Problem solve Get help with specific problems with your technologies, process and projects.

Virtual aggregation: Lifeline for exploding Internet routing tables

Virtual Aggregation offers carriers steps to reduce the impact of the Internet routing table explosion on their networks and extend the life span of routers for up to 10 years, especially with the IPv4 address shortage, IPv6 deployment and fast BGP and routing table growth affecting core Internet routers.

Virtual aggregation technology ... can reduce the size of the forwarding table on a carrier edge router.

Ivan Pepelnjak
IP Expert

Internet routing tables are exploding, which means IPv4 address depletion will only make matters worse by introducing lots of shorter IPv4 prefixes generated by service providers buying and selling chunks of IPv4 address space.

Introducing IPv6 will only exacerbate the problem since it doesn't address the needs of multihomed sites, and we will have to live with the transition of running IPv4 and IPv6 concurrently for years, if not decades.

Another sobering note is that the Internet Engineering Task Force (IETF) recognized in 2007 that routing scalability was the most important problem facing the Internet -- and it still is.

Virtual aggregation technology offers a solution that can reduce the size of the forwarding table on a carrier edge router by a factor of 10 to 20 (see Making Routers Last Longer with ViAggre), extending the usable lifetime of these boxes for up to a decade.

The impact of BGP and IP routing tables on your network

When you try to evaluate the impact of exploding Border Gateway Protocol (BGP) tables and IP routing tables on your network, it's important to keep in mind the multi-layer structure of BGP, IP and forwarding tables implemented in most modern routers. Here's why:

  • BGP tables store all viable information received from BGP neighbors. Each IP prefix reachable in the Internet might have numerous entries in the BGP table, as it could be advertised by several BGP peers. For example, Hurricane Electric receives more than 10 different prefixes for Facebook's IP prefix on its Fremont core router.
  • IP routing tables store the best available routing information collected from connected IP subnets, static routes and various routing protocols (for example, BGP and OSPF). Only the best information from each routing protocol (for example, the best BGP route) competes for inclusion in the Internet routing table.
  • Forwarding tables store the information actually used to forward IP packets in the Forwarding Information Base (FIB). They contain a fully computed version of the IP routing table and additional information (for example, Layer-2 headers and MPLS labels) needed in the packet forwarding process.

BGP tables and IP routing tables are used only by the main router's CPU and can thus reside in low-cost, lower-speed dynamic random access memory (DRAM). The high-speed forwarding hardware (usually custom-built ASICs needs fast access to the forwarding table, which is usually implemented in high-speed, expensive static memory or even more expensive Content Addressable Memory (CAM). The lifespan of today's high-speed routers is thus limited primarily by the size of the forwarding table, which is hard or impossible to upgrade. Upgrading the main CPU's memory is a much cheaper operation.

Extending router lifespan with virtual aggregation

Virtual aggregation technology (formally known as FIB Suppression with Virtual Aggregation) tries to extend the lifespan of today's routers by reducing the amount of information transferred from the BGP table (through an IP routing table) into the FIB. Virtual aggregation design is the opposite of MPLS-only core design used in many large networks today. The basic idea is simple: Older edge routers in a service provider network should need a minimum amount of forwarding information and use default routing to pass the rest of the traffic to the bigger (newer, more capable) core routers that can still hold all of the Internet routing information in the FIB.

Virtual aggregation: A lifeline for exploding Internet routing tables -- Graphic 1

Click the image above to view the graphic in its full size.

The idea sounds simple, but it has a huge stumbling block. Imagine you need to forward packets between X1 and X2 in our diagram, but PE1 has no route for X1 or X2. PE1 will forward the packets toward P1, which will use its more complete FIB to send them back to PE1; we've just created a nice routing loop. (See below)

Virtual aggregation: A lifeline for exploding Internet routing tables -- Graphic 2

Click the image above to view the graphic in its full size.

To solve the routing loop problem, we need to build tunnels that will reach from the core routers (with their full Internet FIB) past the edges of the network. Yet again, MPLS is the solution. With proper BGP configuration, the PE routers propagate the IP addresses of external next hops to the core routers, which can then use MPLS paths (which are established automatically once you configure MPLS in your network) to push the packets past the network edge. The graphic below shows a "simple" virtual aggregation scenario. You could use the same technology to build a hierarchy of virtual aggregation routers or split the whole IP address space into smaller chunks handled by different core routers.

Virtual aggregation: A lifeline for exploding Internet routing tables -- Graphic 3

Click the image above to view the graphic in its full size.

The virtual aggregation technology is documented in an informational IETF draft, which is close to its final form. It can be easily implemented on existing routers running Cisco IOS or Juniper's Junos (and probably on routers from some other vendors). An open-source implementation has been built for the Quagga open-source routing platforms.

About the author: Ivan Pepelnjak, CCIE No. 1354, is a 25-year veteran of the networking industry and a Ask the Expert panelist. He has more than 10 years of experience in designing, installing, troubleshooting and operating large service provider and enterprise WAN and LAN networks and is currently chief technology advisor at NIL Data Communications, focusing on advanced IP-based networks and Web technologies. His books include MPLS and VPN Architectures and EIGRP Network Design. Check out his IOS Hints blog.

This was last published in July 2010

Dig Deeper on Telecommunication networking