NASA Upgrades Weather Research Supercomputer

Cowards Anonymous writes "NASA's Center for Computational Sciences is nearly tripling the performance of a supercomputer it uses to simulate Earth's climate and weather, and our planet's relationship with the Sun. NASA is deploying a 67-teraflop machine that takes advantage of IBM's iDataPlex servers, new rack-mount products originally developed to serve heavily trafficked social networking sites."

Comments

Big Question:

[Penguinisto]

...what are they doing to improve the algorithms used to calculate the results? And if they're transparent (e.g. open for public inspection) - bonus!

(yes, I know that there are only a few folks in the Human race that would even know how to read the things. That said, it would be nice to have something educational, and at the same time open for public scrutiny so as to avoid political accusation, you know?)

/P

Re:Big Question:

[CorporateSuit]

I know that there are only a few folks in the Human race that would even know how to read the things

That sounds like just the kind of challenge that can motivate any slashdotter into becoming an armchair mathemagician.

Re:Big Question:

[Ngarrang]

I would like to think that in addition to throwing more hardware at the problem, that the folks are smart enough to update the algorithms as much as they can. I would be curious, as well, to the general concepts behind their implementation.

Re:Big Question:

[bunratty]

As you may expect, making climate models more accurate is a big topic of climate research these days. You can read about the basics of climate models at Wikipedia: http://en.wikipedia.org/wiki/Global_climate_model

Re:Big Question:

[lysergic.acid]

In the Greek mathematical Forum
Young Euclid was present to bore'em.
He spent most of his time
Drawing circles sublime
And crossing the pons asinorum.*

wait, you said armchair mathemetrician, right?

*not my poem.

Re:Big Question:

[toby34a]

You know, a lot of the climate and weather prediction models are open source. You can download the source code if you want, and run it on your own PC if you have certain compilers. Some links for you for your own perusal: Community Climate Model NASA GISS Model Weather Research and Forecasting Model Regional Atmospheric Modeling System As long as you have access to a Linux/Unix machine, you can get these models yourself. If you want to contribute, you can do so. Just know that you probably need to have taken graduate level courses in numerical methods and actually get the physical terms in the model to make changes that mean something. Science in this case is rather open. People can easily download these models and make changes to improve it if they needed to (or to test sensitivity, etc).

Re:Big Question:

[Leebert]

Lots of the models are open. There's a nice site at: http://modelingguru.nasa.gov/

Re:Big Question:

[SL Baur]

No, no, no!

This is slashdot and the big questions are:
does it run Linux?
FTFA:

IBM said its new server, which runs Linux and is based on Intel's quad-core Xeon processors

W00t!

Can you imagine a Beowolf cluster of these things and can you give me a car analogy of how fast these things run?

I, for one, welcome our new IBM iDataPlex overlords.

Re:Big Question:

[Anonymous Coward]

Here ya go http://ams.allenpress.com/perlserv/?request=get-document&doi=10.1175%2FJCLI3712.1 - the first search on Google Scholar. It has at least a hundred references to work that it's based on. and this is only one model out of many. Inspect away, Mr. Public.

Re:Big Question:

[Kardos]

You must be new here.

This is slashdot and the big question is: does it run vista?

In Soviet Russia, supercomputers simulate you !

There, fixed it for you.

Re:Big Question:

[ShilocK]

I think Al Gore was involved...

Re:Big Question:

[StevisF]

Much of the simulation is CFD (computational fluid dynamics). The formulas governing fluid movement have been well established for a long time. They make a 3D grid of cells, then when something (air, water) moves over the boundary of a cell, the equations have to be calculated. The equations are computationally expensive though. You can always use a smaller cell size, which will increase the accuracy of the model. A good example of this is turbulence. It's not directly numerically computed even though it can be because it would be too much compute time. I don't know if turbulence is relevant to climate prediction, but its effect is more significant with low speed flows, so it seems possible.

I'm not a climate scientest or an expert on CFD though. I maintained a cluster that was used for CFD (indoor air flow and turbo machinery) for a number of years, so that's where I picked this up.

