clustering
Interesting story (in Wired) on clustering Macs:
<a href="http://www.wired.com/news/mac/0,2125,50078,00.html" target="_blank">That's a Whole Lot of Power, Mac </a>
Now, is it possible that when Jobs told Wall Street some time ago of a plan to "address the megahertz myth", he had in mind precisely this sort of thing - some kind of massive clustering technology (maybe via gigawire) that was so easy to set up that any smart 6th grader (or his not so smart 6th grade teacher) could do it over his/her lunch hour?
This would directly combat large gaps in clockspeed at the individual chip level, and provide an interesting little "network effect" in school or business environments: the more Macs there are in any given small area, the more powerful any additional Mac in that small area becomes... and thus the local cluster grows. And grows... and grows... until even the staunchest Windows users find it difficult to resist all of that extra power just waiting to be tapped, if only they had a Mac on their desk.
Of course, the advantage would be temporary - MS/Intel would eventually fight back with their own clustering tech. But for the 1 or 2 years it might take them to do so, there might be a powerful window of opportunity for Apple to double or triple its market share.
Thoughts?
<a href="http://www.wired.com/news/mac/0,2125,50078,00.html" target="_blank">That's a Whole Lot of Power, Mac </a>
Now, is it possible that when Jobs told Wall Street some time ago of a plan to "address the megahertz myth", he had in mind precisely this sort of thing - some kind of massive clustering technology (maybe via gigawire) that was so easy to set up that any smart 6th grader (or his not so smart 6th grade teacher) could do it over his/her lunch hour?
This would directly combat large gaps in clockspeed at the individual chip level, and provide an interesting little "network effect" in school or business environments: the more Macs there are in any given small area, the more powerful any additional Mac in that small area becomes... and thus the local cluster grows. And grows... and grows... until even the staunchest Windows users find it difficult to resist all of that extra power just waiting to be tapped, if only they had a Mac on their desk.
Of course, the advantage would be temporary - MS/Intel would eventually fight back with their own clustering tech. But for the 1 or 2 years it might take them to do so, there might be a powerful window of opportunity for Apple to double or triple its market share.
Thoughts?
Comments
WRAP YOUR HEAD AROUND THIS.
> You want more power? Buy another mac.
> Need more PCI slots? Buy another mac.
> Need more drive bays? Buy another mac.
> Still need more power? Buy another mac.
> Want more ports? Buy another Mac.
> You want all of these things at the same time? Buy another mac.
You get the point. Connected macs would share their resources but function as a single computer -- 1 keyboard and monitor setup. It's the equivalent of that shampoo company adding "repeat" to their hair washing instructions and instantly doubling their sales.
Mhz doesn't matter anymore. The only real drawback is the additional space needed. Just imagine the testoserone overdose you would have stepping back to see 4 (or more) dual 1 Ghz powermacs rack mounted, bus-shared, and cranking out your maya renders and video compressions!!!
The interoperability of clusters containing G3's and G4's and different versions of Mac OS, is REALLY intriguing though. If you can cluster iMacs, you can upgrade every few years and instead of trying to sell your old one, just ethernet it to the new one....a cheap way to bridge the megahertz divide.
Has anyone tried clustering a bunch of iMacs?
How bout all those Cubes that were so expensive because they had limited expansion capacity? We had speculated a year ago about Cubes beibng wireless "nodes" in renderfarms, has anyone clustered them?
<strong>One of the main suggestions I had was the ability for powermacs to automatically cluster when attached via a hypothetical "bus sharing" line.
</strong><hr></blockquote>
Actually, IIRC, there were some SGI workstations which even took this one step further: They had an external connector connected to their system bus (FSB) which could be interconnected (i.e. you then had one multi-processor computer spread through two enclosures).
Bye,
RazzFazz
<strong>
Actually, IIRC, there were some SGI workstations which even took this one step further: They had an external connector connected to their system bus (FSB) which could be interconnected (i.e. you then had one multi-processor computer spread through two enclosures).
Bye,
RazzFazz</strong><hr></blockquote>
Was this the NUMAflex technology or am I messing it up again? Either way, SGI has some _great_ technology and it'd be cool if Apple did that "SGI for pro shops, Apple for Joe" thingie. I think that any "advanced" tech like the SGI one would add up to general cost of the computers in a way to actually scare away people who don't need it.
Photoshop - see Demo, see Demo run. See Demo crush Pentium computers with clustering set up in 10 minute SeyBold Demo. Only available in Photoshop X.
Maya - drool. Again, trying to sell lots of boxen to the serious graphics artist crowd...
iDVD - imagine how much faster you could encode your DVDs with iDVD clustering turned on. 300% as fast as a Pentium! Not! Try my iMac lab at school now encodes 5000% faster than a Pentium.
