Speaker: Nesreen Al-Malkawi, Systems Administration, German Jordanian University, Jordan
I have more than twelve years of technical experience in the field of Information Technology, besides my technical experience, I have a Master of Science in Engineering, my research experience in the field of HPC Cloud Benchmark.
• Master Degree in Computer Engineering, Warsaw University of Technology
• Bachelor Degree in Telecommunication & Software Engineering, Al-Balqa’ Applied University
• Library Database Systems Section Head, German Jordanian University
• Systems Engineer, German Jordanian University
• Laboratory Teaching Assistant, German Jordanian University
• Administrator of the University Library System (Horizon)
• Administration of the GJU library Software/System.
• MyVLib at GJU Library
• E-Library Portal at GJU Library
Abstract: “Benchmarking HPC in the Cloud”
Cloud computing has been considered one of the most important buzzwords in the Information Technology and business industry in recent years and will likely remain so for the foreseeable future. It draws significant attention from researchers due to its widespread applica-tions and substantial benefits. Using the cloud, the resources are offered as a service, in this case, the user does not need to configure and maintain On-Premises IT infrastructure, the resources can be leased from the cloud service providers and used on-demand, furthermore, the user can allocate resources as needed then deallocate them as well, in a totally, it is an elastic environ-ment. Virtualization technology is the main heart that enables cloud computing, using this tech-nology the performance overheads on CPU, memory, networking, and the disk might be affected because of resource sharing. High-Performance Computing (HPC) applications need super com-putation power, so, they might be affected by resource virtualization. When deploying an HPC cluster in the cloud, each node of the cluster is provisioned as a Virtual Machine (VM), this VM shares the other VMs and services the same cloud hardware resources which may lead to reducing the cluster performance. In this paper, HPC Clusters in the public cloud were evaluated. Three public clouds were chosen based on different hypervisor implementations (KVM, Xen, Hyper-V) and their Gartner report score, different Virtual Machine (VM)/Instance types, and specifica-tions were chosen. The efficient MPI version was implemented on the HPC Clusters in the selected three clouds, and two HPC cluster locations scenarios were implemented. In terms of performance benchmarking, the implemented HPC Cluster scenarios were benchmarked using open source programs such NAS Parallel Benchmarks, and HPL. The benchmark results were justified and analyzed to provide a guide to the next generation of High-Performance Computing as a Service.