Re:Big Question:

[Dekker3D]

we need a +1 brilliant sarcasm.

but really.. can it play crysis? all else is unimportant!

Re:Big Question:

[S3D]

Well, I worked with numerical simulation of liquids (not water, melted semiconductors), so I had some exposition to different numerical methods of FD. There is a lot of different numerical methods, so while equations well established, numerical methods for their solutions no way are. There are multiresolution methods, methods numerically conserving different quantities, like different combination of energy, momentum, entropy etc., adaptive meshes, methods with different tricks to fight numerical instability - a whole huge zoo, which is growing all the time. Efficient method for specific simulation can increase precision or/and stability on the orders of magnitude with the same cell size and about the same computation load.

Re:Big Question:

[UncleTogie]

...or for the gamers, try this game!

Pong + Fluid Dynamics = fun! Be sure to try the sandbox mode!

Re:Big Question:

[Anonymous Coward]

The code we run on discover is currently parallelized via mpi but I/O (reading and writing files) remains a major bottleneck - though there have been some recent advances in parallel I/O.

Re:Big Question:

[Anonymous Coward]

To say they want the models to be "more accurate" assumes that they were at all accurate before, and that's a load of crap. BS climate models

Well how long?

[the eric conspiracy]

Does this mean that the forecasting simulation for tomorrow's weather will run in less than 24 hours?

Re:Well how long?

[EaglemanBSA]

Yeah, but it doesn't mean the results are any less dubious :-D

Re:Well how long?

[smittyoneeach]

No, it means that the prediction for last week, a raging financial crapflood, will have increased accuracy.

Re:Well how long?

[jase001]

no, it just means it will be in colour.

Runs Windows

[Anonymous Coward]

Does this compute with the Windows Cray? Imagine a blue screen on this puppy.

Oh wait, it does run Linux. Neverrrr mind

Re:Runs Windows

[_Sprocket_]

While there may be no Blue Screen of Death, it does appear that it is a Big Blue planet.

Bad predictions even faster!

[ohxten]

Now they can make crappy predictions even *faster*!

Re:Bad predictions even faster!

[geekoid]

Forecast, not predictions.
Learn the difference. And compared to 30 years ago, they are a lot more accurate.

Re:Bad predictions even faster!

[DougF]

I'm over 50yrs old and the forecasts are NOT any better. Weather forecasters still can't get temps within 2-3 degrees 5 days out. 7 day weather models are useless for scheduling outdoor activities. They have the same rate of change to forecasts today as they had 30 years ago. On a regular basis I see weather forecasts change by 3 to 5 degrees 24hrs out.

Doppler radar means they can track weather happening NOW, and that has significantly improved severe weather tracking, but not forecasting. Call me when you can forecast the path of a tornado two days in advance, call me when the weather guy/gal can put up a 7 day forecast and it needs no updating/changing. Otherwise, I'll stick to Summer=hot & humid, Winter=cool & rainy, Spring/Fall= darn near perfect (based on middle Georgia weather patterns).

Heck, the NHC can't even forecast the path of a hurricane more than 3 days out, and that's with supercomputers, multiple models, dozens of PhDs, satellites, aircraft, decades of statistics, etc. In the last 10 years of tracking, the cone of variability beyond 3 days has not shrunk at all, the models are STILL in violent disagreement on which ridge will be more/less influential and how fast it will move. I've read the discussions and more often than not, the forecaster uses the phrase "based on my/our judgement...", and "I/We feel..".

And last, but not least, global warming theories will NEVER be accepted at the average joe level until weather forecasting at the local level gets a whole lot more accurate. Scientist can say what they will about forecasting 1 to 2 degree changes over dozens to scores of years, but until you can prove you can forecast the weather at my house, don't try to tell me you can forecast the world's weather.

For the curious...

[Anonymous Coward]

They use SuSE Linux Enterprise Server as the OS. They use IBM's GPFS for shared storage. They use xCAT to manage the cluster. They use Cisco Infiniband for the compute interconnect.

Coincidence?

