Apple is developing a distributed/parallel computing solution called Xgrid, according to ThinkSecret. Apple recently put up a mailing list, “xgrid-users”. According to the page, the mailing list is for discussion regarding use and deployment of Xgrid. No archived messages exist. The page describes Xgrid as “Apple’s solution for parallel and distributed high performance computing.”
Xgrid: Apple to Take on Parallel and Distributed Computing?
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
sure, there are Grid computers at IBM and HP, but those are research and are far from ready for a product…Apple is now stepping in so they must be confident that an Apple grid can help office and home users…if they succeed, they will be selling more grid computing solutions than IBM or HP (just like Unix based OSs, but these are because of the consumer base {number wise} for apple is larger than the consumer base for IBM or HP etc)
It’s nice to see Apple moving into markets they couldn’t before because they simply didn’t have the operating system necessary to do so. Now with a POSIX compliant operating system they are making moves into areas traditionally dominated by Unix vendors, and they have a number of technologies to simplify administration of networked systems, most notably Rendezvous. Perhaps the best example I’ve seen so far was the Rendezvous-enabled RenderMan, in which you could choose via point and click which nodes of a RenderMan cluster you wished to use. I assume Apple will apply the same sort of technology and interface to XGrid, greatly simplifying configuration and operation of clusters.
Sun also has GridEngine for MacOSX (as well as many other platforms):
http://gridengine.sunsource.net
http://bioteam.net/MacOSX/gridengine/
In a somewhat related vein, I just read on /., that Apple’s are seemingly likely to take the No. 2 position on the Top500 supercomputer list.
http://apple.slashdot.org/apple/03/10/12/143259.shtml?tid=107&tid=1…
That fits in nicely, if they want to be perceived as a company that does distributed high perf computing.
I also found it interesting, according to the post a “1100-node dual 2.0 GHz G5 cluster” is rating at 17.6 Tflops, twice the “1152-node dual 2.4 GHz Xeon Linux cluster” is gettting (~7.6 Tflops).
Forget about it.
I completely disagree, and so would many institutions looking for supercomputing power. The most amazing thing about Virginia Tech’s Terascale Cluster is not its performance, but its ridiculously low price of $5.2 million for that kind of power. Universities and research labs typically do serious work; Unreal Tournament is way down on their list of priorities.
As pointed out for a school an apple solution isn’t to expensive, but even more important is if it is a simple plug and go solution. The people using these systems do not want anything complex. They want something they can just use, and a student can figure out in no time.
> It has engineers looking at how to develop a better looking trash can.
You just lost all credibility.
apple is going upstream into higher price and margin business. they are going after the big boys and i hope they succeed. market my word, every few points of penetration that they get in these markets will translate into lower consumer desktops. The future apple is indeed bright. They are the only “other” computer maker that has a consumer base (volumes) and an OS and hardware (G5) to address the mass market and the high-end. Apple is doing things right and i just wish i could buy more stock.
//
> It has engineers looking at how to develop a better looking trash can.
You just lost all credibility.
//
No, they just patented the TrashCan Icon. Is that better? 😛
> No, they just patented the TrashCan Icon. Is that better? 😛
a) This is not the same as “has engineers looking at how to develop a better looking trash can”.
b) Having lost the look-and-feel UI war by mistakenly relying on copyrights, Apple is relying more on patents than it did in the past. The term patent here is something of a misnomer because it really is more of a registered trademark than a patent proper. Apple is attempting to utilize what the law allows to protect the uniqueness of their designs. They are attempting to accomplish this by patenting individual elements of those designs. Right? Wrong? That’s a religious debate. But it is not “engineers looking at how to develop a better looking trash can”.
Well, their trashcan icon had to be unique enough to patent.
How do you think they got a unique trashcan icon, faries?
ryan (IP: —.dyn.optonline.net) you must be very optimistic or simply you don’t know much about it to say that APPLE = NEXT SUN/SGI.
NUMA on cheap linux (pc) or expensive SGI, Where xgrid fits?
