Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
With the biggest chipmakers for data centers keeping Russia out of next-generation devices and mobile and software companies pulling out of that market, it’s no surprise that Russian researchers are working quickly to find ways around the new technologies that will drive the rest of the world.
This is important for Russia right now, but it’s likely that these efforts will lead to similar ones in China, which is also used for teaching sanctions, as we’ve seen with Huawei, for example.
After Russia invaded Ukraine, the US government stopped key tech exports to Russia, like semiconductors, last week. At least AMD, Intel, TSMC, and GlobalFoundries are chipmakers that are following the US export controls by stopping shipments to Russia. Dell, HP, and Lenovo have also stopped sending products to the country, and Oracle and SAP stopped doing business last night.
Since Wednesday, when key Russian banks were kicked out of the system, police have been watching over the Swiss data center that the financial messaging service SWIFT uses.
The effects of these sanctions can’t be overstated, as they affect everything from Russia’s most powerful supercomputers to enterprise systems and the wider world of mobile for both businesses and consumers.
“In the meantime,” says Andrei Sukhov, a professor and head of the CAD lab at HSE University in Moscow, “it’s not enough to just say what the problem is. You need to look for a quick way out of the situation using the resources you have.”
In a timely article for the Association for Computing Machinery (ACM), Sukhov talks about how Russian computer science teams are looking at building the next generation of clusters using older clustering technologies and a lot of open-source software for managing everything from code portability to parallelization, as well as standards like PCIe 3.0, USB 4, and even existing Russian knock-off buses inspired by Infiniband (Angara ES8430).
Such systems would have to be based on what is easily available in terms of processing and networking. Russia (and China, if it comes down to it) have better mobile options that are native to their countries.
For some background on the top end of computing, Russia has seven supercomputers on the Top500 list. The 199-node super Chervonenkis is Russia’s highest-ranked computer, coming in at number 19. Chervonenkis is made up of AMD Epyc processors and Nvidia A100 accelerators. An integrator can’t do much without important parts, such as the InfiniBand.
As it turns out, the 36th, 40th, and 43rd most powerful machines are all similar, with AMD processors and Nvidia accelerators.
“These devices aren’t the most powerful, but their production is completely independent and doesn’t affect the patent rights of American or European companies,” Sukhov said.
Even if all the parts are in place, new boards still need to be made. Sukhov said that this problem could be solved by using wireless protocols to switch between processors, even though the network latency would be bad. This would make it hard to do real tightly coupled, low-latency HPC simulations (which come in handy in areas like nuclear weapons simulations, as just one example).
“Given that the available mobile systems-on-chip are on the order of 100 Gflops, small clusters of high-performance systems-on-chip can easily reach a performance of several teraflops,” said Sukhov.
“Using standard open operating systems like Linux will make it much easier to use custom applications and will make it possible for such systems to run in the near future. It is possible for these clusters to be heterogeneous, with different systems-on-chip for different tasks (or, for example, FPGAs to make specialized accelerators for specific tasks that can be changed on the fly).”
In a short conversation after the article, he told The Register, “No special problems are expected with the supercomputers that have already been built and put to use. Based on Linux, these supercomputers can keep running even without help from the companies that made the hardware and software. So far as I know, all scientific supercomputers, even ones that are more than five years old, are run in normal mode today. They can only be stopped by hacker attacks or commands that are forced on them. But I don’t know of any actions like this that have been taken in relation to scientific projects, like supercomputers.”
“It’s clear that Russia won’t be able to make a new supercomputer in the next few years. Still, the way we’ve suggested can be used to meet all of the current needs in computing and data processing. Especially if we use different types of hardware acceleration for different tasks, “he says.
“It’s important to note that our proposed method is meant to be put into action quickly as a pilot project. During this implementation, software solutions, new protocols for exchanging data, and new computing technologies will be worked out.
“In the future, the cluster device will be able to be made better (for example, to try to launch the release of a new motherboard, which will host several chips connected by a common bus).
READ MORE ARTICLES;
