U.S. Invests $258M in Supercomputing Race with China
wsj.com
wsj.com
Our scientists look at what could be, look at their very interesting datasets, and then gnash their teeth for want of better tooling. People know this. Senior people from industry have screamed at senior government leadership that they are 20 years behind and falling faster. Nothing changes. Why?
Every time someone gets out from under their grad school payoff contract, they run to Google or Facebook.
Very few people are committing for the long term, and that's what's needed. Long-term commitment to steward a program. The people who remain often lack technical chops, have "certificates" in "cyber" something or other, but got through the Acquisitions school and now have authority to just pour money into blades, load Red Hat and mpi4py, and say "look, it's a supercomputer"!
They don't need to invest in iron. They need to invest in people. But they think they can just keep hiring new grads and expect magic.
Much the computing on these machines isn't supercomputing and doesn't need fast interconnects. In fact, many (most?) workloads would be best done in the cloud. Except we already bought these machines. So there's tons of frameworks around meant to use these machines like they are AWS pizza boxes. If it's inefficient due to an architectural mismatch, that's fine. You're still using the machine that was paid for, and the only numbers that matter is utilization, and those metrics aren't even defined in terms of CPU utilization. If you reserve 10 nodes, and only run on one node, it's still reported that a 10 node job ran for 6 hours, not that a 10 node job job ran at 10% efficiency for 6 hours.
As far as why the accounting is this way, it's assumed that users will attempt to maximize their usage since allocations are defined. But even then, you can blow through your allocation and generally ask for more. But that's not necessarily what happens.
In summary: Because we bought these big computers, we have to use them, even if the workload isn't a good match.
Why would China be working on conventional ship-to-ship missiles if they expected a nuclear war? They'd just use ICBMs / long-range missiles.
PS: Remember, we where doing optical recognition for missiles twenty years ago. So, reasonably close is going to be a hit most of the time.
I know, I worked on some of them.
> However, consider people have had a long time to prepare for aircraft carriers without a direct conflict between major powers and aircraft carriers are easily tracked in real time by satellites.
As I mentioned in another reply, it isn't as easy as some people seem to think to find and track a CVBG by satellite. As I also mentioned, I'll go into more detail when I'm not on my phone.
> Further, they might as well be sitting still in terms of a high mach (AKA 10+) missiles.
I'm not aware of any hypersonic cruise missiles in that range, and those that approach it have relatively short ranges. I would need to look at this further.
> PS: Remember, we where doing optical recognition for missiles twenty years ago. So, reasonably close is going to be a hit most of the time.
Longer, I believe. Image recognition for terminal guidance is not foolproof, though. Plus, as with any weapons system the countermeasures groups have not been sitting still.
But the real problem is more basic. They are very high value targets an can be taken out for vastly less money than they cost to produce.
Taken with a grain of salt: China might not actually be 'there' yet, but they are progressing toward that goal.
Anyway, I would need to look into that. Chinese claims have a tendency to be blown out if proportion, particularly by the media of their neighbors.
[1] http://infoproc.blogspot.com/2017/02/on-military-balance-of-...
[2] http://infoproc.blogspot.com/2016/09/balance-of-power-in-wes...
It's a fictional BBC news segment covering a "serious incident" between NATO and Russian forces.
"Tensions between NATO and Russia have been growing over the past few years, and the continuing military build-up in the Baltic region is increasing concerns that a serious conflict might be a real possibility. This dramatization depicts one possible scenario. We hope that it shows how fragile our world is, and how quickly it can be changed by military and political decision-makers who do not speak for the tens of millions of people whose lives they may well destroy."
It's only tangentially related to the conversation, but it's such an excellent work of art that I can't help plugging it. The authors appear to be anonymous, which is interesting. If you were affected by the movie Threads or any other nuclear fiction, you might find this video similarly powerful.
I'm sure that when China does the same, they're not imagining the cost of war against, say, Australia or Mexico. they invoke the bogeyman of American military hegemony.
[0] https://www.theregister.co.uk/2017/06/13/f35_pilots_report_h...
https://news.usni.org/2017/06/14/aviation-leaders-still-unsu...
