And generate a ton of heat! I hear they're only going to sell these things in Canada and Alaska.
😉
Dharvabinky
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
High End Server Rumors?
- Thread starter MacRumors
- Start date
- Sort by reaction score
And generate a ton of heat! I hear they're only going to sell these things in Canada and Alaska.
😉
Dharvabinky
Yeah, the price alone should have clued people in that this rumor is not true.
Look at this webpage for an example of what a server with 64 processors costs: Sun High-End Servers
Unfortunately, most people don't realize that once you scale past 4 or 8 processors, you have to go to a multi-board system. Can you imagine trying to fit 64 processors on one huge motherboard? It would be the size of a small home. So to make it manageable you break it into smaller pieces. Each system board (motherboard) now has 4 CPUs on it, and you have 16 total system boards. The next problem is that you have to make it so that all of these 16 system boards can share the same memory and all of the CPUs can talk to each other. To do this, Sun Microsystems acquired Cray (the mainframe manufacturer), and used their proprietary interconnect technology to build a cross-bar type system that allows all of the system boards to communicate with each other at many many gigabytes per second (massive bandwidth). As you can probably imagine, it's not the 64 processors themselves that cost so much money, it's all of the supporting hardware that's so expensive.
Oh, and 512GB (that's right, half a terabyte) of ECC memory isn't too cheap either... 😉
I don't mean to rain on anyone's parade, but I don't think Apple is going to ever be a competitor in the high-end server market (servers 64 procs and up). The barriers of entry are just too high, and right now it's too competitive. You have Sun, IBM, HP, and Fujitsu to compete with, and they've been making these servers for years.
In the Unix workstation market, however, I see Apple taking a big chunk out of all the companies above.
Rocketman
macrumors 603
Re: High End Server Rumors?
Even if all they did was pre-configure a set of dual G5 Xserve's in cluster mode in a rack, it would qualify to fulfill this rumor. That is what these folks use to perform these complex calcs.
Servers are used for web pages.
If only Apple released a blade format dual they would be in a position to compete in the 3U 8 board market (24 CPU/3U vs 6/3U now.)
If only Apple made a quad processor motherboard they would double the throughput density of a G5 X-serv. The relatively simple possibilities are right there. Going too far afield and making a Big Blue like dedicated 100 processor + box seems too far afield for Apple marketing wise. Unless IBM were to sell it.
Rocketman
Even if all they did was pre-configure a set of dual G5 Xserve's in cluster mode in a rack, it would qualify to fulfill this rumor. That is what these folks use to perform these complex calcs.
Servers are used for web pages.
If only Apple released a blade format dual they would be in a position to compete in the 3U 8 board market (24 CPU/3U vs 6/3U now.)
If only Apple made a quad processor motherboard they would double the throughput density of a G5 X-serv. The relatively simple possibilities are right there. Going too far afield and making a Big Blue like dedicated 100 processor + box seems too far afield for Apple marketing wise. Unless IBM were to sell it.
Rocketman
scan300
macrumors 6502
Despite conflicting reports, bigger servers are still on the cards
Remember SJ's introduction of the Xserve? He said it was a modest introduction into the server market. I don't have his exact wording, but he left me with the impression of a much bigger plan is in the offing.
Apple spend a few years planning and designing their products with a lot of prototypes left on the cutting room floor. (Remember Ive's first attempt at an LCD iMac was ditched and a new 'sunflower' version developed.)
I'm sure Apple have a range of machines, that are wildly souped-up and blow up in 3 minutes.
You can't just follow the pack if you're Apple, so a rumour such as this 64 cpu thing is plausible, if someone was privy to what goes on in the concept lab.
Remember SJ's introduction of the Xserve? He said it was a modest introduction into the server market. I don't have his exact wording, but he left me with the impression of a much bigger plan is in the offing.
Apple spend a few years planning and designing their products with a lot of prototypes left on the cutting room floor. (Remember Ive's first attempt at an LCD iMac was ditched and a new 'sunflower' version developed.)
I'm sure Apple have a range of machines, that are wildly souped-up and blow up in 3 minutes.
You can't just follow the pack if you're Apple, so a rumour such as this 64 cpu thing is plausible, if someone was privy to what goes on in the concept lab.
Hopefully with the internal and external expansion capabilities that the a average workstation user would use. That is internal drive bays, and externally accessible drive bays. Also it is not unreasonable for a work station user to instal raid arrays in their systems so hardware support for that would be nice. Using the same modules as XServe raid would be even nicer.
