 20.0 some day.
More >From the Field blog entries:
Go to Cisco Subnet for more Cisco news, blogs, discussion forums, security alerts, book giveaways, and more.