XGrid? - Page 2

This. Is. Shweet. Finally I can steal my Dad's processing power for my own use.

(Does PS support it yet?)

What is a good size of files to tranmit? I would love to be able to do this for encoding DVD's and photoshop, anything that requires more then a min or so of processing. I understand at some point it won't be good because the network is too slow or there is no need for other computers, so what is the size range for data to have usefullness.

Nope, and not likely to for a while.

Xgrid is a strange beast as it currently stands - you don't fire off an Xgrid job from your application, you fire off your application from Xgrid.

I can see how one *could* do the former, but that's now how it is now.

*COMPLETELY* depends on your calculation time, file size, and network speed.

I have files that are going to be massive XML dumps, maybe .1-1GB... but they are a one-time transmit, and multiple runs can be made on each file, and the calculation time swamps the network wait. So my system will work with Xgrid, even with big files.

I just think it would be nice to have an xgrid option in the sharings prefpane that allows your spare cycles to be used by other computers. Then when an applications could use more it can just look quickly to see if it has any other computers that could help.

Is this unrealistic?

You get precisely that when you install Xgrid, a SysPref pane that lets you set all sorts of items.

And yes, sharing your computer can be as simple as a checkbox.


Eg: the simplest is Rendezvous enabled auto-connect no-password on the same LAN. If that's what you want, then one checkbox will do it for you. Someone else sets up a controller, and voila. Done.

Those are good ideas.

Hop on the Xgrid mailing list and let them know you want them. Xgrid's still in prerelease, so there's lots left to do.

Jeez mon! It's a tech *preview*!

I fully expect a lot more detail can be exposed to power users, (the info about CPUs, etc, has to be buried in there already for the CPU tach, etc) but you have to admit that the idea of a simple checkbox for such a complex technology is appealing.

Excellent rundown can be found here.

POVray on Xgrid! Woot!

NOW yer talkin' !!! 8)

Does Xcode use distcc (distributed c compiler, farms out compiles) or something else ?

I believe that this is one of Xgrid's strengths. Remember, it is designed to enable laboratories, universities, rendering shops etc make use of spare cycles in an already existing network of Macs. Thus, it will 'take over' secretaries' or accountants' computers once they have gone home. Having a screensaver means these workers don't need to remember to set anything, and, in the event that they need their computers, will retain control. My school has several hundred Macs that I'm certain are doing nothing between about 7PM and 8AM - think of the wasted cycles that will become available. Thus, you get high power distributed computing without any significant investment. As things stand, I believe Apple views this sort of system as at least as important as using Xgrid in a specially bought system like Big Mac (for which it is probably far from ready anyway).

Looking into the future, I'm sure we'll see this built in at OS level. In fact, MOSR has it coming in 10.4, though you can probably take that with a large pinch of salt. One thing I think would be an especial advantage of this would be that, if enabled, Xgrid could seek out ANY available cycles on the network, and then ask the user if s/he would mind 'donating cyles' in the background. Of course, it would have to be configured to avoid slowing the donor's system, and there would doubtless be security issues to sort out. The preview won't do this right now; you have to know the cluster's name and password to join.

In the even further future, isn't this one time where we can really say that a shift in the most basic way we design computers is possible? I'm well aware that the OS and apps and so on would require a lot of work to be clusterable, but nevertheless what about this: at some point, processor speed becomes irrelevent. As soon as you fire up Doom 6 or whatever, the OS locates all the cycles you need and allocates them to you. Same goes for your big render, your statistics project, your bioinformatics stuff, your compiles. Is this a fair way off in the future and does it require a lot of development, not least in networking? Yes to all. Is it feasible? I certainly believe so.

It is an exciting prospect.

You'd be surprised. One of the projects right now in my dept is a dedicated uber-graphics-machine that consists of a graphics rendering node on everyone's desk connected through a high-speed backbone. It's used to drive your own desktop, but if someone else wants to use your spare cycles, they're available. Also, you can share desktops (or individual windows, or individual sections of windows, your choice) and stitch them together into one big projected layout... it's quite unreal.

Granted, Xgrid is quite different, but the type of shift being talked about isn't as far off as you may think.

Sorry, but I can't buy those objections. Sure, it's not going to happen for a while, but never? These clusters are generating several hundred Ghz already, but you think there's no significant benefit? You won't always need 'highly optimized, near zero latency' and all the rest to do effective grid computing. The return is already worth it to rendering shops, so it isn't that much of a jump to gaming. Heck, I'm not claiming that this is inevitable, but I certainly wouldn't dismiss it on the grounds you chose. The history of computing is full of those who said 'couldn't be done' or 'isn't worth it' and so on. They were wrong a lot of the time.




