"So we could go from making a million dollars in a day on Christmas Eve to making tens of thousands the next day on Boxing Day," the IT manager said.
Cloud "allows us to scale up very quickly and then be able to go back down to a normal environment without any human assistance... so we don't have to have any sysadmins having sleepless nights."
Prior to the AWS shift, RedBalloon prepared for peak loads by using a lot of caching and having a couple of extra servers on standby. "We didn't have true elasticity like you have in the cloud but you try to provision it, and that's kind of hard with some traditional managed servers where you have to buy servers for two to three years," Keen said.
The company went with Amazon's cloud, although RedBalloon also shortlisted Telstra and Rackspace from a group of some 10 managed services and cloud vendors.
"All three of those were fairly aligned, so we had three criteria we used: local support, cost and technical competency. Amazon were ... probably a year ahead of everyone else in the technology space, pricing - Amazon was cheapest and we actually found the local support in Amazon has been exemplary."
Before the shift, RedBalloon was concerned about latency, especially as Amazon was yet to open a Sydney data centre. "We have run services from the US and they're slow and you can't get around that," Keen said. However, Amazon 'hinted' that the company would be opening a CDN node in Australia, which the company unveiled in June last year.
Now that AWS's Sydney Region has launched, RedBalloon is migrating its infrastructure to the local data centres. However, the company will maintain some services in an overseas region. "The day that we did a million dollars on Christmas Eve was also the day Amazon went down in North Virginia," Keen said.
"I always thought a Region going down was so improbable, I just wasn't going to worry about it. But ... now I've had that sort of warning shot, that actually Regions can go down and if they go down, they're going to go down for a number of hours. So I need to architect around that that and with the cloud you can do that very cost effectively, because you generally pay for what you use.
"We're going to have a couple of servers sitting in a different Region and and they'll just be synchronised. We'll also have all the server images copied across and then if the [Sydney] Region goes down, we can just spin up a new region, make a couple of configuration changes and within about 15 minutes we should be up and running again."
Sign up for Computerworld eNewsletters.