While my biggest issue with obtaining the first round G5's is $$$$$$$, I also think it is wise to wait a bit to see if they come out with something a bit more expandable. I'd even be tempted by an XServe chassis if it supported hardware raid and had a decent built in graphics system or the ability to easly handle the latest AGP cards. That is no heating issues. But who knows, with a bit of overtime I might be able to trash all of the above and have my machine by october.
Dave
While my biggest issue with obtaining the first round G5's is $$$$$$$, I also think it is wise to wait a bit to see if they come out with something a bit more expandable. I'd even be tempted by an XServe chassis if it supported hardware raid and had a decent built in graphics system or the ability to easly handle the latest AGP cards. That is no heating issues. But who knows, with a bit of overtime I might be able to trash all of the above and have my machine by october.
Dave
Re: Despite conflicting reports, bigger servers are still on the cards
I think plausible is a stretch for a 64cpu machine.. this isn't just big iron, it's huge iron.
these machines use NUMA memory architectures, they have massive internal bandwidth, they cost a ton to build, and they need to be ultra redundant and ultra reliable. 64cpu machines from other vendors STILL cost a million bucks a pop.
I think it's plausible that apple is preparing bigger, badder hardware, but I'm not sure where the market is for an OS X 64processor machine.
I think apple could break in with a 4cpu or 8 cpu box if they could release it at a reasonable price point. They may market such a system as a modular computing node (think SGI Altix) that can scale to 64cpus, but this would still require a big R&D cost and high production costs.
There is a reason why customers in Apple's new markets (science and video) don't buy HP superdomes... their applications run MUCH better from a price/performance standpoint on clusters of inexpensive nodes. In fact the applications that Apple wants to target probably run much better on a pure performance standpoint on a large cluster than on a lower-cpu massive server.
I think plausible is a stretch for a 64cpu machine.. this isn't just big iron, it's huge iron.
these machines use NUMA memory architectures, they have massive internal bandwidth, they cost a ton to build, and they need to be ultra redundant and ultra reliable. 64cpu machines from other vendors STILL cost a million bucks a pop.
I think it's plausible that apple is preparing bigger, badder hardware, but I'm not sure where the market is for an OS X 64processor machine.
I think apple could break in with a 4cpu or 8 cpu box if they could release it at a reasonable price point. They may market such a system as a modular computing node (think SGI Altix) that can scale to 64cpus, but this would still require a big R&D cost and high production costs.
There is a reason why customers in Apple's new markets (science and video) don't buy HP superdomes... their applications run MUCH better from a price/performance standpoint on clusters of inexpensive nodes. In fact the applications that Apple wants to target probably run much better on a pure performance standpoint on a large cluster than on a lower-cpu massive server.
oz project
interesting. Canberra is the capital of Australia. Just heard about a project concerning a telescopic array to be located in the outback to view Southern Hemisphere stars. Supposed to be a joint project between NASA and the Australian gov't. Wonder if it's related.
interesting. Canberra is the capital of Australia. Just heard about a project concerning a telescopic array to be located in the outback to view Southern Hemisphere stars. Supposed to be a joint project between NASA and the Australian gov't. Wonder if it's related.
Exactly. The bombs should arrive in Sydney ('cause we do know that that's the real capital) in 30 minutes.
Re: I really don't see how Apple can compete
Sorry, but that is a very near sighted view of IT. It's better to think of Linux as 'free' as in intellectually free... not 'free' as in costing nothing.
If you support linux, you know it is anything but free. What you save in licensing, you make up for in configuration and administration. Sure Windows has high support costs for viruses and hacks, but Linux has plenty of vulnerabilities too and it's only as secure as your admin is skilled.
Sure, Windows is expensive when it breaks (or get's broken into), but for a lot of uses it pretty much just runs. Think about how much support $ you'd spend on a typical secretary if she had a linux box on her desktop instead of a Mac or XP machine.
Cluster (well) isn't easy. It takes someone, or a full staff, who knows their s*it and it takes time and effort to get it running well... and to keep it running well.
I think the bottom line is, linux is more robust and it is technically a better clustering platform than windows. You don't see a huge clustering initiative from Microsoft, even though linux clusters eat into the sales of Windows Server Enterprise edition (for big smp boxes).
There are a lot of rabid MS supporting developers out there but if you want clustering projects, you look to linux.
ux.
