Wednesday, April 10, 2013

Pay for use: the road ahead?

Kostas Glinos made it clear in the opening plenary of the community forum: the EC will no longer accept the status quo when it comes to future funding of e-infrastructure projects. The wonderful scientific discoveries and impressive technological developments resulting from the array of projects that make use of distributed computing must now also be backed up by greater sustainability after projects end, better coordination of resources, optimised governance and, potentially, new models for funding.

Pay-for-use is a model familiar to many in the world of cloud computing; one of the reasons that cloud has taken off in industry while grid has struggled is that businesses understand the idea of paying for a service –  paying for what they use. The marketplace in cloud is subsequently enormously diverse, a benefit due in part to the free marketplace that can quickly adapt to changing customer requirements. Competition means that commercial cloud providers now sometimes provide free cloud services to individuals with the offer of upgrading to a ‘pro’ package – more suited to a heavy user, or a business – for a recurring  fee. So could the pay for use model work to bring similar benefits to grid computing?

Scientific research is not a business: there is the need to be cautious and completely shifting to pay-for-use may not be so popular with the e-science community that is used, in Europe, to central funding for e-infrastructures and individual project funding. On the other hand, some parts of the community are already using commercial cloud providers to do everything from supercomputing, to storing large datasets. Already used to the greater agility, dynamism and adaptability of such services, they could be ready for a pay for use grid.

Warburton toll gate, south-west Manchester.
Pay the price: get to use the road.
(CC-BY-SA David Dixon)
“Until e-infrastructure providers try these models, how would we know what works,” said Sy Holsinger, Strategy and Policy Officer at EGI. After a paper from 2012 was endorsed by the EGI Council, the time was right to turn the ‘thought experiment’ into a real-world test. Exploring potential brokering models, EGI have tested ‘matchmaker’ and ‘one stop shop’ models, in addition to the ‘independent advisor’ model currently used. Matchmaker sees the federator (EGI) allocate resources from resource providers that match customer requirements, with the resource provider paying the federator for establishing the the contractual agreement and the customer paying the resource provider; One stop shop sees the federator doing everything including collecting payment from the customer, then paying the resource provider. For pay for use, both of these models work better, with lower overheads, than the independent advisor model currently used by EGI, where interactions are more decentralised (incurring more overheads); the federator listing services offered by resource providers and being funded by a membership. So pay for use may mean a shift in the role that EGI – and other infrastructure providers – adopt.

Problems and challenges ahead? There are some. Some academic research centres cannot, by law or contract, resell services to third parties. There is the question of taxation. Many member countries have VAT, which could have implications for the purpose of invoicing.

The economic situation in Europe means that, looking ahead, the status quo is not an option. Hopefully, by testing some new ways of working, infrastructure projects can explore how their own sustainability might be achieved, while making the range of computing services available to researchers more diverse and more agile: better resources, and better science. The status quo it is not.

No comments: