The first day at the EGI Technical Forum brought together e-infrastructure’s
movers and shakers to consider the future of grid, cloud and e-science. Not a
small task then. Luis Busquets-Perez of DG-CONNECT at the European Commission
set the scene by outlining the newly formed directorate’s mission to promote
research and innovation, creativity, security and to provide digital goods and
services – all in the spirit of cooperation. The e-Infrastructures Unit within
it covers GÉANT,
digital environments, HPC and grids. The Commission wants to avoid overlap of
research in member states and promote cross-border collaboration. “Ideas are
not like oranges,” said Busquets-Perez. “If I give you an orange, and you give
me an orange, we both have an orange. If we swap ideas, we both have two ideas,
not one.” Sharing of ideas is key – under Horizon2020, 100% of scientific
results will be available through open access.
Steven Newhouse, Director of the European Grid
Infrastructure also looked ahead to Horizon 2020, and EGI’s plans to build its
activities around three themes: community and coordination, the operational
infrastructure and virtual research environments. One exciting new initiative is to recruit EGI champions to bring in new users at an institutional level, coordinated
across Europe.
This sort of scheme has worked successfully for XSEDE in the US, and in individual
countries such as the UK, and EGI is looking forward to this initiative getting underway
on a European basis.
Chiming with this future-facing theme, the conference yesterday hosted an EC
workshop on distributed computing infrastructures for e-science. The participants
were encouraged to do some crystal ball gazing – what does the future hold?
What are the tough questions the community needs to find answers to?
One question that always rises to the top pretty speedily in
these sorts of discussions is how to meet the needs of users? Antonella Galizia
presented a case study from hydrometrology, using the example of a fatal flash
flooding event that happened in Genoa in November last year. The same thing
happened 40 years ago and researchers want to be able to predict these extreme
rain and flooding events more accurately. DRIHMS was set up to provide an
interplay between ICT and domain experts, providing simulation services,
galleries of models and training. User-friendly interfaces were at the head of
the list from users. “Users keep asking to make things as simple as possible,”
said Galizia.
Andres Steijaert, provider of cloud services for SURF, a collaboration of research institutions and
universities in the Netherlands, told us that “users are choosers” - they choose their own software and devices,
they want to bring along their own preferences to the party. According to
Steijaert, the biggest question for users is what happens to their data?
Individually, users rely on many clouds from different providers, fragmenting
data and diluting their buying power. However, by clubbing together through SURF, this gives a potential
market of 1 million users and companies like Amazon and Google are much more
likely to pick up the phone to discuss user requirements. David Wallom of the
Oxford e-Research Centre also stressed the need to be user centric – the
ability to support the users is as important as buying the hardware in the
first place.
A second theme that emerged is how much does it all cost? Fotis Karagiannis presented some preliminary results from the eFiscal project, which compared the costs of public and private distributing computing infrastructures. Early results show that in-house HPC and HTC e-infrastructures are cost effective compared to commercial services, particularly when comparing on-demand costs. Personnel represent about 50% of annual costs and the CAPEX/OPEX ratio is 30/70. However, more benchmarking is needed, particularly when comparing performance. As Mike Symonds pointed out from Helix Nebula, the science cloud project, commercial providers also have to deal with the issue of charging VAT for their services. This is coupled with the e-Infrastructure Reflection Group’s considerations that cloud computing is not yet a solution for all scientific applications and the future is likely to be hybrid – public/private and grid/cloud/HPC. As Tiffany Trader reported from the conference in HPCinthecloud this morning, “The reports of grid's demise have been greatly exaggerated.”
No comments:
Post a Comment