Description of Work 1/8 Cloud services for science e-Infrastructure in 2011 OBJECTIVE The main objectives are to implement and deploy cloud computing and cloud storage services for Norwegian and Swedish research groups and to disseminate knowledge that enables the participating centers and end-users to build competence on supporting and maintaining cloud-based services. 1. Background In 2010, UNINETT Sigma and SNIC participated in the NDGF project “NEON” – a fully Nordic collaboration, coordinated by NDGF through SNIC/KTH - to assess the feasibility, ease-of-use and cost efficiency of cloud computing technology in the Nordic realm. The findings of this project are: • Private cloud technology is not mature enough yet to provide a transparent user experience. It is expected that this will not be the case until earliest mid 2012 • Public cloud technology is mature enough but lacks certain features that will be necessary in order to include cloud resources in a transparent manner in a national infrastructure (e.g. quota management) • Public clouds are competitive in the low end for non-HPC jobs (low memory, low number of cores) on price. • A significant fraction (ca. 20%) of the jobs running on the current supercomputer infrastructure is potentially suitable for cloud environments. This holds in particular for single-threaded or single-node jobs with small/medium memory requirements. • There is sometimes a backlog of “real” supercomputer jobs that suffers from the non-HPC jobs on the supercomputer infrastructure. • Available storage capacity is not accessible in a user friendly way; most storage clouds are only accessible via programmable interfaces. A pilot “cloud-backed storage” service was developed as part of Notur/NorStore’s contribution to the NEON project. This service gives users access to cloud storage via either a web interface or network drive (using the WebDAV protocol). The storage can have a quota per user, and can use multiple storage back-ends. In 2010, an Amazon S3 and a local disk back-end were developed. The aim is to provide end-users a lowthreshold interface to cloud storage resources to manage and share scientific data and eventually build community-specific services with the assistance of selected user groups. SNIC established a multi-site private cloud (Eucalyptus) that is API compatible with the Amazon public cloud. The objective was to monitor the maturity of private cloud solutions to determine their suitability as a low-cost alternative once they reach the same maturity level as public clouds like Amazon. 2. Strategy In the proposed activity, the national computing infrastructures in Norway and Sweden will build on the results of the NEON activity in 2010. The infrastructures will investigate the migration of scientific tasks from Notur/SNIC resources to cloud environments in 2011. As the public cloud has a low threshold for entry-level user groups and communities, another Description of Work 2/8 aim is to reach user groups and communities for which the compute and storage resources in the present national e-Infrastructures are difficult to use. To achieve these objectives, the national infrastructures will be offering cloud computing services and storage services on a small to medium size scale in a public cloud. Both computing and storage services will be gradually made available to a selected group of end users, following the recommendations of the NEON project. Notur and SNIC will work closely with a limited number of projects to get the services deployed on top of the public cloud. This will lead to organizational and operational experience. It should also bootstrap and organize user communities that are able to share computational and data cloud environments. The focus in the proposed activity lies strongly on providing end users a low-threshold interface to both cloud computing and cloud storage. This is achieved by: 1. Funding: UNINETT Sigma and SNIC cover the cost of using the (public) cloud resources 2. User support: partial funding is provided for advanced user (application) support for selected user communities 3. 4. Dissemination and training: workshops are held to train the participating centers and help end-user projects get started (e.g., hands-on boot camps) 5. Repositories: build a catalogue of default configured cloud resources (e.g. standard virtual machines) for the national infrastructure By the end of the project, research projects are able to request access to cloud resources and application support by applying through the regular Notur and SNIC channels. Partners that are interested in participating in deploying the cloud services should have the necessary skills to support end users on the provided cloud services. 3. Activities and milestones Common to the work packages mentioned in this section are user and project management, billing and accounting – investigating and implementing transparent methods to monitor and handle cost for research groups. This includes the national and cross-national view, i.e. how to secure the project allocations from the national infrastructure are mapped onto contracts (cf current consolidated billing in Amazon Web Services). 