[ciaohound]

NASA Engineer: "You know, chief, I've been thinking. I bet we could just about triple the performance of this thing if we supercooled it."
Manager: "Super what?"
Engineer: "Chilled it to absolute zero, like in the large hadron supercollider. Speeds up the electrons."
Manager: "What would you need to do that?"
Engineer: "Oh, I don't know, maybe... a ton of liquid helium?"

Re:Coincidence?

[Crazyswedishguy]

Weather forecast: chilly.

Tweet

[Anonymous Coward]

NASA is deploying a 67-teraflop machine that takes advantage of IBM's iDataPlex servers, new rack-mount products originally developed to serve heavily trafficked social networking sites

Does this mean the Fail Whale might appear from time-to-time?

Re:Tweet

[JohnnyGTO]

I just saw that today, perfect timing

GPGPU

[sdemjanenko]

Seeing as 67 teraflops is going to be the new processing power for this machine, I wonder if a NVIDIA CUDA implementation has been considered. Their Tesla systems are designed for this High Performance Computing, offer a significant amount of processing power and are relatively easy to parallelize code for. I know that oil companies use these high powered systems to find locations of oil, but I guess that its less likely for weather forcasting since there is less money in it. However, it would be interesting to see these cards used for modelling hurricanes and determine their expected strength and path of travel more accurately.

Re:GPGPU

[Anonymous Coward]

There's a considerable investment in time that needs to be spent on doing this, and I think NASA's mission is more in line with getting constant results as opposed to taking a year or two to totally re-write a large, existing, validated model. .... That said, there IS some work on running WRF, a weather model (similar but not identical to climate, obviously), via CUDA. John Michalakes at NCAR is the one behind this. Check out:

http://www.mmm.ucar.edu/wrf/WG2/GPU/

    Bear in mind, that represents only 1% of the code. Most of this stuff is in Fortran and I think the Fortran interface to CUDA is still in development. Definitely interesting stuff, though!

Open models are imperative

[EmbeddedJanitor]

These are the models predicting Global Warming etc. These need to be open to peer review due to the significant impact of getting these models wrong.

Faster does not mean better. I'd rather have less iterations per day on a good model than many of a crap model.

Re:Open models are imperative

[Anonymous Coward]

What makes you think they are not? Anyone can download the both the source and the detailed documentation for any of the current or previous generation models. We use a coarse resolution, but full physics model when we teach climatology. You can go to www.ccsm.ucar.edu/models to download, compile and test on your own the current generation climate model. You may choose to reduce the resolution to shorten up the run times, but that's up to you. This openness is contrast to The Viscount Monckton of Benchley's "model" which of course is too complex for you to understand or possibly SPPI's model

Re:Open models are imperative

[pablodiazgutierrez]

I'd rather have less iterations per day on a good model than many of a crap model.

Well, as long as the simulation doesn't go slower than the weather itself. Sounds silly, but it's a relevant point.

Re:Open models are imperative

[electrosoccertux]

Who needs models? Just call the Russians and ask them.

Other Big Question: Is this NASA's job?

[EmbeddedJanitor]

Is this really NASA's job? Isn't there some other organisation in USA that does weather prediction etc?

No wonder they're not getting anywhere replacing aging shuttle fleets if they are playing with rubber ducks and earth climate modelling.

Re:Other Big Question: Is this NASA's job?

[khallow]

From the National Aeronautics and Space Act (which authorizes NASA and its activities):

(d) The aeronautical and space activities of the United States shall be conducted so as to contribute materially to one or more of the following objectives:

(1)The expansion of human knowledge of the Earth and of phenomena in the atmosphere and space;

(4)The establishment of long-range studies of the potential benefits to be gained from, the opportunities for, and the problems involved in the utilization of aeronautical and space activities for peaceful and scientific purposes;

(5) The preservation of the role of the United States as a leader in aeronautical and space science and technology and in the application thereof to the conduct of peaceful activities within and outside the atmosphere;

Re:Other Big Question: Is this NASA's job?