> The PathForward program [...] address the four key challenges: parallelism, memory and storage, reliability and energy consumption
Energy consumption is becoming a very important challenge, as the article says, current petascale supercomputers consume "enough power to run a small town". Major efforts will need to be made to have Exascale supercomputers that don't consume the equivalent of 10^3 small towns.
[1] https://www.top500.org/news/doe-shells-out-258-million-to-si...
Think about it - what China can possibly lose in such a race? When Americans are trying to maintain its No.1 ranking in the field, Chinese are just trying to further modernize its industry which stands at less than $10k GDP per capita, you can probably beat the next generation Chinese supercomputer, but hey, you are pushing for the creation/improvement of the Chinese supercomputing community!
Think about it again - what China can possibly lose in such a race? America won't lose its overall dominance just because some other countries have faster supercomputers, America is losing its influence around the global when it stupidly using its borrowed resources to force its main competitor (also No.1/2 debt owner) for more R&D spending.
Why Chinese are building supercomputers on their own? Because the US laws prohibit any US supercomputer export to China, including the export of core components that can be used to build such supercomputers.
In case you are not familiar with the matter - the above mentioned core components are not some fancy parts guarded by your NSA - we are talking about parts that can be freely purchased on ebay for $300-1000 each.
There's a commercial need for expensive distributed processing now, with Google/AWS/Azure spending a lot on video cards. The ML craze may make government spending on supercomputers less important, especially if people prefer to distribute calculations on many interchangeable computers.
I wouldn't mind a few NSF or DoD grants for novel computing architectures, but I'm not sure I buy that having massive supercomputers for economic forecasting is "strategically important for nations".
by the end of phase 3 when they actually procure machines its mostly 'just give me one of what you're already shipping, and it better run MPI well'
doe exascale was supposed to be fundamentally different. because by the time you got there all the incremental improvements in power and latency management weren't nearly enough anymore...my guess is that they'll just make big gpu clusters and call it a day
its probably not that much money in the scheme of things, but the sad (or good) depending on your perspective is that HPC got so far out of the mainstream that when corporations finally got around to worrying about scaling, they just did their own thing. so I guess they can thank darpa/doe for infiniband? not really?
and they're still running fortran/mpi
[1] http://wccftech.com/us-government-bans-intel-nvidia-amd-chip... [2] http://www.pcworld.com/article/3086107/hardware/chinas-secre...
CCP has its own "Make China Great Again" thing, officially known as "The great rejuvenation of the Chinese nation".
https://www.amazon.com/CHINESE-DREAM-GREAT-REJUVENATION-NATI...
"What I was trying to say is that my opinion is not informed based on what I have read in the press but based on the fact that I have spent time in the country and have read the local press, and spoken with the local people. The humiliation faced by the Chinese at the hand of the western powers is a common theme of books and movies and TV shows in China."
Not this someone.
Nothing the GGP said about the national Chinese interests was remotely controversial (perhaps subsequent comments are unrepresentative of individuals' thinking). Heck, Obama wanted to pass TPP mostly as a bulwark against Chinese ambitions.
Faster super computers allow you to do all kinds of things that would otherwise be quite hard. Think of supercomputers as weapons that you can fire very quietly, nobody even notices that you just tested some new (possibly nuclear) weapons system or that you designed some fighter jet, submarine or rocket.
Super computers are a game changer, not having them is as much as ceding territory, having them but not knowing how to use them has a similar effect. Just like computers were a game changer in the previous war so super computers will be a game changer in the next, with the added caveat that you might not even notice when it starts or ends other than that you suddenly find out that your counterparty runs the show because they cracked the codes you thought were safe.
You can even get time on machines that aren't in the country. I'm sure Chinese experiments have run on US supercomputers at this point.
[0] http://www.anandtech.com/show/11547/us-dept-of-energy-awards...
http://www.anandtech.com/show/11547/us-dept-of-energy-awards...
Something like the Rex Computing Neo would be more suitable for this.