3.1 Cloud computing To start offering a cloud computing service, the following activities are defined: Work package 1: Cloud computing service setup Start date: April 1, 2011 Delivery date, July 1, 2011. The objective is to set up a ready-to-use public cloud computing service that can be used by research projects and communities. This needs to be prepared thoroughly to provide a smooth user experience from the start. The allocation of cloud services should be offered to users of the national infrastructure in a transparent manner (and at the same level as Description of Work 3/8 existing services in the national infrastructure). Any services rendered in the "cloud" should thus be subject to the same accounting, user and project management policies. The user management and accounting regime from the national infrastructure must therefore be interfaced with the cloud service provider, thereby using the existing administration systems as much as possible. Activities include: · · · · Set up access management, i.e. cloud service provider set up with the current administrative management tools being used by the national infrastructures. Identify and engage communities that are candidates to run on a public computing cloud. Select users/communities with cloud-friendly applications (e.g., life sciences). Establish a support process that guides users to the most effective support resource, ranging from public support forums (e.g. Amazon) to national infrastructure support services. Transfer support knowledge to partners (e.g. site operators) via documentation and workshops; Organize a workshop on how to use the available cloud technologies. Participation is mandatory for partners that wish to deploy the cloud support services locally. Result: After this work package, all preparations have been done for a selected group of end users to start applying for (and use) a public cloud service through the national infrastructure. Work package 2: Deploy cloud computing service Start date: July 1, 2011 Delivery date November 1, 2011. The objective is to deploy the computing service from Work package 1 in the national infrastructure. Users can apply for cloud resources via the normal national allocation procedures. The aim is to train at least one person from each participating centre on providing support for the cloud computing service. These people should be able to support (local) users on the cloud towards the end of the project. Activities include: · · · · · · · End user documentation for using the cloud computing service Raise awareness for the available cloud computing services (and its advantages, e.g., immediate availability, no scheduling, low-threshold interfaces) Implement and fine-tune the support and management process for users and communities of the cloud computing service Build and maintain a catalogue of standard virtual machine images for use in various “types” of research projects. These images are selected based on user community input. Evaluate and forecast actual usage and cost of the cloud computing service Maintain a private cloud solution that is compatible with the public cloud service (for cost comparisons). Description of Work 4/8 Result: The result of this work package is a public cloud computing service with an implemented user support structure. The computing service is integrated in the national infrastructure and includes a repository of pre-configured resources available. The participating partners are capable of supporting local users. 3.2 Cloud backed storage Work package 3: Personal scientific storage Start date: Feb 15, 2011 Delivery date: July 1, 2011. The objective of this work package is to provide researchers with a personal cloud storage service for managing scientific data. The work builds on the results of the NEON activity in 2010 and consists of deploying and polishing the end-user (web) interface and integrating the service with a national identity provider (IdP). The metadata “tagging” system will be enhanced. The service will initially be deployed in the Amazon cloud, though the design must ensure that other cloud service providers can be added with reasonable effort. The service will have a WebDAV interface. Activities include: · · · · · · · · · Design and integrate the web interface with a national IdP. Design and implement support for the WebDAV interface, i.e., “WebDAV disk access”. Design and implement support for a metadata system Improve user-friendliness of the web interface Test and document use cases for the cloud-backed storage Document and publish software as open source; Bootstrap an open source community around the software End user documentation Deploy an initial version on the public cloud Operate the initial version Result: Users will be able to use this cloud service as their personal scientific storage area via a drive interface and web interface, with the option of tagging data and making data publicly available. The provided interface provides essential functionalities for users to manage data. The design and use of personal spaces are properly documented