[lysergic.acid]

perhaps NASA conducts so much peripheral research because there's no dedicated government agency for general scientific research.

i know that we have the NOAA for atmospheric research, but perhaps there needs to be an overarching government agency for scientific research in general. NASA, NOAA, and probably NIST would be branches or departments under such an agency. and all research that is pertinent to our societal advancement, but does not have a dedicated agency such as NASA or NOAA, would be conducted under this umbrella agency.

after all, we should be funding public research into general science, not just space/weather/nuclear energy. if we want to continue to be scientifically & technologically relevant, then we need a broader scientific research strategy, as well as a government agency to coordinate this strategy between the various existing research agencies.

Re:Other Big Question: Is this NASA's job?

[vought]

Thank you for looking it up and posting it. I know most people think NASA just does airplanes and spaceships, but that'd be a sad and narrow scope for an agency that is charged with not only doing missions one we're in space, but with getting through the atmosphere in the first place.

Problem is, those airplanes and spaceships gotta get off the earth. And to do that, they often fly through weather - so yes, it's part of NASA's purview. In fact, the Marshall Space Flight Center often provides imagery of the mid-Atlantic and Caribbean regions to NOAA and other U.S. Government agencies.

NASA has also been active in trying to discover the causes of and ways to predict clear air turbulence, which can be some scary shit.

Re:Other Big Question: Is this NASA's job?

[Red Flayer]

(1),(4),(5)? What happened to (2) and (3), are they still on Mars exploring?

They must have some pretty advanced counting methods over there at NASA. Either that or they royally screwed up the 1, 2, 5 -- no 3! gag from Holy Grail.

Re:Other Big Question: Is this NASA's job?

[khallow]

Well, you know NASA is authorized to do other things than just play weatherman.

Re:Other Big Question: Is this NASA's job?

[Thelasko]

(1)The expansion of human knowledge of the Earth and of phenomena in the atmosphere and space;

Still, NASA should be launching the satellites that collect the data. NOAA should be crunching the numbers.

This only reaffirms my belief that there needs to be a massive restructuring of our government. There are 16 intelligence agencies. 16! How does the Director of National Intelligence keep track of it all?

How does one coordinate a government this big? I guarantee there are multiple government agencies working on the same projects, while some projects get left behind because no single agency has enough resources to perform it. Here's a list of agencies involved with environmental science. Some of those agencies have to overlap. I'm sure the ATF, FBI, DEA, Secret Service, and U.S. Marshals cross paths quite a bit.

What frustrates me the most is when somebody in government realizes this, but instead of fixing it, they create another government agency to make it worse.

Re:Other Big Question: Is this NASA's job?

[Lodragandraoidh]

Is this really NASA's job? Isn't there some other organisation in USA that does weather prediction etc?

No wonder they're not getting anywhere replacing aging shuttle fleets if they are playing with rubber ducks and earth climate modelling.

The parent post was not insightful.

Anything that can effect aviation and spaceflight is critical to understand. Furthermore their charter charges them with not only understanding these phenomena, but also to advance scientific knowledge for applications in these arenas.

Re:Other Big Question: Is this NASA's job?

[ScrewMaster]

There are 16 intelligence agencies. [wikipedia.org] 16! How does the Director of National Intelligence [wikipedia.org] keep track of it all?

Probably another supercomputer.

So, what about chaos?

[Anonymous Coward]

http://en.wikipedia.org/wiki/Chaos_theory

Chaos theory originated when a scientist discovered that changing the accuracy of his calculations totally changed the results of his model. Small changes completely change the outcome. Our weather today was caused by a butterfly flapping its wings in the Amazon years ago.

So, one wonders if one really needs a super computer. No matter how much computing power we have, we still don't seem to be able to predict the weather properly.

Re:So, what about chaos?

[Eternauta3k]

Even if there is some randomness, we can still try to predict short-term events (this week's forecast)

Re:So, what about chaos?

[Urkki]

Even with chaotic systems, and especially with systems where physical laws, especially conservation of energy and momentum, things are pretty predictable for certain time ahead. Chaos comes from uncertainities adding up, until finally that flapping of butterfly can be deciding factor of hurricanes direction. But once a hurricane is moving in a direction, it has a lot of energy and momentum, and it'd take a motherload of butterflies working togheter coordianted to change it's direction... Or one really big one ;-).