:-)
<strong>Clustering is a red herring and not worth Apple's time. Fact is most people don't need that much power. The ones who do buy fast cheap Linux boxes. It's just not worth the effort.</strong><hr></blockquote>
that's kind of the point, cheap linux boxes end up being anything but when you need people with the technical skills to set up and maintain such a cluster... larger initial investment in apple, but you save money in the long run, or so the theory goes <img src="graemlins/hmmm.gif" border="0" alt="[Hmmm]" />
Riddle me this
Command line is a red herring and not worth Apple's time. Fact is most people don't need that much power. The ones who do buy fast cheap Linux boxes. It's just not worth the effort.
<img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
maybe it's just that the people who could put this technology to use don't have access to a less expensive, user friendly, and more viable alternative
[ 01-30-2002: Message edited by: janitor ]
[ 01-30-2002: Message edited by: janitor ]</p>
In short, clustering have ENORMUS potential to be realyl useful in some specific places.
Does anyone know how effective, if at all, clustering with Pooch is if done over the internet on say a 384 sdsl line?
The main drawback is that we need cluster-enabled software. I can cluster my mac with all of yours right now if we have the Pooch software installed on our comps. We all already have a TCP/IP connection to one another. Here's the install page I picked up from the linked PDF page on the wired article.
So who wants to start The Official AI Cluster Farm?!? More importantly, what would we do once we start it?
maybe then we'd get more reliable rumours?
just make a simple 2-4 node cluster and you instantly have super power in your 3-d/animation stuff, best thing is you start with one computer which allows you to get X amount of work done=Y profit then you get a 2 node network going with Y profit you now have 2 computers doing one task doubling your productivity and if your keen and savvy doubling your profit, this is assuming you use your mac, to make money doing graphics work(i.e designing posters, ads and the likes for web companies) or something like that
It could be totally awesome
personally I would be more than happy having even 1 dual ghz g4, cause the most advanced program I use in graphics is carrara, and I'm sure a dual ghz g4 would handle carrara just fine(as well as bryce 5)
anyone with a mac, whenever they log on to the net, will be logged into a Mac cluster server (although any mac user can choose to not be involved, however, then, they won't feel the benefits). When you have a big job that will take your computer some time to do, it will send part of it out to a couple random mac users with free resources. and, of course, when you have free resources, you will receive other people's jobs.
okay.. the current speed of broadband would probably make this impossible.. but this could be the future!
as far as more traditional clustering that you guys are talking about.. it would be great.. and a great move for apple!.. right now, when people buy new computers to replace their old.. they're doing just that.. replacing it.. usually throwing away their old computer or selling it for close to nothing.. but if apple added the ability to cluster, we could still use our old computers. i think this would actually help apple because people wouldn't feel they needed to wait as long to buy new computers!
one last note.. i'm not sure about this, but i think if they could add the ability to cluster into macOS, individual apps wouldn't need to be programmed specifically for it.. basically, the macOS scheduler would handle all of it itself.
in time, we shall see
[ 01-30-2002: Message edited by: confirmed ]</p>
this would be a massive "screw-you" to XP. every lab and graphics house in the country would want that kind of power. i just hope gigawire is a fiber channel networking/clustering technology!
<strong>here's my unrealistic idea for a mac cluster... with the help of broadband.. use the internet!.. something i like to call Anonymous Clustering.</strong><hr></blockquote>
That would be a security nightmare!!!
here's my unrealistic idea for a mac cluster... with the help of broadband.. use the internet!.. something i like to call Anonymous Clustering.
Originally posted by Fluffy:
<strong>
That would be a security nightmare!!!</strong><hr></blockquote>
Are you sure?
The idea of anonymous clustering (ubiquitous distributed computing) is soooooo compelling that I suspect some kind of workable security model will be found.
The *nix underpinnings of OS X would certainly make it easier - protected memory, file access permissions, etc.
Of course viruses (virii?) can override such protections, but to my knowledge the number of *nix viruses is extremely small, and any vulnerabilities they expose are quickly patched.
The idea of a global Mac collective (while beyond current wireless tech for bandwidth reasons) is so seductive that there would be a huge incentive to develop a viable security model.
But in the short term, security won't be such a big deal, because the clusters will all be wired, within single organizations (schools, businesses, labs, the well-heeled enthusiast's den).
and roll with this thought
suppose of the millions of mac(10 million or so) in the US that 1 million were clustered together, further suppose that this would be feasible and sensible(given the cool things unix and os x offeR)your pc friend comes up to you and says "my computer is 3 ghz using rambus..etc" you say "my computer is about 14 terahertz, then walk away grinning
lol
that would be sooo cool
now back to reality where this is impossible
This is where all those/you unix guys can start really contributing to the cause!!!
OS level clustering would be great....I wouldn't want to have to wait for even Adobe to get around to supporting clusters.