Thursday, 3 July 2014

RDM Technical Infrastructure Review - Introduction

1. Introduction

The purpose of this report is to indicate options available for the development of a technical infrastructure to support research data management (RDM) at the University of Sheffield. RDM, its situation within academic research and recent drivers towards change are defined. The processes involved in RDM and the elements of the supporting technical infrastructure are examined. The local context, of RDM technical infrastructure at the University of Sheffield and collaborating institutions, is explored.  The range of technical infrastructure components available and evaluations of these components are reviewed. Instances of fully-functioning RDM technical infrastructure and many of the recent research projects that developed and piloted RDM technical infrastructure components are briefly described. Finally, recommendations for suitable technical infrastructure components are proposed.

1.1.   Research Data Management

The research data collected to test a research assertion must be managed in an appropriate manner to be considered good research practice. Good RDM practice is now required of researchers by many research institutions and by most research funders. Increasingly research funders are demanding long-term curation of some of the data resulting from the research they fund, so that they may be available for re-use. The value of those data and the impact of the original research are increased by re-use. RDM may be considered to involve three broad areas of activity:-
  • Data management planning, during the research proposal and grant application stage.
  • Looking after ‘live’ or ‘active’ data as they are collected, processed, shared and stored.
  • Data Stewardship - Long-term curation of research data and data publishing, making data discoverable and reusable.

1.2.   Research data management drivers 
JISC[i] have supported the development of RDM practice over the last fourteen years by funding projects involving HEIs through a number of programmes, in particular the Managing Research Data Programmes 2009-11[ii] and 2011-13[iii]. The DCC[iv] was established in 2004 with JISC funding, to support expertise and practice in RDM. Since 2011 the DCC have offered tailored support in the development of policy, services and infrastructure for Higher Education Institutions, and are the foremost source for information and advice in the development of RDM infrastructure (Jones et al. 2013). Infrastructure refers to the hardware, software and human resources necessary to support the RDM services and processes. This report focuses on the technical infrastructure, the software and hardware components available.


The need for good RDM practice is recognised by all stakeholders involved in the research process. These include: 
  • Researchers, who may be part of a research project team, which may include members of many different institutions. Researchers need to secure their data against loss or unauthorised access. Making data available to reuse allows verification, promotes integrity and increases research impact.
  • The research institution (may be a HEI) or body employing the researcher and providing the facilities. Research data may be considered part of an institution’s special collections. Institutions will also wish to minimise risk to the data and damage to their reputation.
  • The research funder (usually a research council, charity or a HEI) who may mandate RDM procedures such as the creation of a Data Management Plan (DMP) and the deposit of data to repositories. Research Funders may support facilities for data curation such as data centres.  Funders wish to increase the return on their funding, through the reuse of data.
  • Governments, who fund research councils and other funding bodies, are concerned to derive as much value as possible from publicly funded research.
  • Publishers of research papers, who may publish the underlying data, seeking to add value to the publication process. 
The EPSRC policy framework on research data[v], published in May 2011, puts forward the EPSRC expectations[vi] of organisations receiving EPSRC funding, concerning the management and provision of access to EPSRC funded research data. These nine expectations were developed from seven guiding principles[vii] which are aligned with the RCUK common principles on data policy[viii]. Institutitions in receipt of EPSRC funding are expected to be fully compliant with these expectations by 1st May 2015. In terms of RDM Infrastructure, the pertinent expectations (EPSRC, 2013) are:

“Research organisations will ensure that appropriately structured metadata describing the research data they hold is published (normally within 12 months of the data being generated) and made freely accessible on the internet; in each case the metadata must be sufficient to allow others to understand what research data exists, why, when and how it was generated, and how to access it”

“Research organisations will ensure that EPSRC-funded research data is securely preserved for a minimum of 10-years from the date that any researcher ‘privileged access’ period expires or, if others have accessed the data, from last date on which access to the data was requested by a third party;”

“Research organisations will ensure that effective data curation is provided throughout the full data lifecycle... The full range of responsibilities associated with data curation over the data lifecycle will be clearly allocated within the research organisation, and where research data is subject to restricted access the research organisation will implement and manage appropriate security controls;”
The University of Sheffield Research Data Management Policy[ix] was developed in response to the RCUK principles and EPSRC expectations. Of the eight points of policy, the following (The University of Sheffield, Research and Innovation Services, 2014) are particularly relevant for RDM Infrastructure:

The primary responsibility for effective research data management during the course of research projects lies with lead researchers. However, all researchers, including postgraduate and undergraduate students undertaking research, have a personal responsibility to manage effectively the data they create.”

“Unless the terms of research grants or contracts provide otherwise, data generated by research projects are the property of the University of Sheffield. Researchers should exercise care in assigning rights in data to publishers or other external agencies.”

The University will provide support for research data management, including… ...Additional infrastructure and services for research data management, to be developed in consultation with researchers.”

The research institution is the body responsible for providing the researcher with facilities for research and is therefore responsible for providing the researcher with the necessary infrastructure and services to support RDM. Design of this infrastructure must be based upon the researcher workflow, so as not to burden the researcher with additional work or by changing their practices, and where possible, making RDM processes virtually automatic and invisible to the researcher.

