Instead, Grant opted for a distributed database management system called NuoDB. Unlike traditional shared-disk architectures, NuoDB says it scales out and in dynamically on commodity and virtual machines and has no single point of failure. "It's a single, coherent solution," Grant says. "They built something from the ground up, implementing well-known standards and protocols."
In addition, Grant was interested in NuoDB's support across geographies. "We're an international company with customers in Madrid, London, and New York," Grant says. "In order to deal with the spiky nature of our traffic, we need to be able to reach a server within the geographical region. We can't jump across the ocean." NuoDB will be able to replicate content from region to region via its intranet. "That was the lynch pin that made me an early adopter," Grant says.
Still, Grant was skeptical about the new system. "It sounded too good to be true," he says. "I wondered if they could deliver. I'm a risk-averse early adopter." So he devised a recovery plan if the system didn't work out as planned that would get the company off NuoDB within a few days. "We'd have to throw servers at it and limp along for a bit," he says. "But we'd survive."
After a year of testing and planning, Grant is comfortable will going live on NuoDB next month. The biggest and most unexpected challenge, he says, has been the changes required of his engineering team. "When you go to work with this databse day to day, you're not hitting a single server. You're accessing and communicating with a cluster," he says. "That's created different issues for our engineers." They have to think and work in new ways when, say, opening a firewall for a report. "It creates a different model not just for writing queries, but also deployment, management, and monitoring. You have to take a different approach to get your mind around that and to build tools around it."
The move to NuoDB will creat an immediate reduction in Shout.tv's infrastructure costs as it grows, says Grant, not just in terms of servers, but also management and monitoring. And he's glad he made the decision to rethink the company's infrastructure early. "Everyone talks about success being a great problem to have," Grant says. "But if you wait until it happens, you're in trouble."
Sign up for CIO Asia eNewsletters.