re: NSA data centre power surges & unknowns...
Current quantum computers aren't that big, but they're only good for multiplying numbers up to 15 (I've heard rumors of 21!) As with the early Cray supercomputers, where the size of the air conditioners and power systems were much larger than the round wirey bits, a quantum computer is likely to be pretty small but may have some bigger helium coolers on-site. Remember that you're trying to entangle subatomic particles, and maybe have some magnetic detector things that still aren't all that big. If quantum computers became practical, they'd be using them to crack codes, factoring 1024-bit numbers or whatever. What's really big are the NON-quantum computers they'd be using to attack those problems now. The reason they need all that space is almost certainly for conventional data storage and processing. They want to be able to collect and store everybody's phone calls, locations, credit card purchases, and internet traffic, so they're handling multiple petabytes of data a day, finding the interesting bits, correlating the interesting people with other people who might have been in the same place at the same time, saving the parts that might be useful later. Maybe doing some voice recognition on all their phone calls. It takes a lot of basic horsepower and storage, and the correlation takes a lot of memory. I think they've bought themselves a lot of cheap-ass electrical work, done by people who assume that scaling up power requirements for a conventional data center by a factor of 10-100 doesn't change the principles. At 07:11 PM 10/9/2013, brian carroll wrote:
Lodewijk andré de la porte <<mailto:l@odewijk.nl>l@odewijk.nl> wrote:
The massive quantum computer has unpredictable power consumption.
Lee Azzarello <<mailto:lee@guardianproject.info>lee@guardianproject.info> wrote:
I guess inventing new math to break crypto has some physics problems.
these comments has me contemplating what a large-scale networked quantum computer installation would involve.. and to what extent the image of racks of servers may note correlate with the computing technology inside the boxes, or in some percentage of the data centre.
in other words- how would a quantum installation differ from the classical computer systems of the last many decades- would they be of yet another smaller scale or would an installation inherently be enormous. what if the data centre was all quantum computing in terms of the data throughput - do these numbers correlate or are different approaches to security calculations required
is it possible that a 'computer within a computer' could exist, such that a quantum chip could be embedded and use a classical electronic based system, that is, some form of stealth or hidden computing that occurs in a parallel hidden framework, including for networking data. could ethernet or fibre channel be used or would it need to be different, say tuned resonant circuits computing in other dimensionality.
2013/10/10 Bill Stewart <bill.stewart@pobox.com>
I think they've bought themselves a lot of cheap-ass electrical work, done by people who assume that scaling up power requirements for a conventional data center by a factor of 10-100 doesn't change the principles.
Most reasonable explanation yet. More reasonable than a secret large-enough quantum computer. But I continue to suspect they have one more well hidden. This has the level of "secret that people that "know a lot" know". It kind of satisfies our need to know about secret stuff the government does. The actually secret stuff, well, don't you wish we knew.
On Thu, Oct 10, 2013 at 10:24:36AM +0200, Lodewijk andré de la porte wrote:
Most reasonable explanation yet. More reasonable than a secret large-enough quantum computer. But I continue to suspect they have one more well hidden. This has the level of "secret that people that "know a lot" know". It kind of satisfies our need to know about secret stuff the government does. The actually secret stuff, well, don't you wish we knew.
You obviously have to consider not just the known unknowns, but also unknown unknowns. FWIW, I much doubt they can factor large numbers with QC (if you want to make sure, do a lit review on QC, pull up the list of names, and see whether some of them suddenly stopped publishing, or greatly reduced their publishing rate), but public key cryptosystems do have a slight smell about them lately. We definitely need more diversity in cryptosystems, and should revert to systems which are more well-understood, and focus on future systems that are simple to analyze.
In the design envelope for the data center, perhaps there is a nugget of insight in this thread and the topic it entails. http://www.pupman.com/listarchives/2003/November/msg00428.html Not a physicist, --dan
participants (4)
-
Bill Stewart
-
dan@geer.org
-
Eugen Leitl
-
Lodewijk andré de la porte