|
|
Log in / Subscribe / Register

Southampton engineers a Raspberry Pi Supercomputer

[Posted September 13, 2012 by jake]

The University of Southampton (UK) has put out a press release about a 64-node supercomputer recently built out of Raspberry Pi systems. In addition, the racks were built out of Lego. "Professor [Simon] Cox adds: 'The first test we ran – well obviously we calculated Pi on the Raspberry Pi using MPI, which is a well-known first test for any new supercomputer.' [...] 'The team wants to see this low-cost system as a starting point to inspire and enable students to apply high-performance computing and data handling to tackle complex engineering and scientific challenges as part of our on-going outreach activities.'" There is a guide to building your own, as well as a page for pictures of the supercomputer.
to post comments

Phoronix did something like this with 6 Pandaboards

Posted Sep 13, 2012 21:11 UTC (Thu) by pflugstad (subscriber, #224) [Link]

Back in June:

<http://www.phoronix.com/scan.php?page=article&item=ph...>

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 13, 2012 21:25 UTC (Thu) by atai (subscriber, #10977) [Link] (2 responses)

A challenge: A supercomputing cluster of Commodore 64s

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 11:04 UTC (Fri) by wazoox (subscriber, #69624) [Link] (1 responses)

Someone actually did a cluster of Apple IIs, using the joystick port for signaling IIRC. I must have this in a bookmark somewhere... Got it:

http://home.comcast.net/~mjmahon/AppleCrateII.html

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 14:46 UTC (Fri) by pboddie (guest, #50784) [Link]

I remember when the Acorn Archimedes came out - the tenuous connection to this news story being that it had an ARM2 CPU - one of the benchmarking exercises done by one computing magazine involved comparing the new system's time to execute a program to that of a group of twenty BBC Microcomputers (6502 CPU) working on the same task in parallel.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 13, 2012 22:26 UTC (Thu) by sciurus (guest, #58832) [Link] (22 responses)

As people pointed out on Hacker News, calling it a supercomputer is a stretch when the entire cluster does fewer GFLOPS than a single Intel Core i3. It's still a cool hack, though.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 13, 2012 23:33 UTC (Thu) by BrucePerens (guest, #2510) [Link] (11 responses)

I suppose it's interesting for a student project, in that you might learn the same issues of intercommunication, etc.

Still, I hate when toy projects are promoted as if they are the real thing. Another that seems incredibly silly is building a space satellite based on an Android phone. Space-quality solar cells for a microsat cost $10K. A launch might cost $100K (and that's a bargain). And we're going to skimp on the CPU.

They already skimp a lot. Most student satellites never communicate with the ground reliably because they're using AX.25 as their radio protocol.

AMSAT can provide universities with a much better platform.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 13, 2012 23:58 UTC (Thu) by dlang (guest, #313) [Link] (8 responses)

actually, a typical android phone is going to have a much more powerful CPU than most of what's in space.

the question is if it's going to survive the radiation.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 17:45 UTC (Fri) by BrucePerens (guest, #2510) [Link] (7 responses)

Oh, sure. And if radiation is the only problem, it would survive in low earth orbit because the radiation is higher up. And they aren't getting a ride anywhere else but low earth orbit.

But the components don't have the thermal, mechanical thermal cycling, and vacuum qualification they need to operate in space.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 18:01 UTC (Fri) by dlang (guest, #313) [Link] (6 responses)

> But the components don't have the thermal, mechanical thermal cycling, and vacuum qualification they need to operate in space.

I agree that these won't have the _qualifications_ to operate in space.

the point of disagreement is how badly you _need_ qualifications.

just because something hasn't been certified doesn't mean it won't work.

If you are spending millions of dollars launching something, it makes sense to spend a few thousand dollars to test everything first.

If you are only spending a few thousand dollars to start with, it's probably not worth the extra cost.

I actually expect that a modern phone will work reasonably well. If the solar cells can provide enough power to keep it going, it will probably hold together for the short time that these systems will be in orbit.

you don't think that sputnic-1 had all it's electronics 'qualified' before it was launched do you? that didn't stop it from working.

we're talking about roughly the same orbit (but with 1% the mass, and significantly more capabilities)

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 21:02 UTC (Fri) by Cyberax (✭ supporter ✭, #52523) [Link] (5 responses)

>you don't think that sputnic-1 had all it's electronics 'qualified' before it was launched do you? that didn't stop it from working.
Actually, they were. Sputnik-1 had been tested before the flight.

Also, all of the first Soviet satellites were _pressurized_ exactly because regular electronics don't really work that well in space. In particular, thermal hot spots, that happen easily without convection (never mind the actual lack of air), can quickly wreck silicon.

Many countries tried the 'cheap disposable satellite with non-space-rated parts' route later. It hasn't worked out well enough to justify the costs.

