fotohansel - Fotolia
There is nothing like a deadline to focus the mind on the job in hand, and that is certainly true for utility company ScottishPower, and its recently completed datacentre migration.
With the firm keen to consolidate the number of offices it operates in Scotland, it announced plans in 2012 to sell-off a number of sites across the country, including its headquarters in Cathcart, Glasgow, the location of its primary datacentre.
The project would eventually culminate in the opening of its new HQ in Glasgow in July 2017, where 700m2 of space had been set aside for its new datacentre.
“We were headquartered on the south side of Glasgow in Cathcart for about 45 years, so this was a big move for ScottishPower,” says Allan Ferguson, head of UK infrastructure, operations and communications at ScottishPower.
“Moving the datacentre was going to be one of the biggest challenges, because it had grown organically over a long period of time – year on year, project by project – and you can just imagine the technologies that have come and gone over that time.”
A direct lift and shift of its datacentre equipment from one site to the other was out of the question, because the site it was moving to was markedly smaller, prompting the firm to embark on a pre-move period of rationalisation and consolidation of its server halls.
“The old datacentre is about 1,000m2 and, at its peak, hosted about 1,100 servers,” said Damon Kirk, ScottishPower’s datacentre programme migration manager. “We knew the move was coming, so we looked at trying to virtualise, consolidate and decommission, and managed to get that number down to around 700 or so servers in Cathcart.
“We decommissioned about 130 racks between 2013 and last year containing legacy equipment, where half a rack was filled with just one server. That may have been state-of-the-art equipment 15 years ago, but the processing power on that could be covered by a 2U server now.”
Cracking the Cathcart constraints
Although Cathcart had more datacentre space, the company had reached the limit of what it could do with the site, as its design made it difficult to deploy higher-density servers because of power constraints.
“We’d maxed out the power, which I know sounds funny given the business we’re in, but we couldn’t really expand quickly without investing a couple of million pounds in the site,” said Kirk.
“With new, power-hungry technology coming down the line from the likes of Cisco, Dell or IBM, with 25-40kW racks, we needed to look at the way we designed the new datacentre to create high-density pods that we could build into cloud, hybrid clouds and hyper-converged infrastructure.”
For this reason, the completion of the datacentre migration should pave the way for the company to move into the world of hybrid cloud, once senior management concerns about using off-premise resources are addressed.
“A lot of their concerns are ones organisations commonly come up against when you talk about cloud,” said Ferguson. “For instance, is it a safe place to store customer data? So we are trying to give them that on-premise option too.
“We’re taking a bit of time to educate the business and give them comfort that we’ve covered all these angles for them.”
A shift in cloud mindset
Helping the business get to grips with how to finance cloud deployments is another challenge the company is working through, as IT purchases have historically been classed as capital expenditures (capex).
“That’s the biggest challenge – trying to get them to understand the shift from the capex budgeting position that we’ve traditionally come from, where we buy the asset and gradually sweat it over time, to an operational opex budget, where we pay for what we need when we need it,” said Ferguson.
The migration part of the project was two years in the planning, with the company drawing on the support of its long-standing IT provider Systal Technology Solutions, which had previously worked with ScottishPower during its server consolidation push.
“They were involved in a lot of the pre-work, carrying out datacentre audits to help us shape a picture of our estate, providing resources as and when required in the various technology domains we needed to help us produce run books,” said Kirk.
Two years may sound like a long time, but the team could not afford to leave any part of the migration to chance.
“What was unique about this project was that it had the potential to impact every single business unit that ScottishPower had, because we were moving trading systems, health and safety systems, billing systems and our call centre infrastructure,” said Ferguson.
“The business was very keen, for that reason, to be involved and assured through the planning process that we were covering ourselves off against any potential for error.”
The company relied on planned outages to give it time to move the infrastructure over, and had to factor in that some of these might go on longer than expected, should the team run into technical difficulties.
Read more about datacentre migrations
- Paul Woobey, IT director at the Wellcome Trust Sanger Institute, tells Computer Weekly why the organisation’s HPC workload requirements cannot be fulfilled by a move to the cloud.
- Irish telco Eir shares details of its datacentre consolidation project and explains why investing in a robust UPS was a must.
Despite the sheer size of the task Kirk and his team were facing, there were only two occasions when the company had to call a halt to the migration and roll back some of the changes it had made – which was not bad going, said Kirk.
“There were a couple of rollbacks to do with the trading systems, because we had exhausted all the technical options available to us, so we reached out to our suppliers and just had to make sure the business was back online while we worked through the issue,” he said.
“We then re-planned the move for two or three weeks later, once we had further analysed what the technical problems were, because we just had to make sure all the way through the project that the business came first and were always online when they needed to be.”
The amount of planning that went into the project is something Kirk considers to be critically important to its success. That and the fact that his team had an immovable deadline to hit, given that the original site was being sold.
“On other projects, you don’t often have that finite end date, so maybe there isn’t that same impetus because the clock isn’t ticking so loudly,” he said.
“The line of sight was clear and we spent a great deal of time in the planning process, so we knew what we had to move and where we had to move it from and to, and had agreed the outages, so that kept the programme really tight and the team focused.”
The datacentre move was completed on 22 September 2017, and the next item on Kirk and Ferguson’s agenda is to refurbish its failover site to ensure it is brought up to the same standard.
“It is not as old as Cathcart, but it’s not far off and it has its own challenges with space and power, and that’s going to be another challenge to re-engineer it while in flight, but the goal is to make it as shiny and new as our new HQ,” said Kirk.
Read more on Server virtualisation platforms and management
Emergency services organisations ranked slowest public sector cloud adopters in Eduserv poll
How Forest Holidays migrated its hosted website to a managed service
Goonhilly Earth Station sets out plans to open green colocation datacentre in 2018
Agnes Kirk on the role of CISO, Washington's state of mind