In reading Apple's site on Xserve, they talk about harnessing all the power of all the Xserve's in a rack to get 630 Gigaflops of power. That's fine and all, but is it just a marketing gimmick, because my understanding of clustering is that it's usually pretty useless except for software specifically written for it.
Yes I can have 42 servers each running, and I could have an application server bouncing things around on a cluster, but that's it isn't it?
Or does Apple have some fancy way of "harnessing" this power for normal tasks so a single server could be an entry point and use excess resources from secondary servers? That doesn't sound feasible and network I/O would bottleneck it.
This is what they say on their site:
"Now imagine harnessing the power of up to 84 processors in a 42U rack that packs a 630-gigaflop wallop"
"The dual 1GHz PowerPC G4 processors provide 15 gigaflops of computational power per 1U. That means 42 Xserve units in a 42U rack can deliver up to an astonishing 630 gigaflops of processing power."
To me it just seems like marketing gimmick, it's not like a Sun server with 64 cpu's in it where it actually can use all the cpu's for different processes... here, unless the software does clustering, you've just got 42 different servers.
Please correct me if I'm wrong and teach me something about clustering if you can. I'd also like to know if Apple does have some fancy clustering system built into Mac OSX Server that makes it beneficial to have more than one of these servers working together, because usually it doesn't make a difference with servers, they each run their own process -- and that's it... with the exception of clustering of app servers etc.
Ben
Yes I can have 42 servers each running, and I could have an application server bouncing things around on a cluster, but that's it isn't it?
Or does Apple have some fancy way of "harnessing" this power for normal tasks so a single server could be an entry point and use excess resources from secondary servers? That doesn't sound feasible and network I/O would bottleneck it.
This is what they say on their site:
"Now imagine harnessing the power of up to 84 processors in a 42U rack that packs a 630-gigaflop wallop"
"The dual 1GHz PowerPC G4 processors provide 15 gigaflops of computational power per 1U. That means 42 Xserve units in a 42U rack can deliver up to an astonishing 630 gigaflops of processing power."
To me it just seems like marketing gimmick, it's not like a Sun server with 64 cpu's in it where it actually can use all the cpu's for different processes... here, unless the software does clustering, you've just got 42 different servers.
Please correct me if I'm wrong and teach me something about clustering if you can. I'd also like to know if Apple does have some fancy clustering system built into Mac OSX Server that makes it beneficial to have more than one of these servers working together, because usually it doesn't make a difference with servers, they each run their own process -- and that's it... with the exception of clustering of app servers etc.
Ben