And hmm.. the makret is not so new but highly competitive.
> How do you think they got a unique trashcan icon, faries?
Graphic artists?
You make it sound as if their software engineers are busy doing nothing but making the OS pretty and designing icons.
“NUMA on cheap linux (pc) or expensive SGI, Where xgrid fits?
And hmm.. the makret is not so new but highly competitive.”
Apple is cheaper than SGI and costs more than Linux. But as far as XGrid is concerned Apple can probably deploy a software solution that is easier to implement and deploy than either a Linus or SGI solution so I think their is room for Apple in these non-traditonal markets.
A lot of people will be looking at what VT is doing with the G5 SuperCluster and what kind of performance they are getting with that system. If Apple can make clustering software it will mean more hardware sales.
Its nice to see that Apple really is concerned about their maketshare and is improving upon it!
Now if they can work on lowering the prices of their desktops just a weebit, I would definantly consider buying one.
How can 64 G5 towers be easier to deploy than an Origin 3000 with 128 processors in one rack straight out of the factory preinstalled with an O/S that scales? How can programming for a cluster of machines be easier than for a single system image? Sure it costs a lot more initially but maintaining a cluster is not cheap either.
clustering (as the one on VT) is not exactly the same thing as NUMA. Never heard about NUMA on OS X (but I may be wrong). NUMA can be considered as much more efficient way of clustering.
When I read that Virginia Tech was going to make a “how to” document available I realized that this is a pretty signifigant event. The G5 cluster, while made of computers not intended as blades, is still reaching the #2 spot on cluster speed. 17 terraflops for $5 million is a new record. If you are cost conscious (and who isn’t), are you going to spend $100 million for the #1 slot, or spend $10 million for the new number one slot?
And it was kind of interesting to see that this article showed up on Aljazeera (hey, I picked it up on a newscrawler). So you’ve got to imagine that in poorer countries where ease of use and price are going to be more important than IT job security, this may become a HUGE turning point. If Apple produces a blade version of their G5 they could improve the price/performance and cooling. But this could definitely make more money than iPods.
I realized that this is a pretty signifigant event.
Why? There are already umpteen guides showing how to build a HPC cluster out of commodity systems.
The G5 cluster, while made of computers not intended as blades
I don’t see the significance. The difference between a blade and a conventional tower is merely one of chassis.
17 terraflops for $5 million is a new record.
If you think the VT cluster actually obtains that level of performance, you’re in for a big surprise next month. The figure the BBC reported and slashdot picked up on is a theoretical one.
Is this only for advanced R and D, or could this GRID be used to run a 3D modeling program? Would you need a license per CPU or just for one of them? Would the assistance of the “grid” be transparent to the user (and the program?) So if I had a few dual g5’s in the office, I could sort of harness them all together?
I bet Apple, in their usual way, are working all this out so that it is totally painless for the user. I bet it will be as simple as… run the program and it just works! Boy, do I hope so. If so, my office will become a render-farm at night, and I can produce some cool 3D worlds.
Grid computing has been around for a while, no big news there.
How is VT handling the lack of ECC memory in the G5 cluster?
How can you know your data is correct if there is no way to detect memory errors?
I assume Xserve or blades WOULD have ECC memory?
They have previously discussed this, they use error correction algorithms, no ECC RAM necessary.
> sure, there are Grid computers at IBM and HP …
It’s funny speaking about grid computing without mentioning Sun. A bit of understatement don’t you agree ?
Perhaps we should think a bit. Export controls grew up in the days of discrete computers. If a 2 Gig G5 is no longer convered by export controls and the VT Supercomputer cookbook is freely available do we really want any two-bit country that can pump $5,000,000 worth of oil out of the ground to have a super computer to model thermonuclear explosions, rocket thrust chamber hydrodynamics or whatever their evil little minds can dream up?
If I may be so bold: An easy-to-use Macintosh-based parallel and distributed computing tool was first introduced in 2001. It’s called Pooch.
Please have a look at: http://daugerresearch.com/pooch/