China isn't just building supercomputers, it's creating an infrastructure to create a tech industry, according to Peter Beckman, a top computer scientist at the U.S. Department of Energy's Argonne National Laboratory and head of the DOE's exascale initiative.
China's latest supercomputer, which may be officially cited as the world's fasted when Top 500 global rankings are released mid-month, is running Chinese- made interconnects and software. China-made CPUs are up next, Beckman said.
Tianhe-2 or Milkyway-2, which will have 3.1 million cores, has a theoretical speed of almost 55 petaflops. It's been tested so far at nearly 31 petaflops. A petaflop is 1,000 teraflops, or one quadrillion floating-point operations per second. An exascale system is 1,000 petaflops.
Beckman, in an interview, explains the significance of China's moves, and the power problem facing the push to exascale.
What does China's new system say about that nation's HPC technology development? It is a very clear statement of how serious they are with respect to scientific computing. If you look at their history of investment, this is just one data point in a much longer series of investments.
Was there anything about the home grown elements of this system that caught your attention? The network, and that's a pretty significant part. They have a front-end processor that's their own processor. They have slowly and incrementally woven in their own technology. They designed the interconnect from scratch, and they designed a software stack. They are taking their own approach on how to do parallelism. The two items that make the supercomputer super, the software and the interconnect, they are growing at home. The chips are well on their way. Once they have a chip that competes well it won't just be used for supercomputers.
Will China's next system have homegrown chips? I think so. I suspect that there is a national pride issue happening here as well. They will really work, in my opinion, to make a top machine that will be all (homegrown) tech from top to bottom -- the software, the interconnect and the CPU.
There is international cooperation in developing a software stack for an exascale system. Are the Chinese going their own way? They want to build their own components. They are not racing toward what is the most expedient, easiest way to deploy something. Inside the messaging layers, there were pieces that they were inventing, that they were doing over -- doing a different way. My impression is that their intent is obviously to collaborate and work with the community, but they really want to grow many of the components in-house.
Are they sharing any of this as open source? At this point it's pretty hard to see it. The software that the community is using, none of it is coming from China. It's hard to find, in some sense, on the Web. If you look for some of the pieces like the Kylin (Linux) operating system, it's not easy to find a community of people where this is being used or shared. It's certainly not prevalent yet. Maybe that's to come. I don't want, in some sense, to sell them short. It's very hard to document code in English if you're really writing essentially in Chinese. There may be language issues preventing them from doing this.