Hacker News Reader

Comment by yaacov | Hacker News Reader

> “The largest supercomputers in the world are inside of two companies — Google and Facebook — and where are we pointing them?” Mr. Harris said. “We’re pointing them at people’s brains, at children.”

I'm not sure if I should be concerned that the leader of this initiative is wrong about something so trivial.

https://www.top500.org/lists/2017/11/

View on HN

You should be considered that you are intentionally missing the point by using an overly narrow and domain-irrelevant interpretation of "supercomputer".

Do you think that the users of any of those supercomputers would turn down an offer to run their projects on Google or Facebook's fleet instead of their "one" supercomputer (given some time to adapt their software)

This brings me back to something I discussed here in the past - are there scientific problems that are simply not compatible with running in the cloud, and require a supercomputer? I can't, off the top of my head, think of anything where the performance would degrade enough with the increase in communication latency between components, that the increase in available (on-demand!) resources wouldn't keep up.

Well, 3D physics simulations a la finite difference or finite element are not trivial to break up for parallelization, because the state of individual cells is mathematically dependent on the state of its neighbors, so communication latency between loop cycles would be a bottleneck on a distributed system without some kind of clever optimization.

Edit: for clarity, finite difference and finite element simulations both involve discretizing a 3D volume into individual cubes (I believe other geometries exist, but are less common) and running an update loop which performs calculations for each of millions or more cells. Of course, models with spatially small divisions, and/or those designed to handle high frequency wave propagation, require potentially enormous amounts of memory (e.g. climate simulation, or large seismic dataset processing). Depending on what math you're solving, there are heuristics which you may be able to use to update only a subset of active cells, or divide up the model to split among multiple machines with minimal loss of accuracy while retaining high precision.

Depends how Googles network is set up and how fast and over provisioned it is - how much budget you have for networking also comes into it (that's the power budget btw)

Brad Hedlund has some interesting articles on the challenges

http://bradhedlund.com/2011/11/05/hadoop-network-design-chal...

That's a good point. Thank you

While the wording by the author is technically incorrect in that Google and Facebook aren’t operating a single supercomputer, the computing resources in possession of each company would almost certainly exceed any entry in the top500.

The author used the wrong name, surely what he means is the biggest group of computers working as one system software-wise.

But that really doesn't strike a chord, especially for lay people.

I think it's a metaphor tbh.

Yes, and putting the term between "quotes" (or rather, these 'quotes') would've avoided the confusion/discussion.

Technically true, but if you were measuring raw compute power it's possible that tech companies can beat super computers. (Numbers are hard to come by because companies are secretive about how many servers they have)

Reply on news.ycombinator.com