both for developers and users. Work package 4: Shared spaces for community building Start date: July 1, 2011 Delivery date: November 1, 2011 Users that have access to the personal cloud storage service must be able to create shared storage spaces (much like project areas) with other users that have access to the Description of Work 5/8 service. These shared spaces are set up and configured via the available web interface. After initial set up, they are also available via the disk interface (WebDAV). Activities: · · · · · · Design and implement support for users to create shared spaces Design and implement support for sharing of existing files and folders with full access control Test and document use cases for shared spaces Deploy the version developed in this work package on the (public) cloud Deploy the version developed in this work package in the national infrastructure for scientific data. Operate the deployed versions Result: Users will be able to use the cloud-backed storage to create shared spaces. The design and use of shared spaces are properly documented both for developers and users. Work package 5: Deploying shared spaces for key user communities Start date: May 1, 2011 Delivery date: December 15, 2011. Work packages 3 and 4 provide general tools for storing and sharing data via a cloud backed storage service. This work package focuses on adapting these general services to specific user communities. In this work package, we define the (minimum) metadata that is considered mandatory for specific user communities to document their (shared) data collections, as well as any additional metadata that is required to integrate the data in the national infrastructures for scientific data and computation. This activity is done in collaboration with key user communities that help in defining the metadata requirements (e.g., life sciences, climate, linguistics). Activities: Engage selected user communities; Define relevant test cases in collaboration with these communities Define the minimum required metadata per community Adapt the cloud backed storage service to provide and require the specified metadata per community; A user-friendly interface is built for managing and querying the metadata. Transfer support knowledge to partners (e.g. site operators) via documentation and workshops. The aim is to train at least one person from each centre on the cloud backed storage service. These people should be able to support (local) users on the cloud towards the end of the project Result: A number of shared spaces are made available in the national infrastructure, tailored to specific user communities. The whole service is properly documented, both for developers and users. Description of Work 6/8 4. Resources and timeline The project starts in January 2011, and runs for one year until December 31st, 2011. The project has 3 FTEs (1.5 from each SE, NO). In addition to these 2, the project has 0,2 FTE for coordination – 0.1 per country. Operational resources are taken into account. Costs of using public cloud resources are specified in euros. 4.1 Resources for the project Manpower: Country FTE Name NO 1.0 Maarten Koopmans (coordination), Andreas Bach, NN SE 1.0 To be determined Public cloud costs: The budget for using the public cloud for the described services. Country Service Cost NO Cloud computing service NO Cloud backed storage SE Cloud computing service SE Cloud backed storage 4.2 Work groups Group 1 Work package 1 and 2 Maarten Koopmans, Andreas Bach, NN, To be determined Group 2 Work package 3 , 4 and 5 Maarten Koopmans, Andreas Bach, NN, To be determined 4.3 Time plan Deliverable Description Due date Group 1 Support procedures cloud computing, provisioning May 1 1 2 Initial launch computing service July1 1 Description of Work 7/8 3 Transfer knowledge computing service September 1 1 4 Catalog of shared cloud November 1 images 1 5 Deploy test version April 15 cloud backed storage as personal scientific drive 2 6 Deploy cloud backed storage as personal scientific drive July 1st 2 7 Deploy test version cloud backed storage with shared spaces September 15 2 8 Deploy cloud backed storage with shared spaces November 1 2 9 Identify and engage user communities for cloud backed storage July 1 2 10 Deploy community specific metadata on cloud-backed storage Dec 15 2 5. Risks and measures Risk Impact Measure Cost and quota for services Medium active monitoring; start with a small set of projects. Cloud backed storage for end users is new technology Medium Phased roll out to gain operational experience Little enthusiasm end users Low Advertise and promote services Too much enthusiasm end users High Use as advertising; Conservative sites/operators Medium Most universities already work on clouds in their curriculum in their comp. science dept. (and some other places as well). Activate those users directly Demand for windows support Low accept at most one pilot to bootstrap this for roadmap 2012 and beyond Upon success tender required Medium Evaluate in aug/sept timeframe. Description of Work for continuation 8/8 If needed, start tender process, maybe collaborative.