Sun Super Computer May Hit 2 Petaflops 134
Fletcher writes to tell us that Sun Microsystems has revealed their "Constellation System", a new supercomputing platform that the company hopes will put them back in the running for top dog in the supercomputer race. "The linchpin in the system is the switch, the piece of hardware that conducts traffic between the servers, memory and data storage. Code-named Magnum, the switch comes with 3,456 ports, a larger-than-normal number that frees up data pathways inside these powerful computers. 'We are looking at a factor-of-three improvement over the current best system at an equal number of nodes," said Andy Bechtolsheim, chief architect and senior vice president of the systems group at Sun. "We have been somewhat absent in the supercomputer market in the last few years.'"
Good to see! (Score:2, Interesting)
Once again you have shown us the power of talent, determination, and skill.
Rock Rock On!
Re: (Score:1)
Obligatory (Score:2, Funny)
Re: (Score:3, Funny)
Re:Obligatory (Score:5, Informative)
The first one of these being built, at the Texas Advanced Computing Center, is in fact running Linux, not Solaris (See this Register article [theregister.co.uk]). Sun will support both.
Re: (Score:1)
Will we see the Java Desktop System [sun.com] for Linux OS be recontinued and enter a return to life (Release 3 for Linux)?
Zoolander (Score:4, Funny)
Re: (Score:1)
Re: (Score:2)
IBM has already [slashdot.org] beaten this....(3 > 2)
Tm
Re: (Score:1)
"We have been somewhat absent..." (Score:3, Interesting)
Still, kudos to Sun, for real. Investors may get mad that Sun is full of terrific technology and solid R&D but can't seem to build the business model that will let Sun capitalize on it all. But from my perspective... God, that sounds almost refreshing, doesn't it?
Re: (Score:2)
Funny, I was just wondering when exactly Sun was in the Super Computer market. Sun was a low-end Unix machine company that transformed itself into a high-end Unix server/mainframe company. (I remember drooling over the brand new 64-way specs of the E10000 Starfire.) AFAIK, they never even touched the domain of supercom
Re:"We have been somewhat absent..." (Score:5, Interesting)
FWIW, that's where Sun sees its future. Which makes sense. There's no point trying to compete with Linux for low-end applications (and by "low-end" I mean everything from desktops to simple Web-app servers). Sun has always been good at crafting products for that top 2% of customers who really, really need that high-availability or high-performance component that isn't going to make a difference for the other 98%. And Sun can charge for them.
Re: (Score:1)
Except that, in the long run, the platform of choice for the HPC/high-availability/high-high-peformance market is turning out to be Linux, thanks to IBM and it's HPC business. Ever wonder exactly why IBM contributed NUMA, SMP, numerous networking improvements and JFS? You
Re: (Score:3, Interesting)
Ha, well, yes, there is that. Only I wouldn't lie it completely at IBM's feet. After all, IBM is pretty much the only mainframe vendor still around. They have a vested interest in selling that kind of supercomputer, even though they've obviously seen the writing on the wall for their mainframe business.
Outside the commercial secto
Re: (Score:2)
--Linux-on-mainframe is just waiting in the wings rt now; they're not really "pushing" it yet.
Re: (Score:3, Interesting)
I fell in love with Sun when I first laid my hands on a Su
Re: (Score:2)
Let's rephrase a tad. Sun has been good at SELLING products for that top 2% of customers who really, really need that high-availability or high-performance component that isn't going to make a difference for the other 98%. And Sun REALLY REALLY charges for them.
Sun is in t
Re: (Score:3, Informative)
why the Java market, it's everywhere. of course, Sun hasn't made a thin dime off of it, but the market sure went for it.
Re: (Score:1)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
This, after the strategy of open sourcing all the family jewels has only been running for the last year. That's quite a turn around in quite a short period of time.
Re:"We have been somewhat absent..." (Score:4, Interesting)
Re:"We have been somewhat absent..." (Score:4, Informative)
Re:"We have been somewhat absent..." (Score:4, Funny)
IBM Blue Gene/P (Score:4, Informative)
Re: (Score:3, Interesting)
I'm waiting to see a customer actually purchase one, and for it to be installed, and actually running customer code, before I really care.
Re: (Score:2)
Re: (Score:2)
Everyone in the industry knows who the customers are. Some of the specifics are secret, and what codes are actually run on the systems is not known. Really huge machines do have to be manufactured, and pushed through Q/A before being sent to a customer, even a classified customer. The on-site installation guys have security clearance, but no HPC company has all of the manufacturing and Q/A people cleared. If the secret-sites were buying fastest-in-the-world sized computers, people w
Re: (Score:1)
Re: (Score:2)
Top500 uses a benchmark called Linpack to determine how fast a machine is. Linpack is basically designed to show off big systems. It can squeeze almost every bit of performance out of a system. It offers lots of thread level parallelism for superscalar and vliw designs to exploit. It makes great use of FP multiply-accumulate. The program and the data easily fit inside processor caches. The amount of message passing between nodes is tiny. It's the perfect benchmark for
Re:IBM Blue Gene/P (Score:5, Funny)
Re: (Score:2)
The couple who decked out that kids room, while thoughtful and probably quite intelligent are also dorks (not the good kind!) that poor kid is totally going to rebel just as the comments say.