1.3.   Research data lifecycle

Data collected during a research project will include the research data themselves; experimental, observational, modelled data etc. together with the metadata that describes these data in detail and documentation describing the context of the research, details of the research project and the processes involved. Data here will be defined as the numerical and textual information collected by analysis or measurement from the research samples or objects – but not the samples or objects themselves. For example, a collection of images or of tissue samples will not be considered data until there is textual or numerical information, such as identifiers (names or ID numbers), descriptions and relationships, associated with them. In this document, data refers to digital data, although the same management principles apply to analogue data formats. However, it is best practice to digitise research data, making its discovery and reuse easier.

From the initial drawing-up of a research proposal and grant application, data will be collected and managed. This includes documentation about the project, people and bodies involved, grant application, data management plans, experimental protocols, possibly test data or data collected from previous projects for re-use and a literature review or bibliography.

Once underway, a research project will collect or create raw data, which, during the project, will usually be processed to create derived or processed data. There may be many different iterations of processing, resulting in many sets of derived data. Eventually a set of ‘results’ data will be selected as the basis of the research publication(s) output by the project. All these sets of data can be considered active data, which will need to be quickly accessible and easily shared between collaborators. All these sets of active data will need appropriate documentation to describe the processes involved in their creation and modification.

After the project has finished, researchers will need to select data for curation on a long-term basis. This may have been decided in agreement with the research funder during the initial planning stage of the project. Curation in the context of RDM, refers to archiving, preservation and adding value through transformation and reuse. These archive data selected for curation, may need further processing (validation, cleaning, anonymisation or redaction) before submitting to an appropriate repository. The associated metadata will be needed to provide the necessary information for citation and re-use. There may be the facility to add new metadata or documentation, generated by data reuse, to the curated dataset. Data not required for curation needs to be disposed of in an appropriate manner.

1.4.   Data documentation, metadata and data collections

Data need to be documented to be understood and managed. Data documentation indicates the conditions and processes involved in the creation or collection of the data, the processing of the data and the context of the research. Detailed documentation is essential for verification and reuse. Adequate data documentation is necessary to determine provenance, licensing and access arrangements and preservation requirements. Research data need to be documented at three levels:
  • Project level – providing an overview of the research context and design.
  • File level – describe the relationships between files or database tables.
  • Item level – describing, for example, the meaning of a variable in a table.                                             (Research Data Mantra, 2014UKDA, 2014)
Metadata are a highly structured subset of core data documentation. Metadata are structured so that they may be indexed and stored within a database, thereby facilitating data organisation and discovery, and machine to machine interoperability. By considering its function, metadata may be divided into three layers:
  • Core metadata (Datacite, 2011, p. 8) or Catalogue metadata - creator name, publisher, title and an identifier are required for discovery and correct citation of the data. This could possibly include some subject description or classification details.
  • Detail metadata or Administrative metadata – provides generic dataset description. This includes access, preservation and technical metadata, and is required for the long-term curation of the data. This will include more detailed classification / subject description.
  • Discipline specific metadata (also known as Reuse metadata) – This documents aspects of the dataset that will be of interest to researchers wishing to validate the research process or re-use the data. This will consist of experimental protocols, instrument settings, and relationships with other elements of the dataset, other files within a data collection or other data collections. This will provide very detailed classification / subject description, providing the fine-grained attributes of data necessary for accurate discovery and location of elements within a dataset. Discipline specific information is frequently held in unstructured formats, so could be considered data documentation rather than metadata.                                                                                                                             (Ensom, 2013 and IDMB, 2011)

Data collections are typically organised by reference to a particular survey or research topic and may cover a specific geographic area and time period. The UKDA defines a data collection as typically comprised of three components: data, documentation and metadata. Code is occasionally considered a fourth component (Ensom and Corti, 2012, p. 3).

1.5.   Data repository or Data registry?
A repository is a content management system which may be considered to consist of three elements – a user interface front-end, and a database layer and storage layer back-end. The database holds records of entities, consisting of metadata elements as a series of fields. The storage layer, containing the actual data bitstreams, may be a file system on the repository server, or a file system on a local server or a remote server that is independent of the repository system. This may include cloud storage or a hybrid storage system. Usually in a repository system, only metadata records are held within the database, not the data objects themselves. This is due to the larger size of data objects which results in slower indexing / access speeds. Storage designation is handled by a storage controller or storage resource broker.

Different repository systems may be configured for different organisational structures. Repositories may range in the granularity of data described: The entity described by a repository record, the data object, may be a single row within a database or spreadsheet, a single file, or a collection of interrelated files constituting a dataset. In the Essex ePrints [6.1.4][x] context, the ‘eprint’, the key entity, is the ‘data collection’ which consists of a set of metadata and files (Ensom and Wolton, 2012). Datasets may be grouped as collections or groups, the ‘User’ entities may be grouped as a ‘Community’.