I think the bigger reason (than licensing) is that the people savy enough to build 1000 node clusters for huge corporations know that they need the best tool for the job.
jmho...
Sorry, but that is a very near sighted view of IT. It's better to think of Linux as 'free' as in intellectually free... not 'free' as in costing nothing.
If you support linux, you know it is anything but free. What you save in licensing, you make up for in configuration and administration. Sure Windows has high support costs for viruses and hacks, but Linux has plenty of vulnerabilities too and it's only as secure as your admin is skilled.
Sure, Windows is expensive when it breaks (or get's broken into), but for a lot of uses it pretty much just runs. Think about how much support $ you'd spend on a typical secretary if she had a linux box on her desktop instead of a Mac or XP machine.
Cluster (well) isn't easy. It takes someone, or a full staff, who knows their s*it and it takes time and effort to get it running well... and to keep it running well.
I think the bottom line is, linux is more robust and it is technically a better clustering platform than windows. You don't see a huge clustering initiative from Microsoft, even though linux clusters eat into the sales of Windows Server Enterprise edition (for big smp boxes).
There are a lot of rabid MS supporting developers out there but if you want clustering projects, you look to linux.
ux.
I think the bigger reason (than licensing) is that the people savy enough to build 1000 node clusters for huge corporations know that they need the best tool for the job.
jmho...
Re: Why 970s?
I agree that using 970's for high-end server applications is kinda silly. The key differentiators (way simplified) between Power4 and 970 are these:
1) 970 lacks the dual cores and related circuitry of the Power4
2) The interprocessor interconnects have, IIRC, been simplified for the 970
3) The 970 is designed for speed over stability; the Power 4 is the opposite
4) The 970 includes Altivec.
Seems like the only reason Apple might put 970s into a high-end server would be support for Altivec (which might be important for, as an example, Renderman). But you lose so much by doing such ... seems the more "sensible" approach would be to just convince IBM to slap Altivec onto a Power5 (it's that easy, right? 🙂 )
The 970 is NOT a high-end server chip. It is a low-end server, high-end workstation and desktop chip.
I agree that using 970's for high-end server applications is kinda silly. The key differentiators (way simplified) between Power4 and 970 are these:
1) 970 lacks the dual cores and related circuitry of the Power4
2) The interprocessor interconnects have, IIRC, been simplified for the 970
3) The 970 is designed for speed over stability; the Power 4 is the opposite
4) The 970 includes Altivec.
Seems like the only reason Apple might put 970s into a high-end server would be support for Altivec (which might be important for, as an example, Renderman). But you lose so much by doing such ... seems the more "sensible" approach would be to just convince IBM to slap Altivec onto a Power5 (it's that easy, right? 🙂 )
The 970 is NOT a high-end server chip. It is a low-end server, high-end workstation and desktop chip.
Umm, no. Blades are smaller.
http://www-1.ibm.com/servers/eserver/bladecenter/blade_servers/
ftp://ftp.pc.ibm.com/pub/pccbbs/bp_server/why_blade_servers.pdf
There's a link to some blades. Note that blades slide into a 3U, 6U, or 7U rack mount vertically, fitting 12-16 across. All blades in a chassis share a single power supply and cooling system and management/networking modules.
A 1U server, such as the XServe, has fully self-contained power, network, and management systems. This leads to multiple cables to each 1U server and relatively slower communication amongst multiple 1U server processors than between blades.
Top price is half a mil
First time I read the Bidouille article I also read $50,000. Looking again and it's $500,000. Logical when you think of 64 processors with 16 Gig of memory for each. A top price of $50,000 would mean about $750 for each processor matched to 16 Gigs - we ain't there yet!.
I think we will see them, either when Panther is released or when the 980 is ready. There is no way Apple would invest the development resources to get prototypes "running well" in all configurations if they were not going to go for it. There is simply too much pressure in the other areas - G5 PM, new PBs, refreshing the consumer lines, etc. to put a major effort into this project as a whim.
This is going to be one hell of a year for Apple!
First time I read the Bidouille article I also read $50,000. Looking again and it's $500,000. Logical when you think of 64 processors with 16 Gig of memory for each. A top price of $50,000 would mean about $750 for each processor matched to 16 Gigs - we ain't there yet!.
I think we will see them, either when Panther is released or when the 980 is ready. There is no way Apple would invest the development resources to get prototypes "running well" in all configurations if they were not going to go for it. There is simply too much pressure in the other areas - G5 PM, new PBs, refreshing the consumer lines, etc. to put a major effort into this project as a whim.
