Clouds, Grids, and Standards
The European Commission was putting a focus on Cloud Computing when launching the Digital Agenda. A cloud as a metaphor for our increasingly ubiquitous digital life is very strong and the specific focus helped a lot to gain the public interest. In the case of scientific computing the somewhat simplified message could raise some concerns. But, over time the communication and discussion have matured which, if not before, became evident when joining the 4th Cloudscape event (23-24 February 2012) and its release of the SIENA project’s “Roadmap on Distributed Computing Infrastructure for e-Science and Beyond in Europe”. Neelie Kroes, Vice–president of the European Commission and responsible for the Digital Agenda deepens the message already in her foreword: “…science is increasingly driven by the processing of big data, researchers need access to science clouds and other e-infrastructure that satisfy their requirements. Interoperability of services and applications is a key concern…”. This statement is followed up in the actual roadmap by the conclusion for the community to “…continue the efforts begun in the EGI-InSPIRE project to build the standards and software to support clouds formed by federating multiple smaller cloud providers, with possible extension to commercial public clouds”.
Neelie Kroes’ foreword and the conclusions of the report underlines that to satisfy the researchers’ demanding requirements we need as usual make use of all our IT-skills and capability. We know very well from long-term efforts in HPC, networking, meta-computing, grid computing, and data intensive research that there are seldom any simple suits-all concepts. It is the same with virtualization and cloud computing – it adds new great possibilities but alone it will not satisfy all the different needs. To cite the SIENA Roadmap: “Cloud computing is not a magic wand for scientific computing …”.
“EGI is the largest, most powerful, and most comprehensive distributed computing e-infrastructure supporting research in the world.” Such a position can only be maintained by continued growth and sensibility to the researchers needs. This requires an almost constant transition or evolution of the infrastructure without breaking the users trust. Key factors are transparent infrastructure, resilient services and high availability and reliability.
I can only agree when it in the roadmap is stated “we are now at a cusp in the development of e-infrastructure to support research in Europe. It seems clear that […] variations of cloud computing will ultimately dominate computing of most kind in the future.” The activities to address this prediction started within EGI over a year ago. To make this transition we not only need to understand and solve the technical issues but also analyse the fundamental differences in the concepts of grids and clouds.
To compare the concepts it is stated in the roadmap, “Grid computing enables sharing of distributed computing resources; cloud computing offers computing as a service.” Maybe correct but it doesn’t put the finger on the real difference. Production grids can typically also be considered as service-oriented, based on middleware adhering to a Service Oriented Architecture. Still, grids have indeed a focus on sharing, as one of the problems that grids were designed to solve was collaboration and sharing of resources across organizational borders. And here we are homing to a real difference: while grid computing has an origin in an ambition to make the sharing so standardised and easy that you would consider computing as a utility, cloud computing has its origin in the sale of services or abundant computing resources. Cloud computing also comes with the ambition to fulfil the promise of computing as a utility but with additional factors of market logic: differentiation and customer lock-in.
With this perspective the reason for lack of standards in the cloud space is evident: Cloud computing appeared in large scale first as a commercial offering.
Standardization is key for the take-up of cloud computing within research. At the same time as the Roadmap points to the amount of standardization work left to do, it reminds of the suite of early standards available from the bodies of Distributed Management Task Force (DMTF), the Storage Networking Industry Association (SNIA) and the Open Grid Forum (OGF). To build any new development on current standards should be obvious, and through EGI-InSPIRE and its Federated Clouds Task Force a lead has been taken in the work on gathering required standards and methodologies for federation of distributed virtualized resources. For this I can only agree with the roadmap: “in this respect the future looks bright”!
As a final remark it is noted in the Roadmap: “the EGI-Inspire project has an important role to play in evaluating cloud computing adoption strategies and trends at National Grid Initiative (NGI) level so as to harmonize future steps. Such an approach would reduce duplication effort and fragmentation through bespoke solutions, while offering further opportunities to capitalize on ongoing standardization efforts.”
Federation of virtualized resources is a great opportunity for EGI to enlarge the offering of flexible and reliable computing and data services for the European Research Area and its global collaborations. The importance of the EGI members’ engagement is evident and I can only conclude that knowing through EGI what a collaborative effort can achieve, the future indeed looks bright!