Federated accelerated computing platforms for EGI

Overview:

Accelerated computing systems deliver energy efficient and powerful HPC capabilities. Many EGI sites are providing accelerated computing technologies to enable high performance processing such as GPGPUs or MIC co-processors. Currently these accelerated capabilities are not directly supported by the EGI platforms. To use the accelerator cards capabilities available at resource centre level, users must directly interact with the local provider to get information about the type of resources and software libraries available and which submission queues must be used to submit tasks of accelerated computing. 

EGI-Engage project has implemented the support in the information system, to expose the correct information about the accelerated computing technologies available – both software and hardware – at site level, developing a common extension of the information system structure, based on the OGF GLUE standard, in order to have the capabilities published uniformly by all the sites. Users will then be able to extract all the information directly from the information system without interacting with the sites, and easily use resources provided by multiple sites. The EGI-Engage team has also extended the HTC and Cloud middleware support for accelerator cards, where needed, in order to provide a transparent and uniform way to allocate these resources together with CPU cores efficiently to the users.

A work plan was initially defined to develop a solution enabling accelerated computing support in CREAM-CE for the most popular Local Resource Management Systems (LRMS) already supported by CREAM-CE and, for what concerns the information system, based on the GLUE 2.1 draft schema. A CREAM-CE prototype allowing to submit jobs requesting one or more GPGPUs or MIC co-processors was developed for working in conjunction with the HTCondor, Slurm, Torque, SGE and LSF batch systems. 

On the cloud side, the activity started by reviewing the available technologies, focusing on GPGPU virtualisation in KVM/QEMU hypervisor, and completed by enabling PCI passthrough virtualisation GPGPU cards over an OpenStack based IaaS site which is now part of the EGI Federated Cloud. Documentation and portal has been created for easy access to the GPGPU in EGI federated cloud.

The progress on the roadmap to achieve the federation of GPGPU or MIC co-processors capabilities across EGI HTC and Cloud platforms will be presented, together with a number of use cases implemented by the structural biology and biodiversity scientific communities that already exploited the first HTC and Cloud prototypes.

Target Audience:

Research communities with interest on using accelerated computing in Grid and Cloud
Service providers with interest to offer accelerated computing capabilities
e-Infrastructures for integration and support for accelerated computing.

Benefits for Audience:

Status and roadmap of accelerated computing in Grid and Cloud in EGI
Information how to offer accelerated computing capabilities for service providers
Information how to use accelerated computing capabilities for research communities

Topic 2: Services enabling research