CloudLightning was recently featured at the Fifth National Conference on Cloud Computing and Commerce (NC4) held in Dublin City University on April 12th. Prof. John Morrison (University College Cork, Ireland) gave an overview of the project, followed by Dr Marian Neagul’s (Institute e-Austria Timisoara, Romania) presentation on the CloudLightning Service Description Language.
Dr Christos Papadopoulos-Filelis (Democritus University of Thrace, Greece) discussed simulation in CloudLightning. The project involves three major use cases: Genomics, Oil and Gas Exploration and Ray Tracing. Computationally capable sets of resources, usually composed of accelerators in conjunction with CPUs, are offered by the system with respect to user requirements for use cases. Computational efficiency, in the sense of optimal coalitions of hardware, and power consumption are also major issues, hence necessary parameters affecting the execution of such applications must be a priori known and will be crucial for simulating such an environment at large scale. The simulation is an important step in assessing the performance of such a novel system, which is dedicated towards HPC in the Cloud.
Perumal Kuppuudaiyar’s (Intel Lab Europe) talk was focussed on the implementation of test bed which had integrated with various state of the art software stacks on top of the heterogeneous resources to provide FT/HA clusters, fined grained resource management and containerised workload orchestration for HPC.
Perumal demonstrated the 3D graphics interactive rendering from one of the use case Ray Tracing using Blender 3D model, Intel’s Embree (Ray Tracing Kernel) libraries on Intel’s Xeon Phi co-processor. Rendering jobs were packaged within the container along with application libraries and could be executed selectively either on CPU or on Xeon Phi with direct access using kernel device pass-through. Interactive rendering for remote clients were achieved with SSH X11 forwarding with data compression to optimize the display frame rates and X server (Xming) running at windows clients machine.