High Performance Computing (HPC) corner meets cloud computing

Cloud computing and personal computers have spread all over the world, so what happens when the two come together? How does high performance computing apply cloud technology?

Supercomputers have been used in various fields, energy exploration, meteorological services, physical experiments, image rendering ... Almost all of the areas you can think of can have high-performance computing participation. And the annual two-degree global high-performance computing TOP500 strong list, but also let the computer elite of all countries bow.

June 2011, the latest TOP500 high-performance computers list, the first three are: K computer, 8.2 petajoules per second, Japan, Tianhe -1a,2.57 petajoules times per second, China; jaguar,1.76 petajoules per second, United States. K's computing speed is even faster than the sum of the second to sixth computer operations. The state of the world's top computer rankings reflects the heat of supercomputers.

In China, since 2009, High-performance computing ushered in a new period of historical development. This is inseparable from the continuous construction of the city's High-performance Computing center, but also with the world TOP500 the top of the Tianhe is inseparable. But the most substantial reason is that with the rapid development of China's economy, more and more investment in scientific research, high-performance computing as a powerful lever to promote the development of the subject of a consensus.

As the frontier of scientific research innovation, colleges and universities obviously have great demand. Moreover, from the experience of foreign countries, many colleges and universities have a high reputation for high-performance computing in the world. For example, the University of Illinois, the University of Stuttgart, Germany, Japan's Tokyo University of Technology. China's Tianhe number is also produced from the National Defense Science and Technology University. High-performance computing in colleges and universities, the main reason is that more users, many disciplines, whether chemical, physical, biological, earth, aviation, ocean, etc., have provided a useful field for high-performance computing. In view of this trend, the emphasis on high performance computing centers in the future will only be strengthened and not diminished.

From this point of view, it is not difficult to understand why the High Performance Computing center of the university has been in a new period of development since 2009.

When High-performance computing encounters a cloud

Since 2010, with the deepening of cloud concept, some colleges and universities have begun to build cloud-based platforms. At the same time, the planning of High-performance Computing Center is also being carried out-the main proposition is: High-performance Computing Center as a public service platform, in the new situation, whether to integrate into the school's overall information structure?

Shanghai Jiaotong University, for example, is doing a cloud platform for the whole school to conduct network services, including Web services, mail services, forums, etc. into the cloud service platform. So what about high-performance computing platforms? Are they also included in the overall cloud platform?

This is controversial. On the one hand, the key technology used in cloud computing is virtualization, which allows resources to be scheduled on demand, but it also leads to sacrificing some of the speed. While High-performance computing is the pursuit of "high-speed", the two seemingly contradictory. On the other hand, it is expected to use the flexible application model of cloud computing to solve the problem of system management of the public High-performance computing platform. Guojie, academician of the Chinese Academy of Engineering, said that "the ease of use of cloud computing will affect the traditional HPC computing model, and on-demand cloud computing provides a more interactive computing model for HPC." Like a few years ago using a popular PC server to build a cluster and a recent GPU-accelerated scientific calculation, cloud computing was a paradigm shift for HPC. "So, can HPC actually adopt cloud computing architecture?"

In view of this, Shanghai Jiaotong University made a preliminary attempt, they proposed a fusion high-performance computing, Grid computing and cloud computing new computing framework-high-performance Flexible Computing (HPEC), can manage and dispatch multiple cloud computing resources across the domain, supporting the GPU and the heterogeneous computing environment, Users have the flexibility to apply and manage computing, storage, and network resources that can be virtualized or direct physical resources that support Highperformancecomputingplatformasaservice (Hpcpaas) Multi-node clustering and rapid deployment of upper software, while supporting computational-intensive (MPI) and data-intensive (map/reduce) applications.

And the Beijing University of Technology High-performance computing platform is also the cloud architecture-"We think we can adopt cloud architecture, because we are a public platform, not for one area, so the scheduling of resources is also very important." Professor of Beijing University of Technology, said.

In the current view, using cloud computing ideas to implement high-performance computing, may be more of a trade-off, when you need to do whatever the pursuit of high-speed, it is clear that the adoption of cloud computing and your original ideal is contrary. Versa。