We've all heard the stories of businesses using Amazon Web Services until they realize that at a certain size, it becomes more cost effective to run their operations in house. ScribbleLive, which offers services including a live blogging platform for major media companies including CNN and ESPN, doesn't see that point coming.
ScribbleLive has supported as many as 1 million people concurrently watching content delivered via its platform. In January, people watched over 3 billion minutes of content offered by ScribbleLive customers.*
The company started out using dedicated hosted servers, said Jonathan Keebler, CTO of ScribbleLive. As cloud services emerged, it began to tinker with them. The first it actually began using was Amazon's S3, back in 2008, to store all the media its customers were producing.
"We were a small company. There was no one to do sys admin type stuff. If we wanted to store data on our own servers with RAID and redundancy, that can be a full time job. At the rate we were growing, there was nobody to do that," he said.
Using S3 worked out well and so ScribbleLive started looking at additional Amazon services. "That was our gateway product to the rest of Amazon," Keebler said. The company now uses a long list of AWS services, including EC2 and DynamoDB, and currently runs all of its operations in AWS.
ScribbleLive found that using web services from Amazon allowed it to focus on more important development. Developers there spend a lot of their time architecting ScribbleLive's technology so that it can scale quickly and efficiently, Keebler said. "We find interesting ways to use CDNs and other technologies out there to buffer traffic wherever we can so we limit the traffic getting back to the servers but then we autoscale when needed," he said.
ScribbleLive uses New Relic to monitor server performance and then looks for ways to make things run more efficiently, he said. "We're always looking at how our code is running in production and saying, 'if I do that call once instead of 50 times, I can really reduce the number of servers we need to run that process,'" Keebler said. "We definitely try to write code to solve our problems rather than just scaling up more and more and more."
That work may be aimed at making up for one inefficiency in ScribbleLive's operation. Since autoscaling isn't instant, ScribbleLive keeps enough servers online so that if required it could scale quickly to sustain the largest amount of traffic it has ever supported.
Like other cloud fans, ScribbleLive says that AWS allowed it to move much faster than it could when it was using a dedicated hosting environment. When it made the initial choice to use dedicated servers it was because the consensus in the market was that using dedicated servers resulted in better performance because users didn't have "noisy neighbors" to contend with, he said.
"But what we'd run into was we'd have a big event and our systems would become near maxed out," he said. Then, it wasn't just a matter of buying new servers. "We'd say we need a bigger firewall, for example, and ask for one and that'd be a 30 day install. It became so inefficient for us to scale," he said.
The cost savings are notable too. He said ScribbleLive has realized a 35 percent OpEx reduction by using AWS.
Keebler has looked at other competing cloud services but has plenty of reason to stick with AWS. "We have so much pressure on us to grow and from our customers to build new apps on our platform that we're sort of pushing the envelope in a lot of ways," he said. "It seems that Amazon always has the next piece that we need just as we need it."
There have been a few instances where he's been discussing with his developers a new challenge and very soon AWS will come out with a service that solves the problem. That happened with AWS's introduction of DynamoDB. "When that was announced, we had an app written and deployed and in production in about a month because that was exactly what we needed," he said. "I do look at other cloud providers and I like a lot of what they're doing but it seems that Amazon's rate of innovation is higher."
Looking to the future, Keebler is hopeful for new developments that will help with extracting insights from big data. Real time big data analysis is lagging, he said. But since this is a common complaint he's hearing when he attends industry conferences, he's hopeful breakthroughs will happen. "Right now I bet there are a lot of smart people in big companies dealing with this same issue," he said. "I think there's an opportunity to solve that problem."
*Correction: An earlier version of the story mischaracterized the 3 billion figure.
Read more of Nancy Gohring's "To the Cloud" blog and follow the latest IT news at ITworld. Follow Nancy on Twitter at @ngohring and on Google+. For the latest IT news, analysis and how-tos, follow ITworld on Twitter and Facebook.