Andrew Grimshaw on 2010-03-20

• Propose requirement with title  'Data is equally important as compute'

Etienne Urbah's position on 2010-04-07

• In fact, Computer science has progressively moved from old days 'Number crunching' to current 'Information processing' with more and more emphasis on Data management
• Just an example :
  - At the beginning of hard disks, data was accessed directly by applications on disk tracks
  - Then Files encapsulating several (non-contiguous) sectors of disk tracks were invented
  - Then Names were given to files, and Catalogs were created to consistently store and reference file names with their sector allocations
  - Also Databases where invented to manage large amounts of data having a known structure, and manage simultaneous access to chunks of data by different users
  - ...
• My professional experience proves indeed that 'Data is MORE important than compute' :
  That is essential as soon as we have to consider Access rights management, Transactional applications, Complex workflows, ... which are increasingly required by Scientific communities
• In particular, the success of EGEE does NOT primarily come from its gLite middleware, but from its design as 'Distributed Data Storage and Sharing'
  - permitting access to this distributed data from any gLite client, and
  - permitting, but NOT requiring, submission of remote computing activities (jobs), which MAY use this distributed and shared data
• For clarity, I suggest to change the original title to
  'Data is more important than compute.  The Execution Service MUST NOT try to perform full Data management, but MUST focus only on the management of Activities, which MUST rely on dedicated Data services for Data management'
• Anyway, I vote FOR this requirement

Morris, Balazs and Etienne on 2010-04-27

• Spreadsheet ID = 155

Amsterdam meeting on 2010-04-29

• Open  (still unclear)