Re: (Score:1, Insightful)
Re: (Score:1)
Re: (Score:2)
Now imagine that on Blue Gene
Re: (Score:2)
And yet... (Score:5, Funny)
Re:And yet... (Score:5, Funny)
Re: (Score:1)
Re: (Score:1)
Re: (Score:1)
Re: (Score:1)
Re: (Score:2)
Bigots like you are very tiresome.
Couldn't resist... (Score:5, Funny)
-
In Soviet Russia, TFA reads you!
Constellation class system (Score:5, Funny)
Re: (Score:2)
Re: (Score:1)
Re: (Score:2)
Re: (Score:1, Offtopic)
Great units (Score:5, Funny)
We are looking at a factor-of-three improvement over the current best system at an equal number of nodes
Whoa, slowdown boy, just tell us how many laptop-miles of power this machine has!
--
Great hosting [dreamhost.com] 200GB Storage, 2_TB_ bandwidth, php, mysql, ssh, $7.95
Slow improvements finally paying off (Score:3, Interesting)
Consider 500 top-end T2 systems hooked up to some very fast switching hardware and you're performance per wattage ratio are going to be very persuasive to those running big data centers, although with the T1 systems the only thing which stopped us adopting them was the shared FPU (telephony codec transcoding sucks on them).
Could we see suns equivalent of IBM's BlueGene system appearing next year? I definitely think so
Re: (Score:2)
Though the T2 does offer dramatically improved floating-point performance, as compared to the T1, I've seen no evidence that it out-performs better than quad-core opterons, on the sort of HPC workloads needed for these sorts of systems. The T2 is designed for transaction processing, which has very different needs.
The main reason I don't think Sun would sell T2 for supercompute, is that they haven't mentioned it at all, and t's the sort of thing they
Re:Slow improvements finally paying off (Score:5, Informative)
]{
First Prime Factorization Post (Score:1, Funny)
Throughput: the race is on (Score:5, Insightful)
This SUN machine is a bigger-scale example of the same. It uses AMD Barcelona chips, and derives its power from internally routing data more efficiently than (most of) its competitors. It seems that in the Moore's-law endgame, what makes the chip a star performer is the surrounding components and their engineering for efficiency.
This will be better for geeks, as it makes the skill of efficient design come back into play after years of "bigger is better." Now if it just extends to software as well, we'll all benefit...
Re:Throughput: the race is on (Score:4, Informative)
> more than sheer increase in speed.
> This SUN machine is a bigger-scale example of the same.
No, not really - parallelism has of course been around forever. But its application for high performance computing has been constantly demonstrated on large servers for the past 15 years. This is back when parallelism on intel hardware might at most have meant two cpus. And that was rare.
MPP (Massively Parallel Processors) systems like Teradata and IBM's SP2 (aka DeepBlue - that defeated Kasparov at chess) successfully demonstrated great performance for the dollar back around 1994-1995 or so. These were originally designed around data mining and math computations - but found most of their sales in data warehousing. Meanwhile, CRAY was complaining that not all problems were good candidates for this kind of more cost-efficient hardware.
By 1998 you could put db2 or informix on a hundred-node SP2, each node consisting of an eight-way SMP, each with its own dedicated storage. Queries on that old system were lightning fast compared to most other options. I worked directly on SP2s and worked with a team that has a 128-node one. Oracle & Sun eventually ecliped these solutions with massive SMPs. But much of that was more due to Informix's financial issues than technical merit - since Informix and DB2 (and of course Teradata) on MPPSs easily out-scaled oracle on SMPs. The SMPs were easier to adapt to application design changes, but the MPPs were easier to grow indefinitely large.
These newer solutions are just more of the same thing - you've still got the same challenges in:
- tons of OS and application images that must be consistent
- node communication bandwidth (major selling feature of all these solutions are proprietary internal networks)
- failover (how do nodes failover, especially if they have any dedicated resources)
- scheduling (how do jobs get assigned to nodes)
So, they're much bigger and faster than 10-15 years ago - and I'm sure there's got to be some cool innovation going on under the hood. But nothing looks fundamentally different from then. And nothing here has been inherited from the pc world.
Re: (Score:3, Informative)
I keep hearing this... (Score:2)
Do you really know what you're talking about? Can you be more specific about the specific data paths between components and how the operating system controlled them or interfaced with them?