This is going to be one hell of a year for Apple!
scan300
macrumors 6502
Re: Re: Despite conflicting reports, bigger servers are still on the cards
I agree about the practicality of Apple bringing such a product to market in the near to mid term future. But I still think it's possible for Apple to be developing concept machines in this category.
R&D for product development that's ready to go to market is different to R&D for new concepts. Lots of different industries spend some resources in developing concept products that fall into the 'can it work?' category. They are too difficult to bring to market in the short to medium term, but they know they can make one, (eg Honda's Human Robot project). These concepts drive the company's forward planning; laying out options to respond to marketplace opportunities, if they arise. A lot of these concepts never see the light of day, or sometimes patents are filed and shelved for a later date.
So my thought on this rumour is that its origins may come from the concept lab and/or a mixture of here-say. I obviously can't say the rumour is true, but I can't rule it out.
I agree about the practicality of Apple bringing such a product to market in the near to mid term future. But I still think it's possible for Apple to be developing concept machines in this category.
R&D for product development that's ready to go to market is different to R&D for new concepts. Lots of different industries spend some resources in developing concept products that fall into the 'can it work?' category. They are too difficult to bring to market in the short to medium term, but they know they can make one, (eg Honda's Human Robot project). These concepts drive the company's forward planning; laying out options to respond to marketplace opportunities, if they arise. A lot of these concepts never see the light of day, or sometimes patents are filed and shelved for a later date.
So my thought on this rumour is that its origins may come from the concept lab and/or a mixture of here-say. I obviously can't say the rumour is true, but I can't rule it out.
eric67
macrumors 6502
space in G5 enclosure
Maybe not related directly with subjet of 64processor based machine, but did someone also noticed the space left inside the G5 anclosure??
I mean it is easily possible to fit at lest something in front of the 2 G5 processors covered with their alu box???
OK there are 2 fans in between but really a lot of space???
Maybe not related directly with subjet of 64processor based machine, but did someone also noticed the space left inside the G5 anclosure??
I mean it is easily possible to fit at lest something in front of the 2 G5 processors covered with their alu box???
OK there are 2 fans in between but really a lot of space???
pretentious
macrumors regular
Blades
We use compact blades like that too. I've seen them. They're sweet. Apple needs to come up with a chassis / blade design like that, so it can fit 8-16 proc to a blade, and then 8 blades to a box. Then the UNIX based server OS needs to mature a *LOT* including becomming POSIX compliant, as well as a far better RAID system than XRaid, then the boxes could be something we might actually be able to use. However, the HW may get there in the next two years, the OS is the part I fear will never meet the needs of my work... Go POSIX. Please.
Jaedreth
We use compact blades like that too. I've seen them. They're sweet. Apple needs to come up with a chassis / blade design like that, so it can fit 8-16 proc to a blade, and then 8 blades to a box. Then the UNIX based server OS needs to mature a *LOT* including becomming POSIX compliant, as well as a far better RAID system than XRaid, then the boxes could be something we might actually be able to use. However, the HW may get there in the next two years, the OS is the part I fear will never meet the needs of my work... Go POSIX. Please.
Jaedreth
Eric67, Space
Yes, there is a lot of space, but if you look close, the "empty space" has the ram slots, so you can't put processors there. That empty space is for airflow. If they cram it too packed, then the air will not flow. If they do a quad on the G5, they would have to rearrange it a bit. Either that, or just make it larger. Certainly the fabled Workstation units will be considerably larger.
Jaedreth
Yes, there is a lot of space, but if you look close, the "empty space" has the ram slots, so you can't put processors there. That empty space is for airflow. If they cram it too packed, then the air will not flow. If they do a quad on the G5, they would have to rearrange it a bit. Either that, or just make it larger. Certainly the fabled Workstation units will be considerably larger.
Jaedreth
Re: Blades
a blade is typically about the size of a big PCI card. You are asking for an entire computer on a PCI card with 8 - 16 processors on it. If you ignore everything else and just think about the power consumption of the cpus...
2GHz 970 @~45 watts x 16 processors x 8 blades per enclosure = 5760 watts just for the CPUs. haha... the paint would burn off the blade backplane enclosure and the PCBs would melt.
After the CPUs, you typically need to add a small Hard drive, ram slots, and all the controller chips (and assorted electronics).
Blades can handle 1 or 2 cpus per blade. I've heard rumors of quad processor blades but I've not seen them for sale anywhere. A quad Blade would need to use some low power CPUs anyway.
