Keeping IT in-house: hyperconverged infrastructure at Chesterfield Royal Hospital

IT technical delivery lead David Sawyer on the benefits of infrastructure overhaul and why there's no hurry to move to cloud

David Sawyer, IT technical delivery lead at Chesterfield Royal Hospital NHS Foundation Trust, is in no hurry to move to the cloud. Not that it's no, nay, never, exactly, but he says his team is waiting till the time is right.

"We are holding back. We have some ideas about using cloud backup and recovery but for now we prefer to keep everything internal."

Most of the Trust's applications don't currently require the cloud's elasticity and reach (and of course there is always the NHS's own N3 and Spine for those that do) and using cloud would just add to the complexity. Nevertheless, it could potentially ease the pressure on the server room and support staff.

Sawyer's team supports around 250 applications, ranging from specialised clinical software in areas such as oncology and endoscopy to patient administration systems (PAS) and back-office systems for financials, HR and estate management. These are used across 60 departments and 25 hospital wards.

Like most organisations of its size, Chesterfield Royal Hospital virtualised its infrastructure some time ago to make it more efficient. However, the original 50 virtual machines had mushroomed to almost 400 and the server room was running out of rack space - a problem as they needed to expand to meet rapidly increasing demand. Performance and reliability issues were also starting to emerge.

Sawyer's team looked to optimise the performance, improve deployment flexibility, cut costs, save space, and make the whole thing easier to manage.

Instead of shifting applications and data to the cloud, they looked to upgrade their internal structure by adopting a hyperconverged appliance-based solution from Nutanix. The applications have all been moved to Nutanix apart from two which rely on proprietary operating systems on which the Nutanix tools could not be installed. Racks of servers and SAN storage units have been replaced with 24 nodes of Nutanix.

"We been able to retire about 60 per cent of our hardware," Sawyer said. "We now have four dedicated cabinets of Nutanix and we're not going to struggle for physical space as we expand the clusters."

In terms of read-write performance his team noticed a big improvement.

"Latency is surprisingly less. We came from a traditional three-tier system where you have storage and compute separated by a network. We compared the write speed of a server that had a direct SAS connection to storage and the disk speed was unbelievably fast on the Nutanix virtual environment versus this physical server with direct access to storage."

They also experimented with iSCSI over a VLAN to temporary storage on the other side of the hospital. "We were expecting some sort of performance hit, but we heard from our user base that it was pretty quick," said Sawyer.

Operationally, the new infrastructure has enabled the team to rationalise the way applications are supported. Previously, core applications ran on VMware ESX, with lots of others on standalone virtual machines running the free version of the hypervisor which has limitations on functionality. This was unsatisfactory, being hard to manage difficult to expand, said Sawyer.

"It was lots of bolt-ons rather than being the result of proper thought process."

The cost savings come mostly from the cost of VMware licensing, since the Nutanix Acropolis hypervisor is included in the price of a node. The team is gradually phasing out its use of VMware.

"In the short term we save money because we don't need to buy new ESX licences; in the medium term we're saving money because we won't be renewing ESX licences; and in the long term when we expand the cluster even more or build another cluster somewhere we have no extra cost associated with it," he said.

And when the time is right, Sawyer is hopeful that the hyperconverged infrastructure will ease the journey to the cloud.

"In the long term will keep looking at cloud services for a backup and recovery and testing. Because of the near seamless synchronisation between on-prem and cloud that could be the first step for us in the cloud and that's what we want to test in the near future."