When leading online job board operator Totaljobs Group decided to shift its customer-facing systems to the public cloud in April 2012, following its acquisition by European jobs firm StepStone, the move quickly paid off.
As Computer Weekly reported in the original case study in February 2012 – Totaljobs hires public cloud services for agility and better performance – the business cut annual IT and related costs by $500,000. At the same time it improved availability and agility. For example, the time taken from approving a business case to deploying servers plummeted from three months to three days.
With such impressive statistics, it is little surprise the project scooped this year’s Computer Weekly European Datacentre award for cloud innovation. The judges commended the company’s IT team for the scope of its ambition and its impeccable execution of the move to AWS.
The acquisition meant Totaljobs Group had to vacate the datacentres of its former owner, Reed Business Information, by the end of 2012. It completed the migration two months before this deadline.
It's not a problem if you're a greenfield organisation building new applications, but it's a lot harder to hit 12-year-old systems with a hammer until they're amenable
Edward Vassie, database architect, Totaljobs Group
As well as moving to Amazon Web Services (AWS), the work included upgrading the database environment to SQL Server 2012, introducing peer-to-peer replication and standardising on Windows Server 2008 R2, as well as putting in place new email, backup and load-balancing systems.
Since fully going live on Amazon’s public cloud infrastructure in November 2012, Totaljobs has worked hard to ensure the initial benefits it saw can be maintained and extended. In the process it learned a thing or two, says database architect Edward Vassie.
“We definitely made the right decision moving to AWS, but due to our inexperience with the public cloud and the short timescales of the project, some of our original decisions weren’t the best they could have been,” Vassie says.
For instance, the firm has since realised that, as far as virtual machine rental goes, bigger is not always better.
“We’ve found that, for a lot of our workloads, using smaller machines to scale out is more effective than our original decision to scale up. That’s given us greater cost control,” he says.
“We’ve also been running a bit more leanly, as well as doing other minor optimisations here and there. For example, we’ve been turning off machines during the day if we know they’re only used at night.”
But he says because the company had to move a substantial legacy infrastucture onto AWS in such a short timescale, it was not practical to do so in a way that would easily allow it to take full advantage of AWS’s capabilities.
“Take the ability to automate startup and shutdown of machines based on actual rather than perceived demand. The investment involved in getting that to work is more or less the same as what we’d have saved by doing it,” he says.
“It’s not a problem if you’re a greenfield organisation building new applications, but it’s a lot harder to hit 12-year-old systems with a hammer until they’re amenable.”
Undaunted, the Totaljobs IT team has nonetheless been making plenty of smaller but significant improvements to the system.
“For example, at the technical level we’ve decided to move away from striping EBS volumes to having a one-to-one correlation between an EBS volume and a Windows drive letter, which really improves manageability,” Vassie says.
More importantly, the additional savings achieved are being ploughed back into new IT initiatives that more visibly improve the business.
For example, the company has been trialling a new search mechanism which gives recruiters and candidates more relevant results to their queries. “As a result, we’ve seen around a 5% improvement in job applications,” says Vassie.
More case studies from Computer Weekly
He adds the company is also looking at Amazon’s cloud-based data warehouse solution RedShift for its business intelligence capabilities, and is relieved Microsoft has not made it hard for the system to interoperate effectively with SQL Server Enterprise Edition.
But he is keenly aware the threat of proprietary lock-in still looms large.
“The cloud today is a lot like electricity at the beginning of the last century. Back then, the great variety of voltages and cycle times seemed good to providers because they could lock people into their systems. But the mass market for public electricity only took off following standardisation,” says Vassie.
"Similarly, the public cloud market will only mushroom when suppliers realise they can get a far greater increase in revenues by standardising systems.
“I hope that happens soon, but I wouldn’t be surprised if we have to wait another five years.”