Systematic Literature Review and Survey on High Performance Computing in Cloud
Independent thesis Advanced level (degree of Master (Two Years))Student thesis
Scientific and engineering applications require large number of calculations, to solve these calculations we need more number of processors, shared memory and multiple disks. Supercomputers are capable to process huge amount of calculations. In supercomputer, all of its power is utilized to execute few programs as efficient as possible. Later High Performance Computing (HPC) came into picture to solve advanced computational tasks of various applications in various fields. HPC uses the technology of supercomputer and computer clusters to perform advanced calculations. Now a days, HPC in Cloud came into existence to solve some of the technical and enterprise problems, which occur in traditional High Performance Computing and thereby increasing the potentiality of Cloud Computing. Developing HPC in cloud is one of the challenges of cloud developers and vendors. This paper focuses on documenting some of the research which is already done in the field of traditional HPC applications, HPC applications running in cloud and also their cost and performance aspects by using Systematic Literature Review. Thereafter, we have conducted surveys with few research questionnaires related to HPC in Cloud. Results From systematic literature review we found six major Domains of HPC applications, ten performance measuring tools and techniques, six major security issues. From survey results we found that Cloud can be used for running HPC applications when computational power is spiky and for loosely coupled applications. When utilization is more than 30% it is better to use traditional HPC. From survey & Interactive conversation different security solutions are obtained. From survey result we also got some performance measuring tools and techniques for HPC in cloud, they are HPC benchmarks, HPCC benchmarks, Ganglia, NPB etc. Conclusion From the information attained so far regarding whether cloud is ready for running HPC applications or not, we observed that cloud is not a viable solution for few applications but it is catching up fast. Cost aspects of running HPC applications in cloud, performance measuring tools and techniques, security solutions for HPC applications in Cloud are identified through Industrial experts by using Survey and Interactive conversation. Information gathered through surveys is compared with SLR result and is almost equal.
Place, publisher, year, edition, pages
2012. , 67 p.
Benchmarks, Cloud Computing, High Performance Computing, Systematic literature review.
IdentifiersURN: urn:nbn:se:bth-3689Local ID: oai:bth.se:arkivexF1EBF780A8119A0BC1257AA9003C243AOAI: oai:DiVA.org:bth-3689DiVA: diva2:831000
Lundberg, Prof. Lars