I predict a G5 Bladeserver at MWSF
I have no sources (unless you count the usual rumor sites) but I feel confident in predicting that Apple will announce a 3U bladeserver at MWSF.
The features I expect are:
I believe this is the solution that the long rumored XGrid technology has been waiting for. Apple will present it as a "supercomputer in a box".
Oh and yes, there will be speed bumped G5s also. The main thing there is that the G5 pro towers will all get ECC memory support and faster DDR.
The features I expect are:
- Up to 6 blades per chassis
- 1-2 2.0 GHz G5 CPUs per blade*
- 1 160 GB HD per blade*
- Up to 8 GB 533 ECC DDR per blade
- XGrid software solution
- Redundant power supplies, fans, etc.
- Blades are independantly addressable, manageable
I believe this is the solution that the long rumored XGrid technology has been waiting for. Apple will present it as a "supercomputer in a box".
Oh and yes, there will be speed bumped G5s also. The main thing there is that the G5 pro towers will all get ECC memory support and faster DDR.
Comments
I remember reading an article on Macrumors that someone in education pressed Apple about a G5 xServe, and they were told it was 4 months away (this back in October or November). I think the xServe with a G5 will come first, then maybe the 3U blade server.
IBM JS20 blade servers
IMHO it wouldn't make too much sense for Apple to replicate IBM's development efforts here - I've seen the JS20 blade in person at IBM's Böblingen Research Center (it was developed here in Germany) and was told the architecture is quite similar to a G5 PowerMac - but IBM didn't yet try to run OS X on it.
I couldn't read most chip designations due to the mounted heat sinks, but I could see that an AMD8131 Hypertransport<->PCI-X bridge was used as well as two Broadcom Gigabit ethernet controllers.
Does anyone know if Apple also uses these in the G5s?
Originally posted by Michael Engel
IMHO it wouldn't make too much sense for Apple to replicate IBM's development efforts here - I've seen the JS20 blade in person at IBM's Böblingen Research Center (it was developed here in Germany) and was told the architecture is quite similar to a G5 PowerMac - but IBM didn't yet try to run OS X on it.
You're right, it wouldn't make sense, and that's not what I'm suggesting.
I'm talking about a 3U chassis that will hold about 6 blades. (Note: I just realized that I left that part out of my original post. Sorry.) Each blade having two CPUs. Why six blades? Because that would effectively double the CPU density over Apple's current 1U XServe product.
Not only is that not the same solution as the IBM Bladecenter solution you linked to, it doesn't have anywhere near the density or associated costs. Apple needs a cost competitive, middle end solution in the high density rack space market if they want to be taken seriously in the enterprise market. They can leave the high end to companies like IBM, but they need something for the middle to low end.
I think the time is right and all the pieces are in place. If not this January, then certainly this summer.
Originally posted by Altivec_2.0
I doubt apple will release any really kinda heavy duty server hardware, besides like the xserve. I don't think that the Networking/Server is there kinda business.
Then why the XServe and XRaid?
No, I'm afraid you're thinking about the Apple of the past. It's a new ball game now.
My understanding of blade-based server sales volumes is that they are currently a niche within the server market. I think that the Apple server market would need to grow in size before Apple will start extending out the product lines much further. I believe that a larger rack size (>= 2u) 4-way G5 would be introduced before a blade model.
One thing that I would like to see Apple do would be to implement SATA drives with a hardware RAID solution in the standard G5 Xserve models. The current Xserves have four drive bays, so RAID 0, 1 and 0+1 options would offer better overall performance if implemented via hardware, instead of software (the latter is currently used for the XServe G4s). You can handle RAID in software, but it hits your CPU cycles; I suppose the imporance of hardware vs. software really depends on your intended use for the Xserve.
This would not step on Xserve RAID's toes too much, as I view Xserve RAID as a solution for: a) much larger scale storage applications and b) greater levels of data integrity/security (i.e. RAID 5).
Originally posted by Chagi
My understanding of blade-based server sales volumes is that they are currently a niche within the server market.
I'm not so sure that's true today. But you may have a point. Apple would certainly be loath to invest the research $ if there's no chance of any real return.
I think that the Apple server market would need to grow in size before Apple will start extending out the product lines much further. I believe that a larger rack size (>= 2u) 4-way G5 would be introduced before a blade model.
