Feb 22, 2016
WLCG today• Successfully supported LHC run 1• Many lessons have been learned – already
several significant changes to the computing models
• Experiments pushing to higher and higher data rates- Already in 2012 x2 data volume vs a “nominal” LHC
year as planned• Funding for future computing is a problem
- Flat budgets are the (optimistic) working assumption
July 17, 2013
Requirements vs resources• Simple extrapolation of what is optimistically affordable will
barely accommodate minimal likely requirements- Significant increases in need anticipated from LS2
• 2015 will already be a problem if we start from the 2014 base
July 17, 2013
+34 PB/yr+363 kHS06/yr
2008-2012 was essentially a linear increase – with ~flat budgets
WLCG strategy• Live within ~fixed budgets• Gain x10 - x?? in event processing throughput
- Find additional resources (additional budgets)- Make (much!) better use of existing resources
• Optimisation of cost of [CPU, storage, network, power]• Invest (limited) available effort where it is important
• Software, data management
• Collaborate with other science communities- Share expertise, experience
July 17, 2013
• HEP now risks to compromise physics because of lack of computing resources- Has not been true for ~20 years
From the 2013 update to the European Strategy for Particle Physics
g. Theory is a strong driver of particle physics and provides essential input to experiments, witness the major role played by theory in the recent discovery of the Higgs boson, from the foundations of the Standard Model to detailed calculations guiding the experimental searches. Europe should support a diverse, vibrant theoretical physics programme, ranging from abstract to applied topics, in close collaboration with experiments and extending to neighbouring fields such as astroparticle physics and cosmology. Such support should extend also to high- performance computing and software development.
July 17, 2013
i. The success of particle physics experiments, such as those required for the high-luminosity LHC, relies on innovative instrumentation, state-of-the- art infrastructures and large-scale data-intensive computing. Detector R&D programmes should be supported strongly at CERN, national institutes, laboratories and universities. Infrastructure and engineering capabilities for the R&D programme and construction of large detectors, as well as infrastructures for data analysis, data preservation and distributed data-intensive computing should be maintained and further developed.
High Performance Computing
WLCG Strategies• Update of the computing models – to show that every effort is
being made to make best use of resources- Ready for LHCC in September - Covers the period of LHC Run 2- Will have estimate of the 3 year resource needs
• Concurrency forum- Encourage work on all aspects of software – - Critical in optimising resource use- Critical in preparing to use modern architectures, HPC, etc.
• Reduce operational effort so that WLCG Tiers can be self supporting (no need for external funds for operations)
• Position ourselves so that the experiments can use pledged and opportunistic resources with ~zero configuration- (Grid) clusters, clouds, HPC, …
July 17, 2013
Implications:• Must simplify the grid model (middleware) to
as thin a layer as possible- Make use of any resources made available- Make service management v lightweight- Centralise key services at a few large centres
where possible• Rely on the networks as a resource
- Push to ensure that all Tier 2s are connected at realistic bandwidths
July 17, 2013
External funding?• WLCG benefitted greatly from funding from EC, US
(DoE/NSF), and other national initiatives• This funding has largely stopped now• Prospects for future funding exist – but the boundary
conditions will be very different:- Must demonstrate how we benefit other sciences and
society at large- Must engage with Industry (e.g. via PPP)- HEP-only proposals unlikely to succeed- Also it is essential that any future proposal is fully engaged
in by CERN (IT+PH) and experiments and other partners
July 17, 2013
HEP value?• Building and operation of the world’s largest globally
federated, distributed, infrastructure• Management of multi-petabyte data sets and facilities• Record of collaborating with other scientific domains
(EGEE, OSG), and industry (openlab, Helix Nebula, …)• And more…• Other sciences now need to address some of the same
problems as HEP: we must collaborate- This is one reason why we must avoid HEP-specific
solutions as much as possible, we don’t have a good record of building broadly useful tools
July 17, 2013
Speculation – future e-infrastructure?• Will need to cover many aspects:
- Facilities:• Networking and federated identity services• An academic cloud resource (most sciences, esp smaller groups need
this)• Experienced and sustainable data archive centres
- Software tools:• to allow a science community to federate their own resources (e.g. HEP)• Tools for data management, workflows, application support• Tools to aid integration of daily activities with the e-infrastructure
(collaborative tools, “dropboxes”, etc….)• Software tools to build citizen-cyberscience infrastructures
- Also need investment in software – • Today we benefit from commodity hardware;• Today our software is not that efficient on the current hardware• Future CPU and storage commodities may be very different• We need to adapt our software (potentially continually)
July 17, 2013
Outlook• HEP can no longer expect funding without
demonstrating its relevance• Need to broker collaborations with other relevant
sciences- Data-intensive or relevant to society- Must be mutual benefit, but we can also learn from other
communities (e.g. HPC)• Funding, expertise, and effort will be harder to attract
and retain in next few years- Focus on key issues and what we need for the long term
• However, HEP has significant expertise and experience that we can build on
July 17, 2013
Evolution – key points• Need to demonstrate that we are doing as much as possible to make
best use of resources available• Software
- Parallelism, new architectures, etc – significant challenges and lack of expertise- Requires some investment
• Commonality- Between experiments- With other sciences
• Simplicity- Reduce complexity where possible:
• Grid services• Deployments (e.g. central service is simpler)
• Focus HEP efforts- Where we must – e.g. data management tools- Cannot afford (nor should we) do everything ourselves
• Collaborate- To bring in expertise and to share ours
July 17, 2013
Data Management/Data Preservation• Data Management: LHC has demonstrated very large scale data management.