A ‘Data registry’, or ‘Data catalogue’ or ‘Metadata store’, is a repository system that holds only metadata records. The data themselves are held on a local or remote file storage system, so the metadata record points to the data store filepath or URL. By using the appropriate metadata standards, metadata records may be exchanged between registries, giving rise to the possibility of national and international registries. Repository systems were originally designed to curate textual digital objects, but are now being modified to curate digital objects of all formats and sizes, with the view of extending their purpose to curate research data (Gutteridge, 2010). As they have been designed to manage, curate and publish research outputs, they may perhaps provide the ideal platform for a catalogue for the data underlying the research outputs.

The ANDS[xi] strategy has been to separate the data storage function from the cataloguing function, the dissemination function and access control are provided by the metadata store (ANDS, 2011b).

1.6.   Research data ecology
In considering the implementation of an Institutional RDM technical infrastructure, it can be useful to consider the ecological approach (Robertson et al. 2008), to gain a better understanding of the interactions between repositories and services. The local infrastructure does not exist in a vacuum and must interact with, and is dependent on, a diverse range of entities and processes in the information ecosystem.

In determining the place of a repository, registry or other infrastructure component in the overall information ecosystem, it may be helpful to identify a range of components by considering their data storage coverage and specialisation (ANDS, 2011a), which may be:
·         Local – personal, project or departmental server for active data storage.
·         Storage associated with an instrument or facility.
·         Institutional storage for active data – networked filestores.
·         Institutional repository for archive data.
·         Multi-institutional project data storage (CARMEN [6.2.5] for example).
·         Research council data centre – for archive data including longitudinal data.
·         Discipline based repository – for active and archive data. National or International coverage.

Metadata storage will have a similar range (ANDS, 2011b), which may be:
  • Local, project and instrument based metadata store – spreadsheets or databases associated with the data.
  • Institutional data registry or repository.
  • National data registry (ANDS).
  • Discipline based metadata store – may be international, national or multi-institution based.

A number of practitioners have suggested that, regarding research data curation, “the Institutional repository is the repository of last resort” (Haywood, 2013), since discipline based repositories are better configured for the types of data and specialised metadata formats associated with the research community they serve. However the importance of the institutional research data services (tier 3) in the hierarchy of rising value and permanence (Figure 1. below)  is emphasised in the Royal Society report ‘Science as an open enterprise’ (2012).



Figure 1. The data pyramid - a hierarchy of rising value and permanence (Royal Society, 2012)

This is reiterated by Simon Hodson (2012), who maintains that Institutional research data services are essential because:
  • The institution is where the data are created and can be captured. Institutions implementing RDM infrastructure will make data discovery and curation possible.
  • Joining the gulf between curated data in national / international data services and uncurated and inaccessible data in individual or project collections.
  • Elevating data to national / international data services from temporary and inaccessible individual collections. Important data collections may emerge as they become discoverable.

The data catalogue component of an institutional infrastructure would ideally adhere to the formats and schemas used by a national data registry under development. The existence of tools for interoperability and for deposit to the major data centres should also be a consideration in the selection of a repository system.

1.7.   Development of RDM services

The DCC have created a guide to developing RDM services at HEIs (Jones et al. 2013), which breaks down the development, process into a number of components, as visualised in Figure 2. 


Figure 2. The components of an RDM service as envisaged by the DCC (Jones et al. 2013, p. 5)

The approach to the development of RDM Services for HEIs, recommended by the DCC involves:
  • Assembling a steering group composed of senior representatives of stakeholder group – senior researchers and institutional support service managers.
  • Appointing an RDM service development group to undertake the work.
  • Carrying out a gap analysis, to determine gaps between current position and the aimed-at future position, and requirements gathering surveys to determine stakeholders needs.
  • Development of RDM policy and strategy. Developing a policy first may be useful as a motivating factor, but may lead to problems if the proposed infrastructure and services cannot be realised. Alternatively, the development of policy may be subsequent to the defining of strategy. 
  • Designing services to meet local and external needs – putting in place the infrastructure required to support these services.
  • Piloting these services to test that they are fit for purpose. 
The DCC have published a case study detailing the development and implementation of an RDM strategy (Rans & Jones, 2013). Design of the technical infrastructure required to support the RDM Service is considered below.




[i] Joint Information Services Council (JISC) http://www.jisc.ac.uk/
[iv] Digital Curation Centre (DCC) http://www.dcc.ac.uk/
[v] Engineering and Physical Sciences Research Council (EPSRC) policy framework on research data http://www.epsrc.ac.uk/about/standards/researchdata/Pages/policyframework.aspx
[viii] Research Councils UK (RCUK) common principles on data policy http://www.rcuk.ac.uk/research/datapolicy/
[ix]The University of Sheffield Research Data Management Policy  http://www.shef.ac.uk/ris/other/gov-ethics/grippolicy/practices/all/rdmpolicy
[x] Essex Research Data http://researchdata.essex.ac.uk/ [see section 6.1.4 for information]
[xi] Australian National Data Service (ANDS) http://ands.org.au

No comments:

Post a Comment