But that would provide no benefit (in terms of CPU density) over dual 1U G5 XServes. Don't get trapped into thinking the G5 can't be used in a 1U case, it can. Especially at 90nm.
Originally posted by Tomb of the Unknown
I'm talking about a 3U chassis that will hold about 6 blades. ... Not only is that not the same solution as the IBM Bladecenter solution you linked to, it doesn't have anywhere near the density or associated costs.
6 blades (more like bricks) in 3U is the same density as IBM's 14 blades in 7U. Your proposal would cost more than BladeCenter because it's different and lower-volume. Why not just piggyback on the existing chassis?
Originally posted by Tomb of the Unknown
I'm not so sure that's true today. But you may have a point. Apple would certainly be loath to invest the research $ if there's no chance of any real return.
But that would provide no benefit (in terms of CPU density) over dual 1U G5 XServes. Don't get trapped into thinking the G5 can't be used in a 1U case, it can. Especially at 90nm.
I fully agree that 1u G5 Xserve is a moot point - it will happen, and soon.
Originally posted by wmf
... Why not just piggyback on the existing chassis?
Originally posted by Kurt
Why not just resell IBM hardware with a version of OS X on it?
Right, why re-invent the wheel? IBM would love to OEM their Blade Centers to Apple, and adding OSX would be the "value added" that differentiates the Apple "Blade Center" from IBM's product.
Apple and IBM announced their "partnership" when the G5 was introduced. The far reaching significance of this has not yet been realized.
The introduction of such a product, built by IBM, powered by OSX, and sold by both, would show their hand as to how far and deep the collaboration between them exists.
Originally posted by Aphelion
Right, why re-invent the wheel? IBM would love to OEM their Blade Centers to Apple, and adding OSX would be the "value added" that differentiates the Apple "Blade Center" from IBM's product.
I suppose IBM would OEM the Blade Centers for Apple if the money was right, but lets not forget that there are some significant obstacles to this happening not the least of which is whether or not Apple wants to compete with IBM in this space.
The introduction of such a product, built by IBM, powered by OSX, and sold by both, would show their hand as to how far and deep the collaboration between them exists.
It might, but that assumes that a) such a deep relationship exists and b) both parties are comfortable competing for the same server dollars.
We haven't seen anything that would indicate that Apple and IBM are partnering on anything beyond IBM supplying design and fabrication capacity for Apple's G5s so I think any talk of a real partnership (in the sense you mean it) is premature at this time.
As for the second item, IBM might be comfortable competing with Apple (after all, they stand to lose little or nothing) but I'm not so sure Apple wants to go against IBM in that space. All they would have to differentiate themselves in this scenario is service and IBM has a great rep for service and support in the enterprise whereas Apple is unproven.
As far as CPU density goes, the Bladecenter is less dense than I had realized so that argument doesn't hold. So it boils down to whether or not Apple thinks it can provide a competitive product at the right price. If Apple thinks they can deliver faster CPUs using a different form factor than IBM's Bladecenter, for instance, then the value proposition might be enough to tip the scales.
I still stand by my prediction, although I concede that the odds against it are somewhat higher than I had initially considered.
Originally posted by Tomb of the Unknown
... I'm not so sure Apple wants to go against IBM in that space. All they would have to differentiate themselves in this scenario is service and IBM has a great rep for service and support in the enterprise whereas Apple is unproven...
There is plenty of space for blades in the server rooms around the World, they make space, not consume it.
But in the market space which is what you refer to, there is also "space" for OSX powered blades (RE: Virginia Tech). Think of the G5 - OSX cluster kits that VT is putting out. Their whole 1100 CPU system in a dozen or so racks.
All that aside though, the benefits for IBM to build more blades and blade chassis is the same as fabing 970's ~ economies of scale. So I believe that IBM would be more than happy to OEM their Blade Center and individual blades to Apple.
What would Apple gain from this? Outsourcing it's ultra high end server space kit would give them credible hardware in this area without much R&D work or money needed.
Develop a launch pad for high end use of OSX.
IBM is also solidly backing Linux these days and working closely with Red Hat, so the possibility of Apple licensing OS X to IBM are slim.
The only way I see an Apple/IBM partnership on blades is if Apple feels that the pros of having an offering in that product segment outweigh the disadvantages of reselling someone elses hardware. It could be that Apple will be taken more seriously for just having a blade, and that increases the sales of their other offereings. It's kind of a stretch though.