Also, there's nothing magic in space-rated parts. They are just produced in low volumes so the economics of scale doesn't have a chance to kick-in.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 21:36 UTC (Fri) by dlang (guest, #313) [Link] (4 responses)

the countries that tried the cheap route, how long were their satellites expected to run? how much did they cost to launch?

as I said, if you are spending millions to launch something, spending more on space rated parts makes sense.

for $8k and a few weeks of life, it's far from clear that the space rated parts are worth it.

Assuming that space rated parts only double the cost (and it's probably worse than that), are you better off with two launches of non-rated parts or one launch with rated parts?

if the non rated parts reduce your effective lifetime more than 50% then the rated parts are worth it, if the reduction in effective lifetime is less than 50%, you are better off with two launches of non-rated parts.

remember we're only talking about a couple weeks here, so a non-rated part that will last a month in space is just as good as a rated part that will last 10 years)

also keep in mind that two launches let you learn from the first and alter whatever experiment you have for the second launch.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 21:48 UTC (Fri) by Cyberax (✭ supporter ✭, #52523) [Link] (3 responses)

Lots of sats failed just after the launch. And to be useful you need at least a couple of years of service. And then there's a question of reliability - if you ground infrastructure can't trust the satellite to remain available tomorrow then you need to plan for this.

The problem with non-space-rated parts is that probability of failure is cumulative. And it's already non-trivial even for space-rated parts so duplicate and triplicate redundant systems are used.

> remember we're only talking about a couple weeks here, so a non-rated part that will last a month in space is just as good as a rated part that will last 10 years)
A couple of weeks is not useful for anything but publicity stunts and some rare experiments.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 21:52 UTC (Fri) by dlang (guest, #313) [Link] (2 responses)

> A couple of weeks is not useful for anything but publicity stunts and some rare experiments.

And this is exactly what we are talking about

the $8K package puts the satellite in a low orbit that will only last a few weeks before re-entry, so you aren't going to get multiple years of service in any case.

nobody planning a multi-year service satellite is going to be planning to use an Android phone to power it, but for an experimenter, this is a reasonable thing since it gives you a lot of stuff in a small, low-power package at a low cost.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 16, 2012 4:39 UTC (Sun) by BrucePerens (guest, #2510) [Link] (1 responses)

That $8K doesn't include your ride to space.

It's a sin to send crap into space. At least the LEO orbits don't pollute orbital space for long, they de-orbit fast enough. But given the cost, and the fact that so few people can get their experiments flown, it is the experimenter's responsibility to not be frivolous. Sending up something that can't dissipate its own heat, that outgasses the electrolyte from its capacitors, etc., is just stupid and wasteful.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 16, 2012 20:00 UTC (Sun) by BrucePerens (guest, #2510) [Link]

Oops, they say it does include a ride to space. But they are a year late in launching any. It seems a very low cost for such a ride, and the satellite design looks kind of lame too - solar panel area is really small, transceivers very low powered, etc.

And they are also advertising a lunar sample return mission!

File under "too ambitious to be real" for now.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 0:27 UTC (Fri) by fluxion (guest, #62978) [Link] (1 responses)

Interorbital Systems (I think they secured contract with NASA for launching payloads alongside SpaceX) has a satellite launch kit for $8k:

http://interorbital.com/TubeSat_1.htm

Definitely in the realm of a tinkering enthusiast willing to save for a few months to put their hobby creations in space.

Interesting times are ahead...

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 1:07 UTC (Fri) by dlang (guest, #313) [Link]

very nice. .75Kg max, for just over $8k they provide a kit that includes solar cells, radio, antennas, CPU, development tools AND the launch.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 10:47 UTC (Fri) by epa (subscriber, #39769) [Link] (2 responses)

I imagine that 'supercomputer', like 'mainframe', refers to the architecture of the machine rather than to any absolute measure of performance. A Cray-1 is a supercomputer and this machine is faster than that.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 16:18 UTC (Fri) by andrel (guest, #5166) [Link] (1 responses)

Specifically, a supercomputer has a balanced architecture. In a true supercomputer the I/O, memory, and network subsystems are capable of keeping the CPUs going at full speed.

This system doesn't have that sort of architecture.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 22:14 UTC (Fri) by ballombe (subscriber, #9523) [Link]

You forgot the most important: the cooling system.
No mention of it either.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 17:00 UTC (Fri) by daglwn (guest, #65432) [Link] (6 responses)

Moreover, it's networked using Ethernet. A supercomputer isn't defined so much by its FLOPS as it is by its interconnection bandwidth and latency. And Ethernet ain't it.