I mean, for example, if you wanted to make a video playback or slideshow program could you get the hard disk controller to blit directly to the screen bypassing main memory, or perhaps over the system bus, or a secondary channel? (I doubt it). Or could you describe a similar
Re: (Score:1)
Agnus was the memory controller chip (like the Sun chip), and had the blitter and copper which had independent access to the I/O bus and main+video memory (well video memory for the screen you were using could be anywhere in ram, allowing simple double buffering etc.) The copper has a simplified instruction set (move, wait and skip) and was tied to raster timings and good for screen effects while the blitter did blitting, and you could per
Re: (Score:2)
It has always been Intel's POV that they should be the sole deliverer of silicon for PC's.
The reason for introducing MMX was the following. In 1995/1996 Philips Semiconductor was working on a project for creating a multi-media chip, which would have been an addition to the PC architecture. Intel did not like this and introduced MMX, so that they could say that such additional hardware was not necessary.
3,456 (Score:5, Funny)
Re: (Score:3, Funny)
Re:3,456 (Score:5, Informative)
Re:3,456 (Score:5, Informative)
This is not an unusual arrangement for existing infinaband networks. The distinction is that they have all of these 864 switch modules in a cabinet, and the wiring is probably traces on a backplane, rather than flexible cables. This improves the reliability, reduces the cost, and makes it a whole lot easier to install. That may sound silly, but you're talking about 10,000 cables, each with endpoint connectors on each end. Even buying in bulk, that's a lot of money in cables.
From Kill Bill Vol 1 (Score:1)
Now, someone help me with trees. I, for one, am unclear how a tree is applicable to ports. If I have an Ethernet router with 5678 ports--perhaps this is even better than 3456 ports-- I don't want some ports to talk immediately to the root node of a 12-ary tree while other ports have to send data meandering through twigs, branches, and trunks. If a 12-ary tree is good, a 3456-ary tree is better. Better yet may be every one
Re: (Score:2)
Do you feel lucky? (Score:5, Funny)
Super Computer (Score:1)
Sounds to me like a giant USB-Hub to create a Beowulf-cluster of these [slashdot.org].
And before somebody asks: Yes, it DOES run Linux.
Lovely. (Score:1)
Future servers will need 3456 ports... (Score:1)
Blue Gene Vs. Constellation (Score:5, Interesting)
This is hardly a fair comparison. IBM installed a 131,000 core BG/L 2 years ago, and it's been running customer code for more than a year. The sun system won't be built until late this year, and probably won't be running real customer code until this time next year. Furthermore, the BG/L machine is designed with a low-power node, assuming that a larger number of cores would be used. In IBM's older BG/L design, there are 2048 cores in a rack. Sun is packing 768 opteron cores in a rack. So a per square-meter measure gives IBM's 3 year old design only a 20% disadvantage to Sun's not-yet-released machine.
All of that is moot, of course, as theoretical peak performance is a crappy way to measure supercomputer performance anyway. The opteron is a great processor, and infinaband is a decent, though not remarkable interconnect. I'd be a little concerned, were I to buy the sun solution, that the infinaband bandwidth is being shared by 16 processor cores. That's quite a bit less interconnect performance per processor than IBM's Blue Gene, power5, Cray's XT, or SGI's altix. There's certainly plenty of memory on each of these constellation blades. That said, there are a list of applications that perform very well on Blue Gene, and Sun has a lot of ground to make up in terms of OS, software, and establishing a relationship with the HPC customers.
It's nice to have more options, however.
Re: (Score:2)
I'd like to agree to be concerned about the bandwidth, but I'm not sure I can be. Looking at my friendly local supercomputer over the last year, I see the most processor-hours used by 64 proc jobs (33%), followed by 32 proc jobs (27%). If jobs keep fairly consistent on the number of processor cores
The look! (Score:2)
And of course, the inevitable comparison (Score:2)
Geez, Sun and IBM getting into a prick-waving contest.
Overkill? (Score:2, Funny)
Re: (Score:2)
Code named Magnum? (Score:3, Funny)
Re: (Score:1)
-Bad car analogies
-Outdated "profit!" jokes
-Dupes
etc.
Re: (Score:2)
In practice, it means that you can be less careful when preparing your startfile. Imagine screwing up one input parameter or line of code 30 years ago, that
In what capacity.. (Score:2)
The switches do sound awesome though, will give infiniband a run for their money.
Re: (Score:2)
not news (Score:2)
Great! (Score:2)
YAWN. (Score:1)
Kool (Score:2)
Hi! (Score:1)
imagine a beowulf cluster of these ... (Score:2)
Whooshhh... (Score:5, Funny)
Sun Super Computer (Score:5, Funny)
There ya go
Re: (Score:2)
Beowulf rules (Score:2)
Re: (Score:1)
Re: (Score:2)
Re: (Score:2)
Re: (Score:1, Informative)
(OT) Re:Finally! (Score:1)
Just following convention (Score:1)
http://en.wikipedia.org/wiki/SI_prefix [wikipedia.org]
Re: (Score:1)