Not long ago, Farhan Thawar woke up to a phone call from his sister. There was a power outage in Toronto, she told him, and it included the offices of I Love Rewards, a company that offers branded corporate incentive programs where he works as chief software architect.
Thawar didn’t exactly jump out of bed.
I Love Rewards offers HR departments a way to reward and recognize employees by letting them earn points to win prizes like an iPod or the chance to drive an F1 car. Its services run in a software-as-a-service model, which means downtime would be a major problem. In this case, however, I Love Rewards runs much of its IT infrastructure on Amazon’s Elastic Compute Service (ECS), which means the relevant applications were hosted off site in the cloud.
“We were so glad that a lot of our services had already moved over,” Thawar says. “When my sister was calling and saying there was no power, I knew that our servers were in Seattle.”
Although I Love Rewards had a couple of legacy platforms that required physical intervention if a hard drive went dead, the company has been a steadfast customer of Amazon’s EC2 and also its Simple Storage Solutions (S3) service. A little more than a year ago, the firm moved to a completely virtualized environment.
With the Cloud Computing Conference set for Feb. 11 at the Metro Toronto Convention Centre, ComputerWorld Canada quizzed Thawar for some advice on how to make the transition from on-premise to the cloud a smooth one.
Upgrade your way in
Thawar has already been approached by peers who he says are interested in cloud computing but are nervous about it. “I think they don’t want to be at the hands of a vendor where they can’t go in and (physically) change things,” he says. “I encourage companies to try it with their next server. We may have one or two physical servers we use internally for things, and we can easily move our site onto one of those. But if you just think of it as a way of buying your next machine, it seems a little less of a leap.”
Set your expectations
Like many businesses, I Love Rewards was affected by a service shutdown of Amazon S3 last year, which occurred right around the time Microsoft announced its Azure cloud service. Thawar says companies need to assess the maturity of the cloud service being offered and what’s acceptable from a service delivery standpoint.
“Of course because you’re going to be affected by that kind of outage because you’re now reliant on their SLA,” he says, referring to service-level agreements. “Some of these services have not been out of beta for a long time, but Amazon now has a 99.5 per cent uptime guarantee. That’s less than four and a half hours of downtime a year. We’re comfortable to live within that range.”
Be creative with your options
S3 is typically considered a means of storing server data, but Thawal suggests it could also be used for general backup purposes. Other cloud services may allow an organization to grapple with bursts of demand. This happened at I Love Rewards, where using cloud computing meant not having to buy more servers to deal with the Christmas rush. “We have pre-built image instances we can use whenever we want to add that capacity,” he says.
Have more than one throat to choke
Although he’s an avid Amazon client, Thawar says it makes sense to explore what’s being offered from IBM, AT&T and the plethora of other players entering the market. “If you want to be fully redundant, don’t just stick with one vendor,” he says. “Maybe you could have one server on Amazon, one on Slicehost. And then maybe one in a dedicated server facility. Looking at that from a failover perspective, that could be the best of all three worlds.”
Thawar will be part of a panel discussion at the first-ever Cloud Computing Conferenceto be held in Toronto on Feb. 11.