University College London (UCL) is to build a data storage and private cloud infrastructure to house research generated by its over 3,000 researchers and collaborators.
The facility, which is being built on WOS object storage and GRIDScaler hardware from DirectData Networks (DDN), will initially have the capacity to store up to 600TB of research data, although this will scale up.
Dr J Max Wilkinson, head of research data services at UCL, who is leading the project, told Cloud Pro: “The research data that is generated here is a valuable asset ... and we see it as our responsibility in supporting our researchers in managing these data."
One of the reasons for choosing to build a private cloud was due to the type of data being produced, not just its quantity.
“Research data is raw, it is unstructured, it passes across all domains and format types, it can be unpredictable and it can belong to virtually any field of study,” he said.
“This makes it different from business ‘big data’, which is comparatively well typed and predicted” he added.
The second reason, Wilkinson said, was to make sure the data remains available to researchers both inside and outside UCL on an ongoing basis.
“All around the world, people are starting to recognise the value that is being lost because we are not able to manage data the way we used to,” Wilkinson said, claiming that a particular problem with modern data is not just the variety of outputs from different machines, but the obsolescence of digital media on which results are currently stored.
“When our data consisted of observations or measurements that were recorded in lab books, it was really easy to share that information and preserve it.
“Since the introduction of computers, data has been written to all kinds of disk formats, like USBs or CD-ROMs or 3.5” diskettes and that data is effectively lost because we eventually lose the capacity to read some of those media types. So we need to put the data onto an environment where they can be looked after better,” he added said.
While the project is starting as a 600TB single-site data centre, Wilkinson anticipates this will grow, which is why a scalable platform was chosen.
“The history of IT is littered with assertions that ‘this is going to be more than enough than anyone is ever going to need’, so we have pitched high. We are not going to put a 100PB data centre in UCL’s Bloomsbury campus – there’s simply not room, nor is there the energy supply. But I do foresee us, at some point in the future, having a 100PB distributed system which could incorporate an element of shared services, as well as a variety of other media in addition to disk, such as large tape libraries,” he concluded.
