Virtualization Scalability: Bumping Into a Cloud Ceiling

Pascal-Louis Perez tapped cloud services when he became CTO of startup kaChing.com, a website that helps retail investors make better trades. Soon after, his cloud service provider, Xignite, which streams financial market data to kaChing.com, began delivering spotty service.

“Some of their clients’ traffic and a bug affected their system, and therefore our system,” Perez recalls. “Trading was interrupted twice. If we’re unable to get real-time prices, then we’re unable to allow our users to trade.”

Xignite was bumping into capacity limits as new customers continued to flock to the cloud. Perez told Xignite that it needed to get some headroom—and Xignite did exactly that by becoming a customer of Amazon’s Elastic Compute Cloud (EC2) and Simple Storage Service (S3) Web services.

[ Learn about the promise and peril of cloud computing. | Read CIO’s report on the case against cloud computing. ]

Cloud computing has attracted lots of pioneering companies, both as customers and service providers. But these companies often lack the resources and expertise to deliver enterprise-class services. Problems of scalability, reliability and response times crop up. Now companies, mostly cloud providers, are turning to infrastructure services from tech stalwarts like Amazon to shore up their offerings.

A few years ago, Xignite was serving a few million page view data requests a month for the average customer. “Today, it might be 50 million a month for one client and 100 million for another,” says Stephane Dubois, CEO of Xignite. Marquee customers include ING, Forbes.com, Wells Fargo, NetSuite, Starbucks, among others.

Xignite’s datacenter hit capacity overload during a sudden spike last year, and later a customer’s system had a technical glitch that resulted in a massive number of requests sent to Xignite’s servers. Both times kaChing.com experienced slowness and timeouts. “The need to scale orders of magnitude was the triggering event” for Xignite to go to the Amazon platform, Dubois says.

At first, Dubois was dubious of how much it would cost to use Amazon services. Pricing fluctuates depending on usage, and traffic is hard to predict particularly in the finance market. “Cost could be significant,” Dubois says. “We took a bit of a risk.” After rollout though, Dubois says the real cost turned out to be only a quarter of his projections.

Today, Xignite leverages Amazon’s EC2 and S3 services to power its applications for many customers. The services enable Xignite to assess bandwidth, storage and processing needs for each customer and then scale them up and down depending on traffic loads.

Not all of Xignite’s customers were initially thrilled with the move—one, ironically, being kaChing.com. Perez mostly fretted over potential data synchronization issues. “I was concerned there might be a delay,” he says.

When service providers move infrastructure to the Amazon cloud, explains Dubois, they add a new step to the process. “We have our data center where we collect data from the [market] exchanges and essentially mirror this information to the cloud,” he says. “We have to make sure that there’s as little latency as possible in the process.”

Engineers from both Xignite and kaChing.com are working together to make sure that the latency issue isn’t a problem. “Before we go to production, we’ll have a test period for the two systems running in parallel,” Perez says, “and make sure the new system [with Amazon’s services] performs better than the old one.”