Guess this now needs reconfiguring http://worrydream.com/ClimateChange
Okay. That's the kind of crap that was in the 1800's, that's still in the 2010's, unfortunately.
Your words sound like something from a grown up bully jealous of their victim's achievements in life.
I wish the best for China and the Chinese people. I hope for a peaceful and prosperous future. That however does not involve the threat to take Taiwan by force or claiming all of the South China Sea as theirs.
What I was trying to say is that my opinion is not informed based on what I have read in the press but based on the fact that I have spent time in the country and have read the local press, and spoken with the local people. The humiliation faced by the Chinese at the hand of the western powers is a common theme of books and movies and TV shows in China.
I've listened to my grandmother talk about her grandparents, who are from the 1800's.
Of course we're drawn to movies and books and TV shows from that era. Same reason why teenagers in Europe are interested in Medieval stuff and Roman empire stuff. It's part of their history.
Is China taking Taiwan by force? Taiwan has been developing peacefully for almost 70 years since the civil war, I don't see any violence over there.
..."claiming all of the South China Sea as theirs"
Lots of governments do lots of crap, does that mean we can say the whole country and all its people want it?
And what does U.S. has anything to do with any of this stuff, anyway. Sure I understand if China was doing things in Mexico or the Caribbean, but I don't understand why they want to meddle in Ukraine, Syria, Afghanistan, and South China Sea, and then call Russia, Iran, China as agitators.
Do I think that's what the U.S. public support? Meddle in countries half way around the world? No!
And thanks for your best wishes.
China has publicly acknowledged they are pointing DF-16 missiles at Taiwan.
China has publicly stated they will use force to prevent an independent Taiwan, despite the fact that Taiwan is a functioning democracy, unlike China.
China arrest and brutalized dissidents.
The rule of law in China is what ever the current leader says it is - Where is 吳小暉 (Wu Xiaohui)?
China kidnapped booksellers from Hong Kong in violation of the treaty that governs Hong Kong because of their political views.
I could make a list a mile long.
I stand by my statement that China's rise to power is scary because the motivations of the government of China are not driven by democratic principles that provides for checks on power.
The USA could have had an empire after WW2 and there was NO ONE IN THE WORLD that could have stopped them. Instead the USA paid even more of their money to re-built the world and help the countries which they fought.
The USA makes stupid mistakes sometimes, but they alway eventually sort it and sort it on the right side of history even if it takes time.
There are several areas where the US has not yet done so (including some that are older than the US, like treatment of Native Americans, and including some where it's heading in the opposite direction of the right side), so the idea that the US always does so is empty nationalistic sloganeering.
A non-exhaustive list of the countries the U.S. owes in terms of reparations for wrongdoing:
1. Iran
2. Nicaragua
3. Dominican Republic
4. Haiti
5. Cuba
> And what does U.S. has anything to do with any of this stuff, anyway
This is whataboutism and has far less relevance in foreign affairs than in interpersonal ones.
The U.S. Government has incentives to keeping the South China Sea's international boundaries in line with the U.N. treaties. Much of, if not most, of the world's trade passes through there. Since the U.S. economy depends on this trade, it is considered to be one of the most important things to protect.
The U.S., just like every country, is primarily self-interested. It makes sense for the U.S. to want to contain what it perceives as aggressive moves to claim what have previously been considered international waters.
In the spirit of fairness, the Russian Federation considers NATO enrollments of neighboring countries to be aggressive behavior on the part of the U.S. The U.S. isn't the only country threatened by apparently threatening behavior.
The U.S. would defend its involvement in Syria, Iraq, Afghanistan, and in Pacific Rim countries in terms of its self-interest for stability in those reasons, stability being a euphemism for unrestricted trade.
> Ukraine
I'm not familiar with U.S. involvement in Ukraine; it's not a NATO member and the U.S. hasn't (publicly) committed forces there.
This crosses into incivility and personal attack, and those things are not allowed here. Please edit all that out of your comments when you post.
We detached this subthread from https://news.ycombinator.com/item?id=14565533 and marked it off-topic.