Building a REAL supercomputer requires a tremendous amount of work to get everything scaling together nicely.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 18:01 UTC (Fri) by Cyberax (✭ supporter ✭, #52523) [Link] (5 responses)

10G Ethernet is quite popular in supercomputers. New 100G Ethernet begins to get popular as well.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 20:53 UTC (Fri) by daglwn (guest, #65432) [Link] (4 responses)

In clusters, maybe. Not in a real supercomputer. There is a LOT of communication that happens on a big IBM or Cray machine. That's why those companies build custom interconnects.

Infiniband is _maybe_ somewhat adequate. For certain codes. And certain problem sizes.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 21:31 UTC (Fri) by daglwn (guest, #65432) [Link]

100G Ethernet is 100 Gbits/s. With terrible latency and horrendous congestion issues.

On a supercomputer we're talking bandwidths of ~10 GbBytes per port (meaning ~150 G-bytes into/out of a node, compared to 100 G-bits/s for 100G Ethernet), scaled up to a bisection bandwidth of ~5 T-Bytes/s with latency in the 1us range compared to ~3ms for 100G Ethernet.

And then there's I/O bisection bandwidth of 1 TByte/s or more.

Numbers from here:

http://www.theregister.co.uk/2010/05/25/cray_xe6_baker_ge...
http://www.nccs.gov/jaguar/
http://www.networkworld.com/news/2012/091212-brocade-100g...

Note that Jaguar is a relatively old machine these days so the HPC numbers are conservative.

It's mind-boggling just how huge these machines are, in FLOPS, I/O and network capability.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 21:47 UTC (Fri) by dlang (guest, #313) [Link] (2 responses)

the problem boils down to how you define the term 'supercomputer'

There was a time when a system with the specs of this system would have been a supercomputer with no arguments from anyone.

you can try to define a supercomputer by it's technical specs (including interconnect speed, etc), but you have to constantly redefine your spec list.

or you can define it based on the architecture and how it's used.

While the performance of this system and the systems in the 'top 100' list are wildly different, they are actually far closer to each other in terms of architecture and how they are used than any of them have in common with the machine you are using to view this post, even if that machine is superior if you just look at the specs (flops, storage, memory)

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 23:31 UTC (Fri) by daglwn (guest, #65432) [Link] (1 responses)

> the problem boils down to how you define the term 'supercomputer'

Here's one attempt: A machine the can run the large challenge problems of the day in a reasonable amount of time.

> they are actually far closer to each other in terms of architecture and
> how they are used

If by, "how they are used" you mean, "programmed with MPI," sure. I can program the multiple cores on my desktop with MPI too. But the architecture of a Jaguar-class system and that of this Raspberry Pi cluster are vastly different. Sure, they have processors, memory, network and I/O but so does every other computer in existence.

Think of it this way. This Raspberry Pi thing is about taking some existing components and hooking them up in an interesting way. Building HPC systems is about designing a SYSTEM from the ground up, tailored to meet the most demanding computing challenges of the day.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 23:42 UTC (Fri) by dlang (guest, #313) [Link]

umm, many supercomputers were built by taking existing components and hooking them up in much the same way that this thing was. special hardware is optional, not an inherent requirement.

if a system is only a supercomputer if it can handle the most demanding challenges of the day, is something 1/10 the power of the most powerful device still a supercomputer?, what about 1/100 the power?

note that the "top 500 supercomputer" list has the top entry with a rating of 16324 and the bottom entry with a rating of 60. how can a machine so weak still be considered a "supercomputer" according to your criteria

the Raspberry Pi thing being discussed here is a cluster of machines, not a single machine. It has separate memory, storage, network, etc for each node. This is a vastly different environment to work with than a desktop system.

No, this cluster isn't going to do any groundbreaking research or solve any "large challenge" problems. But as something to teach people about supercomputers and let the experiment with and learn what does and doesn't work for HPC computing, this is a much better thing to use than an equivalently priced SMP system.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 4:56 UTC (Fri) by smurf (subscriber, #17840) [Link] (3 responses)

Everybody else may still only get one Pi at a time, and they grab 64.

I hate them.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 5:36 UTC (Fri) by dlang (guest, #313) [Link]

there are quite a few places where you can order them a bunch at a time.

I ordered one from one of the early vendors and still haven't received it, but I ordered three from another vendor a few weeks ago and got them in two days.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 8:23 UTC (Fri) by ekj (guest, #1524) [Link] (1 responses)

The restrictions on pi-ordering have been lifted, anyone who wants to, can order as many as they like. Atleast that's what the raspberry-pi homepage says.

Delivery-time is still high though, 11 weeks.

Southampton engineers a Raspberry Pi Supercomputer

Posted Sep 14, 2012 15:48 UTC (Fri) by dlang (guest, #313) [Link]

> Delivery-time is still high though, 11 weeks.

It depends where you get them. from http://www.mcmelectronics.com/product/83-14277 the delivery time is days


Copyright © 2012, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds