NeIC Conference 2013: Report on "Closing Plenary"

"Bioinformatics" (Tommi Nyrönen)

CSC development manager

• ELIXIR mission statement: protect information created in bioinformatic research. Will have societal impact.
• Bioinformatics underpins all stages of life science research, from molecule dynamics up to population ecology.
• Android app: Power of minus ten.
• Societal driving force: From molecules to medicine.
• Many layers of science must be linked (through collaboration) to generate new knowledge.
• Data discovery (of existing, stored) and integration is key.
• Societal driving force: Make better crops that can withstand draught.
• Societal driving force: Better animal models for medical testing. Dog genomes are not sensitive the same way that human genomes are. A dog genome costs 50€.
• These tasks are to big for individual nations, or the Nordics. May be feasible on a European level.
• Preparation phase ended 2012, now we are in construction phase, until 2016, when permanent phase starts.
• BioMedInfra collects ESFRI projects BBMRI, ELIXIR, EATRIS, with CSC and FIMM.
• CSC have opened clusters and storage to research institutions. These have IT departments that maintain cloud images.
• CSC supports institution IT admins, who in turn support users.
• First step of allocation process: Operational agreement of terms and boundaries. Then start small and then scale up.
• Currently exists as pilot project, 20+ test cases.
• FIMM, chipster, Danish elixir node.
• FIMM has 10Gbit/s OPN to CSC cluster. Users have not noticed the migration from local resources to CSC resources. (this is very good feedback).
• "Tommi, when is your cloud becoming available?" --a finn researcher, when he had actually been using it unknowingly for half a year.
• Chipster: Missed this, little help please Tommi?
• ELIXIR: Web service hosting using CSC cloud resources for backend.
• CSC BioMedInfra cloud: "Service for service providers".

Prediction in 2013:

• 200+ nodes
• 3000+ cores
• 14+ Tb ram
• 6 cloud head nodes
• 600+ Tb shared storage

Top three lessons learned:

1. Provide computable memory.
2. Provide computable memory.
3. Memory.

Enabling sensitive data access:

• Data Access Committees (DACs) control research access to sensitive data resources.
• ELIXIR AAI (authentication) effort.

Next actions:

• Nordic ELIXIR Nodes
  • LoI to NeIC
  • Nordforsk grant for meetings
  • neIC IaaS for LS
• Cloud & Secure computable storage and data access
  • Biobank-derived data
  • Need DAC
• Missed this, little help please Tommi?

Take-home messages:

• ELIXIR most viable framework.
• LS need computational data access - Nordic focus is in sensitive data.
• Cloud tailored for LS is good way to access/provide e-Infrastructure resources.

QA

• Possible for Nordics to make use of CSC IaaS? ELIXIR Nodes are pilots, Swe: SciLifeLab, Denmark: CBS, ...

"Closing Keynote" (Erik Lindahl)

User, tool developer and e-Infrstructure provider

• Membrae proteins responsible for all things SEX, DRUGS & ROCK'N'ROLL! Examples: Anesthetics, Alcohols, Propophol, Drugs.
• Neurons work by emitting neurotransmitters (chemicals) that interact with membrane proteins (receptors).
• We model membrane proteins using bioinformatics, and the more you know, the more difficult it gets. Some open on stimuli, others close.
• With more CPU power you can learn more details about the process.
• Molecular dynamics are newtonian motion, very simple stuff. But many of these. That makes for the difficulties.
• Howard, Murail, PNAS 108, 11727 (211), discovery of binding sites in silico before experiments.
• Can we reverse the effects of drugs? Yes! Proven in recent publication.
• Can we reverse the effects of Alcohol? Yes! BluCetin. "Herbal supplement" in the US.
• This type of research was impossible 20 or 5 years ago.
• Science should drive e-Infrastructure. But e-Infrastructure can drive science.
• 96 cores in -96. Now 64 cores in 1U rack server.
• Projection from 2010: 1Bcores in 2024. (projection holds since then).
• "How will you use a billion cores?"
• How can we achieve longer simulations of milliseconds, with a time step of 5fs.
• Cores will not get faster, but slower. We will get more, but still, slower.
• GROMACS needed to be pretty much rewritten for vectorization. This was done in collaboration, with e-Infrastructure. Primarily, with people.
• GPUs are awesome! And are the way forward for throughput.
• A 4x performance gain means Big Money on big systems.
• GPUs suck! They are horrible to program. But this is what you will have to use in the future.
• NVIDIA: 256DP TFLPOPS on a card in 2014.
• Move to GPU or be left behind.

e-Science & Nordic collaboration (provocative, bad stuff):

• Bad examples of e-Infrastructure:
  • Tier-2 funding for not-quite-infrastructure projects providing services for nobody.
  • Job security and sustainability.
  • Local Phallus projects.
  • Money to coordination at lots of levels.
  • Never use existing standards when you can be funded for new development.
  • Interact with infrastructures, not users.
  • Give many small shares so you have a nice year book.
• Why Nordic, and not Global, or with arbitrary groups anywhere?

Aim higher!

• Nordic collaboration is great if we dare close national ones.
• Evaluate cost vs quality, let best center do the job.
• Dare to criticize and say no.
• We should be best at some things, not fail at all.
• Meritocracy before democracy.

QA

• Colocalisation important? Yes! It's good to have someone in your face telling you that your idea is stupid so you can have quick turnaround. Not easy for distributed meetings etc...
• Shouldn't e-Infrastructure support all the science that is funded? Yes! But larger projcts should get higher priorities.
• Bio is a small field. Are you arguing that you should not get e-Infrastructure support? Same answer as before. Prioritize support for projects that actually did get funded. For reference, only about a third of SRC applications get grants.
• Should e-Infrastructure be consolidated, localized, less centres? It would be better to move in that direction.

"Conference conclusions and closing" (Pentti Pulkkinnen)

Chair of NeIC board

• 2003: Grid computing tecnological push. Early NDGF.
• 2007-8: e-Science, 2nd evaluation.
• 1012-13: Almost teenager NeIC, new strategic areas, new strategy. BMS launched, possibly new areas in very near future.

• Nordic collaboration must be flexible to be successful.
• e-Science is now considered real. e-Infrastructure is not a goal unto itself, but a means to do research.
• WLCG: Sustainability important for Tier-1. LHC will operate for another 15yr.
• only one event in 10^13 interesting.
• Humanities has long collaboration tradition, wide variation of activities.
• Cloud computing: Users are the driving force, must be served. Users with money can choose the services thay want, effectively designing e-Infrastructure themselves.
• Data services. Need of data management and long term funding.

Lessons learned:

• See report summaries in the NeIC wiki.

Conclusions:

• We have entered the third path of science.
• The Nordics are a natural playground for r-I collaboration.
• Nordics have similar goals and sufficiently complementary strengths for synnergy.
• There is political support.
• People are The Key Part for e-Infrastructure.