The British Medical Journal (BMJ) has deployed NTT’s private cloud to test and deploy new applications it can deliver to healthcare staff around the UK.
The body still publishes its medical journal, but now also provides apps and training tools for NHS workers.
However, it was getting expensive to launch new products on BMJ’s infrastructure, because each one took a whole server to support it.
BMJ decided to take development to the cloud instead, opting for private over public because of the compliance requirements when working with the NHS.
Head of architecture Jon Peterson tells Cloud Pro: “The reason private cloud was of interest to us is that we do a significant amount of business with the NHS and they have a lot of particular requirements about how we connect to the private NHS backbone network.”
All NHS hospitals and GP surgeries are connected via the N3 private network, run by BT, meaning BMJ must deliver data down that network.
“The only way of doing that was by having BT physically connect us to that network with our own physical endpoint in the datacentre and a dedicated firewall and router for that purpose,” says Peterson.
With that in mind, BMJ looked at products from Rackspace, NTT and Amazon Web Services (AWS).
Peterson felt that both AWS and Rackspace were mainly commodity cloud services, unable to offer much to complement their Infrastructure-as-a-Service (IaaS).
He wanted additional services at the network level and at the support level, which led him to select NTT.
Its solution is a combination of private cloud and dedicated hosting in a load-balanced implementation of the Linux operating system.
Peterson says this provides BMJ with a mix of the resilience of a dedicated platform and the flexibility of cloud, meaning it can work on new apps and reconfigure its cloud environment in real time.
NTT also provides 24/7 support to BMJ, helping to solve any issues quickly.
Migration
The move to the private cloud began around four years ago and took just two months.
“During that time, applications were simply migrated from old physical servers to new virtual ones,” he says.
The hardware itself consists of four large IBM servers running VMware, providing all the virtual machines BMJ needs.
Each box sports 32 cores, and the cloud is situated in NTT’s Hemel Hempstead datacentre.
Peterson says that two-thirds of the infrastructure that runs atop the hardware is a straightforward Java stack.
“We run Tomcat, Apache and mostly bespoke code written for that platform and that covers a lot of our major products,” he explains.
“In addition to that we have quite a lot of LAMP stack software, in particular Drupal and applications based on that.
“We run Varnish as a caching layer on top of that and make use of a variety of content delivery networks, migrating to a new one at the moment.”
The only part of the infrastructure BMJ couldn’t move was some services running on two old Oracle servers running on dedicated hardware.
Peterson says: “[It’s] just to keep Oracle and their licensing rules happy at the time.”
BMJ’s long-term goal is to get rid of Oracle.
The cloud is managed using VMware management tools but NTT does a fair amount of the low-level work for BMJ, with the body giving NTT trouble tickets to deal with.
Benefits
Since moving to the cloud, BMJ has started to deploy applications using a wider range of technologies, starting to bring in Windows VMs to run applications from a company BMJ is integrating with.
Before the move, Peterson says the main cost was initially is running applications on dedicated servers and then having to commission and install new hardware.
“Even with standard hardware that's in stock at the hosting provider, that's a pretty time-consuming process,” he admits.
“You have to do a build of the machine, install our entire layer and put the applications on.”
Now, IT can create new VMs in less than a day.
“We have much better tools on our side to rapidly deploy our own layer by pointing at a particular VM and having the whole lot installed and released,” Peterson says.
“It is that ability to spin up servers and, indeed, spin them down again that's one of the major advantages."
Another benefit has been resource sharing.
“Previously if we were reaching capacity on a load-balanced pair of servers, we either had to physically install new cores or RAM, which requires downtime and leads to a chain of problems,” he explains.
“Or you have to migrate bigger dedicated servers or again buy and commission a new box and add it into the pool.”
Now, the body can just assign extra RAM on its VMs.
“That's a huge bonus, especially on applications that have burst the load requirements,” Peterson proclaims.
This process of commissioning new VMs represents an 80 per cent time saving on BMJ’s previous methods.
“It is more the agility it gives us,” adds Peterson. “Before, if someone wanted to try something out and they knew it required at least a ten-day lead time just to get some servers, that was a disincentive to doing it.
“Now if someone has an idea, we can say we'll sort that out by the end of tomorrow. We can try things.”
BMJ has now moved onto Google’s cloud for its internal office.
Sharon Cooper, CTO at BMJ, adds that with new customers outside of the NHS, the body will also take another look at AWS.
“Over the last five years our client base has become much more international,” she explains. “Our reliance on the NHS as a customer is much less now.”
