The company's streaming service runs on RealNetworks' RealVideo and Microsoft Corp.'s Windows Media Technologies servers. WorldStream's service hosts archived movie broadcasts, for instance, on the Web. Corporate events with video feeds, such as earnings-report conference calls, run from 8 Kbps for low-end voice to 1.5 Mbps for a broadcast-quality, fat video stream. Because RealVideo and Windows Media Technologies operate at the TCP packet layer, Intira's network uses Layer 4 switching in Foundry Networks ServerIron switches, mostly for load-balancing among the WorldStream servers through a process called transparent redirection. It also facilitates caching so that WorldStream can support, for instance, several thousand participants behind a corporate firewall in a streaming session over a T1 link, Suver says.

But, like WorldStream, Intira is still new to the market and its network is far from at capacity, so the streaming service hasn't really been tested. "Intira is nowhere near saturated, and we have yet to host an event with 10,000 to 15,000 audience members. We are constantly monitoring and tuning" our service, Suver says.

Increasing usage by hosting a high-attendance event, for example, would call for establishing an SLA (service-level agreement) along with storage-area-network capacity and more redundancy in Intira's network. This would involve a higher-priced premium service from Intira based on usage, Suver says. For now, WorldStream is negotiating an SLA with Intira for 99.99 percent uptime--the current SLA is 98 percent. That may not sound like a big difference, but an outage 2 percent of the time would be a big problem for WorldStream's streaming service, Suver says. "It's been reliable, so we haven't had to pull out the SLA," he says.

And it's impossible to prevent a network failure, especially when different ISPs are involved in peering relationships. A few months ago, WorldStream had trouble connecting to Intira's network, and as packets were dropped some of its customers got jerky video. A peering point on Intira's network was the culprit--specifically, a BGP (Border Gateway Protocol) routing problem between two providers. Suver says WorldStream is working with Intira to eliminate all single points of failure on its network.

Size of IT staff: 14, most of whom work directly with Intira

Suver's average workweek: 60 hours

Biggest challenge: Keeping availability high while depend-ing on service providers and their peers to sustain links

Latest projects: Implementing BGP to exchange routes with service providers; eliminating single points of failure with HSRP, redundant hardware and improved port-level monitoring tools