The point of that is simply to ensure that the grid will automatically start up. Thus, clerical workers can forget it's even there, while Sys admins can go home knowing it'll kick in. Sorry, it may seem that I meant to include unused cycles while the computer is in use; this isn't possible with Xgrid AFAIK, though it might be a goal for the future.


I couldn't agree more. But I don't think these things will be limited to traditional supercomputing apps (as Kickaha suggests). I'm sure that this will evolve in ways we can't begin to predict. Of course, some will always need the fastest processor, but nevertheless wouldn't it be benficial if one could make use of others' cycles?

I will be attending an Xgrid presentation, given by Apple, on Feb 3. I'll see if I can ask them about some of the things on this thread.

Kickaha and Amorph couldn't moderate themselves out of a paper bag. Abdicate responsibility and succumb to idiocy. Two years of letting a member make personal attacks against others, then stepping aside when someone won't put up with it. Not only that but go ahead and shut down my posting priviledges but not the one making the attacks. Not even the common decency to abide by their warning (afer three days of absorbing personal attacks with no mods in sight), just shut my posting down and then say it might happen later if a certian line is crossed. Bullshit flag is flying, I won't abide by lying and coddling of liars who go off-site, create accounts differing in a single letter from my handle with the express purpose to decieve and then claim here that I did it. Everyone be warned, kim kap sol is a lying, deceitful poster.

Now I guess they should have banned me rather than just shut off posting priviledges, because kickaha and Amorph definitely aren't going to like being called to task when they thought they had it all ignored *cough* *cough* I mean under control. Just a couple o' tools.

Don't worry, as soon as my work resetting my posts is done I'll disappear forever.

If you want to see where something like this could contribute to gaming, look at where corners are currently being cut, or possibilities discarded, in the name of preserving framerates and playability.

For instance: Computers on a grid could be handed the task of pre-rendering upcoming areas, so that the transition was seamless regardless of the size of the level. Or, cycles could be spent on compute-intensive things like natural language parsers, freeing designers from fixed dialog trees at least to some extent. Or, game designers could borrow a page from systems designers and use short-, medium- and long-range AI. Right now all AI is short-range (tactical). Any medium or long term planning is calcified into scripting. Spare cycles could be used to have NPCs react "off-screen" to what the player is doing, so that the player would not necessarily have to be restricted to whatever options the designer decided to allow. Even things like the contents of containers and storerooms could be offloaded, since the access latency could be hidden behind animations of doors opening and such. This would free up precious low-latency RAM on the main machine for things that the game really needed to access now without sacrificing depth of detail.

Those are just a few options off the top of my head. I can see RPGs and character-driven games and MMORPGs benefitting especially, but a clever designer could doubtless find ways to use that extra power.

Well, Amorph and the MacserverX have by now thought the gaming thing through much more thoroughly than did I. Though I never doubted the possibility of using grid computing for gaming, at least at some point in the future, I mentioned it merely because it seemed an appropriately intensive task that might benefit. I certainly didn't think of those ideas. The prerendering and AI ideas are particularly intriguing.

I think it all goes to show that there are any number of uses that we are currently only vaguely aware of. I can only assume that the hardware costs MacserverX mentions will decrease as time goes by.

The presentation will probably be dominated by scientists, since it is taking place in a research university. I'm sure things like bioinformatics and nuclear physics will be the popular topics, but I will try to ask about rendering/gaming if I get a chance. Hopefully I'll get an idea of Apple's general plans for Xgrid.

Edit: macserverX, I was thinking that people wouldn't need extra comps at home, since they could 'find' others by rendezvous or some similar method. Of course, you would be able to block others from using your cycles if you wanted to. Maybe Xgrid will continually enlarge the 'area' from which it could efficiently accept/disperse tasks as bandwith and network speeds increase?

That's exactly the sort of thing I was thinking of (at least as much as my gaming example), though I don't for a moment pretend to know the ins and outs of implementing it.

Heh that sounds like some pretty specialized stuff......

I'll try to ask the Apple guy about that sort of thing. It sure is fun!

Actually, I remember the post you just dug up. True, It's not exactly Xgrid - but not a million miles away. Perhaps Wolf had some more advanced stuff that hasn't yet made it to Xgrid's feature set.