Reading University builds AWS-style research cloud based on Nutanix

The University of Reading has found a novel way to increase its income and rankings: build its own research public cloud.

The idea behind the new technology came from wanting to open up computing power to all academics and departments by making it cheaper, and also to free up time for the IT team, according to Ryan Kennedy, Academic Computing Team Manager at The University of Reading.

Kennedy explained that IT staff were spending the majority of their time maintaining and supporting a mix of five different compute and 10 storage platforms, and so needed a way to consolidate and simplify the infrastructure.

After realising that moving everything to the cloud would be too expensive, the computing team decided to build its own on-premise Reading Research Cloud, sat on Dell EMC XC Series appliances running the Nutanix Enterprise Cloud Platform.

Kennedy said that the Research Cloud is the university’s take on a public cloud like AWS, but run out of its own data centre. The team wanted to be able to deliver the speed and agility of the public cloud but with more control than is offered by available versions from firms like Amazon and Microsoft.

Before signing up with Nutanix, Reading did some testing on Microsoft Azure. Kennedy said the university spent a fortune on the public cloud system but couldn’t get the control it wanted over what users were doing. With the Nutanix Self Service Portal, users have flexibility, the university gets control and the costs were about 60 percent cheaper than compared to the Azure cloud.

Normally for new IT projects, it is the vendor that has to persuade the customer to seal the deal; not in this case. Kennedy said it was Nutanix that was uncertain, and he had to borrow some kit from them to prove it would work.

The main challenge is, we run just about every environment you can think of. I can remember the first meeting we had with Nutanix and they all looked puzzled and panicky that it couldn’t work. It can – it’s quite happy at doing multiple mixed workloads independently of each other but not interfering with each other, which is a big benefit for me of what I have to manage.

Some of the requirements of a university are quite unique. I work solely in the research sector, and the workloads that I run, especially those on Nutanix, are all different and are very random and very ad hoc. I would have an academic that might turn around to me and say, ‘For the next five days I need 600 cores, what can you do?’ The benefit of Nutanix is, I could buy new nodes and have them up and running in a matter of hours if not days.

Eco-friendly

The Research Cloud has 18 Nutanix nodes in all, 10 for storage and eight compute-intensive nodes, and the storage and compute environment is now managed as a single entity. As well as ditching its old Red Hat ClusterFS storage system, the university also switched from VMware to the Nutanix integrated AHV hypervisor for virtual machine resources.

Kennedy said the biggest problem with the Red Hat storage product was around scalability. With the old system, it took two to three weeks of work to put in a bigger node and transfer the data across; by comparison, Nutanix nodes are plug and play to get an extra 60TB of capacity, he explained.

Being more eco-friendly was also an objective of the project: the cloud system has allowed the university to go down from five to two half-filled racks, meaning power consumption is tiny compared to what it was before. Kennedy said:

That’s always a big tick for us as we try to be as green as possible.

Also to be able to offer a service to everybody in the university, research wise. One of the things we found from the business side is, if you were in cash-rich departments like meteorology they had their own cluster and could do it. But the departments that didn’t have the big income just couldn’t afford to go out and buy their own cluster. I wanted to offer something that everybody could use because you only need to buy that little bit of it.

Reading started initial discussions with Nutanix just over a year ago, and rolled out a four-month proof of concept in collaboration with Dell. Then from January to May, the project effectively went on hold:

We had the usual public sector issues of getting funding.

During this time, the university used Nutanix’s Community Edition to build small clusters and hone the system to its exact requirements. By June, it had funding and the official cluster environment was in place.

To get the actual funding wasn’t complex once we actually put together the business case and showed how easy this tech would be. We didn’t need more operational money for staff because it was so easy to manage.

Usage

Since the Research Cloud’s launch, uptake has been “massive”, according to Kennedy. Three options are available to academics: a quota of CPU, memory and storage that the user is free to do whatever they wish with, plus a limited support model; a fully managed service where they get a VM for their own use; and a scheduled service, where they can make use of a fully managed box, but cannot have their own software installed onto it. These different options were crucial to meet the requirements of the different academic types, Kennedy said:

We’ve got some very intelligent people, computer scientists, meteorologists that know what they’re doing; we have the other academics that believe they know what they’re doing; and then you have the ones who say we don’t know what we’re doing. We offer a range of services. Everything goes through central IT, they just come and say we want some compute.

Quite often in higher education, once one person does it, everyone else is watching and then they follow behind. Higher Education is always quite ‘we’ve always done it this way, so we’re always going to do it this way’. I’m one of those people that’s happy to take the gamble. With research data to an extent you can take more of a gamble than with HR and finance data.

Next steps

Thanks to the success of the project, the university is already starting to plan for the next step, which will be deploying Nutanix Calm to provide a hybrid cloud. This will enable the Research Cloud to scale using resources from Amazon, Azure and Google public cloud platforms.

So if an academic needs 100 cores tomorrow, with all the will in the world, Dell won’t be able to deliver that in time. That way, we can burst into the public cloud, or use Amazon spot sales. It will be called Research Cloud Self-Service Portal 2.0.”
Ultimately, Kennedy believes the Research Cloud will increase the status and earnings of Reading University.

To get a research grant, you usually have to have done some research beforehand to prove that the concept works. We’re able to now offer a platform, where we can say for the next two weeks, you can use it for free, get your data that you need to prove it, then they can go to the research councils with the evidence to get the research income.

That will allow more and more income into the university and help our rankings, which is the ultimate goal of the university.

Kennedy also had a warning for other universities yet to embark on or finish their own digital transformation projects. Reading completed its own comprehensive three-year overhaul two years ago, where external consultants came in and gave advice on every area, including installing a brand new finance and HR system. But research was completely forgotten.

It was a big thing for the university, but nobody really knew why we were doing it but everyone was doing it. It was rushed. We got asked as management in IT to write down all the things that we wanted to do, and then we were given six months to deliver them all. IT projects never run that smoothly.

Now, two or three years after the big transformation was done, I’ve put my hand up and said can I have some money please to do it now, rather than when there was the frankly millions of pounds being spent on every widget you could imagine being installed everywhere.

Image credit – Image sourced via author