Must build on this success:- Improve our own efficiency – improved models of data placement, caching, and
access- Work with wider scientific community to build community solutions or move towards
open standards if possible• Should drive/explore collaborations with other data intensive sciences and
industry to consolidate data management tools in the long term- Build a broader support community, helps our sustainability
• Data preservation and open access to data is now very important and highly publicly visible - policy and technical investments are needed now
• CERN could provide the leadership to build a set of policies and technical strategies to address this for HEP, building on the work of DPHEP and collaborating with other big data sciences
• It is already urgent for LHC experiments to ensure the ability for long term use and re-use of the data
July 17, 2013
e-Infrastructures• Grid computing has been very successful in enabling global
collaboration and data analysis for LHC• We should generalise this (with technical evolutions already
under way) for a more general HEP-wide infrastructure- Several requests from other experiments to use the WLCG- Must understand how this would fit with regional infrastructures,
national and international efforts etc., and how to integrate with other science communities
• We must plan for the next generation of e-infrastructure (e.g. on 5-10 yr timescale), making use of new technologies, e.g.:- Terabit networks – could enable true remote access- How does “cloud” computing fit? – need to understand costs of
commercial clouds vs HEP (or science) clouds: cloud federations would be needed for our collaborations
July 17, 2013
Investment in software• There is a growing consensus that HEP
code needs to be re-engineered to introduce parallelism at all levels
• We need to act now or run the risk that HEP software will not run at all in the future
• afternoon session
July 17, 2013
Long term – LHC experiments• Recognise the importance of re-thinking computing
models for post-LS2 timeframe- Could easily imagine order(s) of magnitude more data - Current models simply will not continue to scale
• Recognition that experiments’ solutions today are not so different- Must devote effort on effective areas:
• CPU-heavy algorithms, simulation, data management, etc- Frameworks and data serving mechanisms could be
common- Willingness to bring these ideas together and start to plan
what needs to be done
July 17, 2013
Long Term strategy• HEP computing needs a forum where these strategic issues can be
coordinated since they impact the entire community:- Build on leadership in large scale data management & distributed computing –
make our experience relevant to other sciences – generate long term collaborations and retain expertise
- Scope and implementation of long term e-infrastructures for HEP – relationship with other sciences and funding agencies
- Data preservation & reuse, open and public access to HEP data- Significant investment in software to address rapidly evolving computer
architectures is necessary- HEP must carefully choose where to invest our (small) development effort – high
added value in-house components, while making use of open source or commercial components where possible
- HEP collaboration on these and other key topics with other sciences and industry• CERN plans a workshop to discuss these topics and to kick-start such an
activity
July 17, 2013
How does HEP computing adapt?
• HEP-wide project on future computing?- Coordinate all the projects in HEP that address future
evolution of computing and software- Governance by the HEP community itself- Launch new projects in areas where there are identified holes- Act as the focal point for collaboration with other sciences and
industry• We have experience in this level of collaboration• We have a bigger problem now than we have had in the
past• HEP computing needs to take a change of direction now
July 17, 2013
Summary• WLCG/LHC experiments recognise the need
to plan for the future• HEP has significant expertise/experience
now on core technologies and has an opportunity to engage with other sciences and industry- Essential if we are to maintain funding and attract
people• Plan community-wide
discussion/engagement on key themes
July 17, 2013