Communications of the ACM
Managing Scientific Data
Disk with silicon sensors as an endcap of the ATLAS silicon strip detector in its testbox at the Nationaal Instituut voor Subatomaire Fysica, Amsterdam, The Netherlands.
Credit: Ginter, P / CERN Photo #: INDET-2006-002
Needed are generic, rather than one-off, DBMS solutions automating storage and analysis of data from scientific collaborations.
The full text of this article is premium content
Comments
CACM Administrator
The following letter was published in the Letters to the Editor in the September 2010 CACM (http://cacm.acm.org/magazines/2010/9/98018).
--CACM Administrator
The article "Managing Scientific Data" by Anastasia Ailamaki, Verena Kantere, and Debabrata Dash (June 2010) explained that data generated by research projects is valuable only when annotated with metadata describing the data's provenance, context, and meaning. However, a given data item can be annotated in more ways than one, for two reasons:
Provenance. A multidisciplinary project can track its progress with basic metadata indicating the provenance of its samples and their associated data. Each data item can also be annotated in a more detailed way through tools particular to the technique used to generate the data item; these annotations are themselves interpretable by people (and software) in the relevant discipline; and
Assumptions. By definition, a research field involves a basic set of concepts used to understand the field but that is not yet agreed upon. Annotations beyond where and when the data was recorded incorporate assumptions that may be contentious among experts.
Data storage and metadata should thus be decoupled. A data repository must be capable of returning any data item stored within it, along with a list of places needed to find the relevant metadata. A metadata repository must be capable of identifying the schema it adheres to and respond to queries about specific data items with relevant annotations.
Decoupling the architecture this way eases develop of an ecosystem of repositories and annotation schemas.
Chris Morris
Warrington, U.K.
--------------------------------------------------
AUTHORS' RESPONSE
Separate (and multiple) metadata stores are indeed essential for a number of scientific applications and should be available to user scientists as an option. However, because data queries likely need to combine information held in separate metadata stores, processing them requires appropriate mechanisms for distribution, access control, and the merging and branching of the stores.
Anastasia Ailamaki, Verena Kantere, and Debabrata Dash
Lausanne, Switzerland
Displaying 1 comment