VIDaaS

The report “A sufboard for riding the wave: towards a four country action programme on research data” was published recently by the Knowledge Exchange (KE) and builds of the “Riding the wave” report. The KE is formed by partners from Denmark, the Netherlands, Germany and the UK and aims to create a layer of openly available scholarly and scientific content in which research data plays a key role. The vision set out in the document is that of a collaborative infrastructure that supports seamless access, use, re-use and trust of data.

In order to achieve the vision four key drivers are identified: incentives, training, infrastructure and funding. These four elements are an excellent framework to analyse the RDM challenge as researchers are put at the core. It´s crucial to incentivize researchers to re-use and share data through recognition, and they need to be equipped with the data skills needed in their research domain.

Other stakeholders with prominent roles include libraries, scientific organizations, funders and journals. Libraries are positioning themselves through the emerging data libraries support services to help researchers access secondary datasets, and to manage and share primary data. This will result in libraries absorbing some of the costs. National and international scientific organizations should issue rules of scientific conduct specific to data to stimulate researchers. Funding agencies need to set data management requirements as part of grant applications. Editorial boards of journals have to press authors to provide access to replication data with the articles.

The report acknowledges the existence of a diverse data infrastructure with two levels: institutional and domain specific. Data management could initially be carried at the local level (researcher, institute) and the curation at higher levels (domain archives). In spite of this there are still many “orphaned datasets” without appropriate repositories and researchers´workflows tend not to be integrated with institutional services.

The action plan outlines a range of possible actions with long term objectives for making datasharing part of the academic culture and data logistics an integral component of scientific professional life, and for a sound infrastructure operationally and financially.

It´s remarkable to see such international collaborative effort in this field; this may help to avoid reinventing the wheel, and provide more coherent frameworks to address the data management challenge.

The ‘VI’ part of the VIDaaS project is now well underway. Partnering with VMware and using loan hardware kindly supplied by Cisco and EMC (specifically, a UCS blade centre and a VNX 5100 SAN), an initial implementation of the Oxford private cloud is now complete. The virtualisation platform is VMware vSphere, with vCloud Director and vShield running on top to provide the cloud abstraction layer. In the cloud layer we are running several prototype VIDaaS VMs, based on our chosen technologies of Debian Linux, PostgreSQL and JBoss.

While development of the DaaS software continues in parallel, the virtual infrastructure team has been investigating best practice for design and implementation of a production private cloud for Oxford. We will shortly have to return the loan equipment on which we have developed the prototype cloud, but the deployment of our live environment is already underway. For this we’ve selected Dell as our hardware provider, and we’ll be using both their blade and SAN technology (the storage element being provided by Compellent Storage Center).

We’ve also been doing some investigative work into how we might move workloads between the private and public cloud. Workloads in VIDaaS are primarily based on what we are calling project nodes, which are essentially Linux VMs running Postgres and JBoss, and it is these nodes that we have been moving between cloud services. For this initial testing, public cloud facilities have been offered by Colt Technology Services. Colt is a leading cloud service provider, certified by VMware through their vCloud Datacenter Services program. Our collaboration with Colt and VMware has been very fruitful and we have successfully migrated several prototype DaaS nodes between the Oxford and Colt clouds.

We are pleased to announced that the VIDaaS’s Project’s Researcher Requirements Report is now available from our website.

The Report is the product of a requirements gathering process lasting several months, which involved interviewing researchers from a range of disciplines, and conducting a national survey. Our chief aim was to gauge interest in the Database as a Service (which – fortunately for the project – turned out to be considerable), and to establish exactly what people would like to see offered by such a service.

However, the Report also provides an interesting snapshot of academic researchers and the IT staff who support them, and of the projects they work on. For example, it became clear that collaboration is very important to many researchers – and that a substantial proportion don’t currently have access to tools that permit them to share research data with colleagues as easily as they would like.

We asked about attitudes to making research data publicly available – a real hot topic at the moment, as research councils are increasingly requiring this as a condition of funding. While researchers seem to have mixed feelings about data publication (under half were happy with the idea of making their data generally available at the end of their project), many liked the idea of having a straightforward way of putting a particular subset of data on the Web to accompany publications such as journal articles – particularly if the dataset had a persistent URL or DOI that would allow it to be cited.

Evidence from elsewhere suggests researchers have good reason to be interested in this possibility: a presentation by Kevin Ashley at the recent DCC Roadshow in Oxford reported a study indicating that papers for which accompanying data was available were cited more than twice as often as those with no data available.

We also made some interesting discoveries about favoured software and data formats. Spreadsheets and statistical analysis packages were both common – the latter particularly among social scientists. Relational databases were also widely used, though it was noticeable that IT support staff were more than twice as likely to report relational database use as researchers were, perhaps indicating that academics find this method of managing data most useful when they have ample technical support available.

One slightly surprising finding was the prevalence of use of XML documents, particularly among humanities researchers: almost two thirds of this group make some use of them, and nearly a quarter said this was their chief method of handling structured data. On the other hand, document-oriented databases do not yet seem to have achieved the same level of popularity, with a quarter of survey respondents revealing that they weren’t even sure what these were.

All these desires and preferences (and many more that there isn’t space to talk about here – see the full report for details) were taken into account in compiling the prioritized technical requirements list described in an earlier blog post. This will guide the work done by our technical team over the next few months.