Apple has already proved it can deliver supercomputer technology with a cluster installation at Virginia Tech, so what does Xgrid bring to the table? In typical Apple fashion, Xgrid provides an easy-to-implement solution for two common problems in the scientific community: The lack of cheap computing power, and wasted cycles on unused desktops.
Now this is something that seems like a big plus for the Apple world. An upgrade would mean adding more nodes on your network. I hope to see Maya and Mental Ray rendering done on Xgrid. Imagine distributed processing for all of your cpu intensive ops right in your personal studio at home.
Hmm, When I get my G5 (next year?) I’ll just have to try it with my G4/Dual 867… Hey, maybe I should try it with my G3/333 laptop (yeah I know it’s not recommended, but what the he11…
Jb
This technology might create less computer waste because people can now use/ will be able to use their otherwise obsolete computers. Any old computer can potentially be put to good use.
I was thinking the same thing, it is a reason for people to keep old Macs. They already hold value like a BMW. Good for Mac users, bad for people looking for used Macs
Apple’s Xgrid website doesn’t mention this, but I wouldn’t be surprised if Xgrid required OSX 10.3. Which means that your old unused G3’s will remain unused.
Of course, not too many supercomputing centers are built one racks and racks of old Blueberry iMacs.
“I can’t stress enough how easy it was to set up this software and get Xgrid up and running. Total time invested was under an hour, including downloading the software and uninstalling it after my tests. For any scientific computing environment wishing to take advantage of unused computing cycles or wanting to achieve some form of supercomputing on a budget that doesn’t allow for InfiniBand cards and multiple racks of Xserves, this is a great application.”
—
Me personally, I’m waiting for somebody from *nix land to make the post that ease like this is evil and for people who don’t want to learn to use their computers and a real scientific user would type download the binaries and compile them and then install via apt-get because thats the way a real OS should make the user do things.
I’ll check back in 15 posts or so …
http://don.cc.vt.edu/g5sinracks/slides/IMG_2353.JPG
I was looking at Virginia Tech website for the computer and I was sort of suprised to see they are using stacks of dual 2GHz G5 towers rather than Dual 2GHz “cluster nodes”. I am sure they got a bulk rate or something but they are both priced at $3 grand on Apples site. So why the trend to use Apple desktops for clustered computing and not “cluster nodes”?
OT, but 10.3 does run on many G3s
Hi there,
does anzone know, how does XGrid function exactly, I mean there are at least two ways, how to use grid:
1. The software is not modified for using Xgrid. The software checks what computer is in idle-mode and sends a normal job on this computer. This is how Sun Grid Environment works (or LoadLeveler at IBM)
2. The software has to be modified, it should be splitted in several modules, each independent from the others, so it can compute stand-alone on a client (eg MPI). Then XGrid takes care which clients can compute MPI jobs and sends packets to them.
If the second variant is the case with XGrid, how do I code my programs? Is XCode above MPI so it only administrates the clusters, or is it a MPI replacement, so there is a new paradigm and the programs have to be heavily modified in order to take advance of XGrid?
Thanks for reading and answering
Anton
The cluster nodes were announced last week, the Virginia Tech super computer was built last year.
At the point in time Virginia Tech designed the cluster only G5 towers were available. G5 Xserves were announced only last week.
From the marketing fluff, this seems like the greatest thing apple have done in ages as far as universities are concerned (and there plenty of computers being bought by universities) – It seems to good to be getting so little coverage.
I am really hoping this turns out as good as it seems.
According to the documentation, Xgrid requires OS X 10.2.8 or later, so it will run on any Mac that can run Jaguar.
I’m certain that many at Virginia Tech are regretting not having waited for the G5 XServe, if only for the ECC RAM support. I can’t even imagine what kind of support nightmare they must be having dealing with RAM failures…
XGrid is cluster management software, akin to something like the Sun Grid Engine, although markedly different in several ways.
Certainly it could be used in conjunction with an MPI application, but it certainly isn’t MPI specific… in fact I believe Apple is yet to provide their own MPI implementation for OS X, although MPICH is free and seems to work fine…
Is there any interest on your side of the campus in implementing this technology at the U?
It looks like XGrid itself is more of a frontend using Rendezvous to browse/publish the applications as services, while the brunt of the work crunching is based on BEEP ( http://www.beepcore.org/ ).
As yet, XGrid is unable to run MPI jobs. From the XGrid documentation, it is written that XGrid “makes it possible” for you to write an XGrid plugin equivalent to ‘mpirun’ so that your MPI jobs are started on the XGrid like on a cluster with mpirun.
This being only a tech preview, i’m looking forward to the ‘real’ version when it comes. If an MPI plugin is integrated, that would make a *very* elegant computing solution. If not, that’s ‘only’ a new proprietary way of writing/running distributed applications.
According to two Mac rumor sites, Apple and Virginia Tech had an agreement already worked out, to replace the G5 desktops with X-Serves when they became available. Look for that to happen within weeks.