And that's weather. Climate is another thing, but pretty much same things apply, except in global scale and over longer times and with different uncertainities.

Forecasting chaotic systems can be summed up in one word: inertia.

MTD

[Anonymous Coward]

Great....now Hansen has more ability to mangle the data!

Technology exchange gone backwards

[Anonymous Coward]

Weren't we supposed to be getting technologies from NASA for things like social websites, not the other way around.

Too many administations

[EmbeddedJanitor]

NASA= National Aeronautics and Space Administration
NOAA=National Oceanographic and Atmospheric Administration
Administrations are just that: administrations, run by administrators - primarily to the advantage of amdinistrators. Adding another level of administration won't fix that.

It doesn't count the same way.

[Junta]

When an xhpl score says '67 teraflops' and nVidia/AMD gpus spout off about the ludicrous number of gigaflops they have, it simply isn't the same.

For example, the PS3 variant of the Cell processor claims 410 gigaflops. It's hpl score, however, would be about 6-9 gigaflops. Even the new cell processors 'only' get 200 gigaflops by xhpl count.

32-bit precision scores aren't comparable directoly to 64-bit operations.

Yeah, but will they....

[darkonc]

Use rubber duckies in the cooling pool?

$800 Billion

[Anonymous Coward]

NASA's current yearly budget is $17 Billion. It peaked in 1966 during the height of the apollo program at $32 billion (in 2007 dollars). Imagine what they could do if we gave them $800 billion!

Hey, maybe we can do that next week.

Seems there may be a limiting factor

[SpaceLifeForm]

Ducks (G&R)

It only tripled?

[treblecaster]

Is tripling the performance really that news-worthy? Either they update their hardware infrequently (every few years or more) and are falling behind everything else in the computer industry, or they are doubling or tripling every year or two, which makes this not an uncommon event. It seems like orders of magnitude would be a scale of real accomplishment - or even better - if they tripled the accuracy of the weather predictions.

Recursive...

[3.14159265]

I wonder if they take into account the heating that toy generates when making those calculations! ^_^

it requires a 3x upgrade?

[nimbius]

what, does this one run server 2003 HPC?

climateprediction.net

[itof500]

Isn't this small potatoes to the power of the distributed project climateprediction.net?

duke out

Turf War?

[SlowGenius]

Not to knock NASA (I'm rather fond of them), but what are they doing in the weather prediction biz, anyway? Last I checked, weather and climate studies were in NOAA's domain.

Cross-agency collaborations are great and appropriate, but in general I'd just as soon see NASA's budget dollars stay invested in space research.

Eeek

[Zashi]

iDataPlex? Really? I am a tester at IBM. We've just started to qualify various hard drives and IO cards for the iDataPlex systems. They're very oddly designed and in general suck. The firmware (BIOS/uEFI) is really crappy but it usually is at this stage of testing. I'm sure it will get better over time. The thing that most likely will not get better is the horrible, horrible physical design (which was specially request by Facebook). I would say the reason is unknown, but from what I've heard it's because Facebook didn't want to upgrade their racks/rails so they had IBM design servers to fit them.

There's lots of curious and pointless design features. They're almost like big-ass blades, designed to slide out of a larger outer-housing that contains the PSU and fans, but several cables and wires connect the machine to the outer-housing making it impossible to remove without also removing the outer-housing from the rack. In one variant, the pci-slot is literally in the middle of the system (imagine a card slot in the middle of your motherboard, that, when a card is inserted into it, acts as a locking bar).

All the ports are in the front of the system: vga, usb, ethernet. Except for power. Power is in the back, attached to the external shell. There are also ps/2 ports (a rarity among newer servers) but they are completely blocked by the faceplate.

My overall reaction: meh.

Enjoy it while you can NASA

[gatkinso]

As much as I hate the thought, your funding is about to go bye bye.

For proper comparison

[KiwiCanuck]

plz provide a fps spec for Crysis. ~:-)

Mangle and be famous! (and get funding)

[Anonymous Coward]

You noticed. :-)

Doesn't matter, only morons like politicians accept hockey-stick graphs at face value.

Time will tell, a.k.a. the scientific method.