Why UK Internet had a bad hair day yesterday
LINX recovers from major problem
LINX, the UK's biggest Internet peering exchange, has resolved a major networking problem that affected the performance of the Internet across Britain yesterday.
Early yesterday morning, an unnamed member of LINX (London Internet Exchange) began spewing broadcast traffic onto the exchange. The Foundry switch it was connected to then suffered packet loss.
The problem persisted even after the offending port was disconnected and resulted in the failure of some peering sessions throughout the day.
LINX has a dual vendor network featuring Extreme Networks Black Diamond i and Foundry Networks BigIron 8000 and 15000 10Gbps-capable layer two switches (see diagram here).
With major problems with one of the core BigIron 8000 Foundry switches, the other Foundry switches started suffering. The Extreme switches weren't affected and traffic was still able to flow over them, however in many cases this meant following a most circuitous alternative path.
Because traffic wasn't flowing smoothly around LINX's network yesterday throughput was severely down as can be seen by the exchange's jagged bit rate graph for yesterday. Normally LINX handles around 11.7 Gbps of traffic but this was probably halved yesterday.
Vanessa Evans, LINX's sales and marketing manager, described yesterday as "hairy" but said things had now returned to normal. Engineers have identified the cause of yesterday's persistent fault and completed repairs to bring the peering mesh at the exchange back to normal. ®
Sponsored: Benefits from the lessons learned in HPC