Is November 12 D-Day for 512k prefixes?

During his presentation on October 7th at NANOG on the trends for prefixes over the past half-decade, Jim Cowie, chief scientist at Dyn Research, found something interesting: We may be reaching “512K day” as soon as next month.

Back in 2009, a typical IPv4 routing table contained only 269k entries. Today, in 2014, it is around 471k, and it is projected to be 519k in 2015… if not sooner. See the chart below from Cowie’s presentation.

Many older but still-in-use Cisco routers can only handle 512k border gateway protocol (BGP) routing entries in their TCAM memory. Indeed, a major outage occurred on August 12th of this year, when an accidental de-aggregation of 20k prefixes pushed the consensus routing table size over that 512k limit.

Was this an aberration, or an industry wake-up call to the fact that we are rapidly approaching a persistent 512k number of BGP entries? The chart above certainly suggests that latter. Cowie predicts that as the consensus BGP routing table hits 512k organically, we’ll have a much larger and longer outage as everyone works to upgrade or replace the older routers’ TCAM memory.

In that way, the outage on August 12th can be seen as a mixed blessing; because of it, a number of routers will already have been upgraded by the time the consensus routing table reaches over 512k entries.

As we stated in a previous blog post about that August 12th event, the visibility and proactive monitoring that route analytics provide would alert IT teams to “danger signs” when things are over capacity. One of the reasons to be optimistic about BGP is that software defined networking will allow routing configuration changes to happen automatically, in near-real-time. This is important since we may be reaching the 512k mark by mid-November, according to Cowie.

Just one question: Weren’t we supposed to switch over to IPv6 in June of 2012?