Interop network all about redundancy

15.05.2009
When you’re responsible for the network that powers – one of the premier networking technology conferences in the world – avoiding downtime has to be one of the top priorities.

So when network architect Geoff Horne and his team designed this year’s InteropNet in Las Vegas, they set out to achieve levels of redundancy that did not previously exist at the show. Horne embraced virtualization in a major way , and this year he has begun replicating virtual machines to , a hosted online backup and disaster recovery service. Secondly, the InteropNet on the show floor was built as two parallel networks -- each containing its own server, storage and networking equipment – which can fail over to each other if necessary.

“We really built a much more redundant network. … The biggest challenge was splitting the network in half and designing it accordingly,” says Horne, a network design consultant who has been volunteering for the InteropNet the past 14 years.

Uptime and reliability have become increasingly important over that time span, Horne notes. When he first started, a 15-minute outage was no big deal. Now if the network goes down for five seconds people start complaining. Hundreds of exhibitors need network connections at their booths, while thousands of attendees rely on wireless access points to get e-mail and use the Web.

Working with Horne are 17 volunteers, and about 100 IT pros overall, mainly supplied by the vendors who donate equipment. The InteropNet is built and dismantled in a single week, and in past years included nearly $10 million worth of vendor-donated equipment and software. In addition to the Las Vegas show, the InteropNet will be reassembled for the New York conference in November.

More than 20 vendors are supporting the Vegas InteropNet this month. The network is using 28 blade servers in the IBM BladeCenter H Chassis; about 2TB of EMC storage; Enterasys switches and routers; and VMware’s , a software layer designed to virtualize and aggregate servers and storage into a single computing pool.