The Science Cloud Cometh
Mankind is currently engaged in some of the most important scientific research of our age: the discovery of the elusive Higgs particle to validate our modern understanding of physics; genomic sequencing to enrich our understanding of life on Earth and to fight diseases like cancer; and the global monitoring of the earth from space used to analyze and one day predict everything from earthquakes to volcanic eruptions, to climate change or next year’s crop yields.
These monumental scientific undertakings have very different goals, but one important feature in common: the huge amounts of data that must be processed efficiently in order to yield accurate results. Unfortunately, the advanced computer infrastructure required to handle these big data needs are also exploding in size leading international scientific institutions such as CERN, the European Molecular Biology Laboratory (EMBL) and the European Space Agency (ESA) to look at additional sources of capacity to complement their existing in-house deployments. Without access to the right resources, researchers within these organizations can become limited by computing capacity in delivering and analyzing results.
The answer to this dilemma may lie in one of today’s most innovative computing delivery technologies: cloud computing. By taking advantage of powerful cloud computing platforms, these international scientific institutions can continue to add scale to their compute environments in a competitive and convenient way. With this dynamic in mind, a consortium of European cloud computing companies and international scientific institutions recently launched Helix Nebula, the ‘Science Cloud,’ with the dual purpose of fostering a healthier economic climate for the cloud, while giving the scientific sector access to innovative technology to promote research and scientific progress.
The key aim is to provide a multi-cloud solution that allows scientific institutions to deploy workloads seamlessly across different providers and locations. This involves harmonizing provisioning, networking, software environments and more. In this way, such a cloud environment is able to offer a fully-scalable and customizable infrastructure that can support the varying nature of scientific research computing requirements and the high volumes of data. To put things into perspective, at CERN alone, 25 petabytes of new data are stored per year and 250,000 CPUs are in use around the world to process LHC data. The efficiency of biomedical labs sequencing DNA has outstripped Moore’s Law significantly in recent years. This has created a bottleneck with the downstream bioinformatics pipelines that rely on high performance computing infrastructures. These requirements are increasing rapidly over time.
To satisfy these high-performance computing (HPC) environments, there are several factors that need to come together to create a successful solution:
Many clouds have adopted traditional web hosting methodologies that rely on low utilization from customers and over-provisioning. Large customers – like those participating in the Helix Nebula initiative – with heavy, data-intensive workloads and HPC needs break that model. Advanced infrastructure is a required fit for that purpose. High-speed networking, between both end user sites and clouds as well as cloud to cloud, is essential. Advanced storage strategies and intelligent multi-cloud procurement and provisioning are needed to provide expanded scalability. These are to name just a few key areas of work within the Helix Nebula consortium.
Open Software and Networking Layers
Having a flexible software layer that is able to run existing systems easily is a crucial component. With an open software layer, HPC users can easily port their data and applications to the cloud with little modification – for example, CERN used the CERN VM image for workloads conducted thus far within Helix Nebula. In more restrictive cloud deployments this would not work natively. HPC users have very specific use cases and large existing installed bases, so they need the cloud to work with and not against their existing applications and knowledge.
Being able to tune cloud infrastructure to fit directly with each use case is critical. HPC users care primarily about price/performance, which is delivered through a combination of efficient resource purchasing and good performance levels. The ability to tightly fit the application layer through the virtualization layer to the actual hardware can be very important in achieving these parameters. The ability to tailor cloud infrastructure to fit the use cases closely is therefore highly desirable. In big data, for example, many applications require a great deal of RAM in comparison to CPU. The fixed server model of many dominant public cloud providers can cause significant over-provisioning of resources and destroy the economics of using such public cloud providers. Part of the Helix Nebula consortium’s efforts therefore covers ensuring participating suppliers of cloud resources are able to reflect the requirements of the scientific institutions.
HPC needs are often temporal – at least at a project level. For instance, CERN runs its accelerator chain in long campaigns followed by maintenance windows which change their compute consumption requirements over time. Each individual DNA sequencing and assembly run lasts for a set period. A purchasing model that can match these usage profiles as closely as possible can improve utilization and therefore cost effectiveness for research institutions. A seamless model that can accommodate the purchasing of capacity in a reserved fashion but also absorb on demand needs is very important for HPC users. Delivering this behavior using multiple cloud providers offers a greater degree of scalability and is a key aim of the Helix Nebula consortium.
There is a lot of discussion around the benefits of public and private cloud environments when it comes to business and consumer services, but a flexible cloud infrastructure without deployment restrictions suits big data and HPC needs in the scientific research sector. Such flexible cloud platforms can carry the weight of projects like those from Helix Nebula members because their approach to cloud computing emphasizes performance and flexibility, without overburdening infrastructure or overprovisioning resources, and combines that with a multi-supplier deployment model. By tapping into cutting-edge developments from the leading cloud infrastructure providers, organizations like CERN, ESA and EMBL can continue to better the world through research, without the potential future roadblocks of limited computing infrastructure resources.
About the Author
Robert Jenkins is the co-founder and CEO of CloudSigma and is responsible for leading the technological innovation of the company’s pure-cloud IaaS offering. Under Robert’s direction, CloudSigma has established an open, customer-centric approach to the public cloud.