HPC cloud
HPC cloud refers to the use of cloud computing services and infrastructure to execute High Performance Computing (HPC) applications. Normally these applications are executed on Supercomputers due to their processor and memory needs. In late 2000s, cloud computing providers became popular and began offering several resource options, which were able to serve the first wave of HPC applications. Researchers started to investigate the benefits of using cloud computing versus on-premise cluster[disambiguation needed]s. The main findings of these studies were that (i) cloud is suitable to handle peak loads of HPC applications that require no inter-process communication; (ii) cloud providers needed to offer high-speed network, such as InfiniBand to support tightly coupled parallel applications (those with high inter-process communication). For the applications with no high speed network requirements, which are a large number of scientific and engineering applications, cloud providers already offer hardware with compatible performance compared to various on-premise HPC facilities. Apart from handling a large subset of HPC applications, cloud brings not only cost reductions to maintain a computational infrastructure, but also the speed to access HPC resources.[1][2]
After identifying bottlenecks in cloud platforms, researchers and practitioners developed solutions to optimize the performance of HPC applications in these platforms [1]. The performance improvements came mainly from resource managers able to identify how applications should be mapped to computation infrastructure, platform selectors to help users identify which resources are suitable to their needs, and application performance predictors to help on both application-infrastructure mapping and platform selection.
One of the promises of cloud computing is to facilitate the use of HPC resources by encapsulating all the complexity of the underlying infrastructure to the user. One of the increasing areas of HPC cloud is to transform HPC applications traditionally executed in clusters into services to be hosted in cloud platforms. By doing so, users do not need to submit jobs to Job Schedulers; they simply make queries to the service to obtain their results.[3]
There are various studies on how cloud computing is helping companies serve their HPC needs, which be found in the UberCloud Experiment reports.[4]
See also[edit]
References[edit]
- ↑ 1.0 1.1 Netto, M.; Calheiros, R.; Rodrigues, E.; Cunha, R.; Buyya, R. (2018). "HPC Cloud for Scientific and Business Applications: Taxonomy, Vision, and Research Challenges". ACM Computing Surveys. 51 (1): 8:1--8:29. doi:10.1145/3150224.
- ↑ Gupta, A.; al, et (2013). "The Who, What, Why and How of High Performance Computing Applications in the Cloud". doi:10.1109/CloudCom.2013.47.
- ↑ Balis, Bartosz; al, et (2017). "Porting HPC applications to the cloud: A multi-frontal solver case study". doi:10.1016/j.jocs.2016.09.006.
- ↑ "UberCloud Experiment Reports". Retrieved 23 January 2018.
This article "HPC cloud" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:HPC cloud. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
