Internet outages expected to abate as routers are modified, rebooted

A flood of updates to databases inside Internet routers caused intermittent outages on Wednesday and connectivity issues for companies, but experts expect the long-forecasted hiccup to be resolved soon.

Internet traffic is designed to flow in the most efficient way, which means frequent updates to routers that describe how networks should connect. Some routers can only accommodate 512,000 of those updates in memory without further tweaks.

Some of those routers have hit their limit, which caused some networks to go offline. The situation, while inconvenient for some, is being regarded as more of a technical bump rather than the meltdown of the Internet, but one that may keep network operators on their toes for the next few days.

“This situation is more of an annoyance than a real Internet-wide threat,” wrote Jim Cowie of Dyn. “Most routers in use today…have plenty of room to deal with the Internet’s current span.” Dyn acquired Cowie’s network performance company Renesys in May.

Cisco warned its customers as early as May that the rising number of route entries could pose trouble, wrote Omar Santos, incident manager for the company’s Product Security Incident Response Team, in a blog post Wednesday.

In just six years, the number of entries in routing tables has doubled, from 256,000 to upwards of 512,000, he wrote. The problem is that as routing entries exceed 512,000, the Ternary Content Addressable Memory (TCAM) contained in switches and routers will run out unless modified. TCAM is a type of memory that is faster than RAM, Santos wrote.

Older products from Cisco are configured by default to only accommodate 512,000 routes. Products that may be affected include the company’s Catalyst 6500 switches, 7600 series routers, ASR 9000 and 1000 Series Aggregation Services Routers in certain configurations. Cisco has published workarounds for that equipment.

The flurry of routing changes appeared to have originated from two networks run by Verizon, wrote Andree Toonk, founder of BGPmon, a network monitoring and security company in Vancouver. The changes from Verizon appear to have pushed global routing tables to 515,000 entries, past the default limit for some types of equipment, he wrote.

Verizon appeared to make some changes that lowered the number of routes, which fixed some issues. But “the Internet routing table will continue to grow organically and we will reach the 512,000 limit soon again,” Toonk wrote. Verizon did not have an immediate comment.

The Lansing, Michigan, based Web hosting provider Liquid Web moved quickly after it experienced issues. It upgraded the memory allocation to its core routers as well as rebooted some, the company wrote on its support blog.

Send news tips and comments to jeremy_kirk@idg.com. Follow me on Twitter: @jeremy_kirk