Also, do you realize that being POSIX compliant means that it supports standard Unix conventions? OS X IS functionally POSIX compliant, but Apple does not choose to apply for official POSIX compliance from the licensing group because a) it costs a lot, like a couple hundred grand every time you submit for review, and b) it takes a long time to go through the certification process.
Apple is very clear to its developers that though OS X doesn't official have the POSIX badge, it is nearly entirely POSIX compliant (it actually may be POSIX compliant but they don't bother with certification).
If Apple bother with official POSIX compliance, you would see much longer spans between major system upgrades.
hehe.... that's entertaining...
a blade is typically about the size of a big PCI card. You are asking for an entire computer on a PCI card with 8 - 16 processors on it. If you ignore everything else and just think about the power consumption of the cpus...
2GHz 970 @~45 watts x 16 processors x 8 blades per enclosure = 5760 watts just for the CPUs. haha... the paint would burn off the blade backplane enclosure and the PCBs would melt.
After the CPUs, you typically need to add a small Hard drive, ram slots, and all the controller chips (and assorted electronics).
Blades can handle 1 or 2 cpus per blade. I've heard rumors of quad processor blades but I've not seen them for sale anywhere. A quad Blade would need to use some low power CPUs anyway.
Also, do you realize that being POSIX compliant means that it supports standard Unix conventions? OS X IS functionally POSIX compliant, but Apple does not choose to apply for official POSIX compliance from the licensing group because a) it costs a lot, like a couple hundred grand every time you submit for review, and b) it takes a long time to go through the certification process.
Apple is very clear to its developers that though OS X doesn't official have the POSIX badge, it is nearly entirely POSIX compliant (it actually may be POSIX compliant but they don't bother with certification).
If Apple bother with official POSIX compliance, you would see much longer spans between major system upgrades.
Apple and IBM
Well, here are a few things to keep in mind:
IBM's major cash cow besides enterprise server software and services, is enterprise servers, aka AIX and those wonderful high end servers that are not available to anyone but corporations (and then those that are the strict domain of the US Government). Trust me, a good friend of mine does Level 3 AIX support, Install/Restore.
Sure, IBM is getting *back* into Workstations. They had dropped them. But notice what their workstations are running on. They are actually Intel based. IBM is trying to keep up with the market while trying to *develop* a new market.
I'm sure IBM would love to have *nothing* to do with Intel, their competition for chips. I'm sure IBM would love to have *nothing* to do with Microsoft. And I'm sure they would want to help Apple make that happen.
Sure, use of IBM Smartdrives (which was consequently sold to Hitachi) showed some semblance of Apple's and IBM's in-bed-ed-ness, but not much. This new G5 architecture with Apple system controllers being fabed *by* IBM shows this to a greater detail.
Such rumors of IBM assisted high end workstations from Apple sounds like good business for Apple and IBM. Especially since Apple also needs help in the Servers market. These high end workstations should not be mistaken for servers. Sure these machines will be more awesome than XServe, by a good bit.
That means Apple needs to get *serious* about servers, and go to using 1U Blades especially in its cluster configurations. The main server that controls the cluster can be a souped up xraid, but Apple needs Blades, true to form.
Also Apple need a far better RAID solution. Apple needs to be able to put a lot more storage in one XRAID box. Apple needs to use smaller drives that hold more, and thus fit more in. Even if it takes up U4 or U5 to do so, if Apple wants to be a serious contender in the low end server market (competing with GUI based servers, not unix based, ie, with MS), each XRAID must be able to handle 10TB, and as a base configuration of fully stocked. (eg, all drives in place, but with the lowest drive config for XRAID, so you can still order an XRaid with less than 10TB by not filling it up, but you can also get much more by using larger drives.)
I know that sounds *extreme*, but we here where I work wouldn't consider them otherwise. As for the Blades, they would have to have a minimum of 8-processors per blade, and a minimum of 8 blades to a 1U. Because that's what we're using. (And no, you can't find them to buy publicly, only available to corporations like the one I work for. And they're made by Sun, and larger than a PCI card, but thin.) Then it would still have to be cheaper *and* for our software and OS to work on it. We are Unix based, but OS X is simply *not* a mature enough UNIX based OS for us to do with OS X what we do here.
That's why I'm so adamant that Apple must have *better* equipment than the competition for *less*, as well as a fully 64-bit and POSIX compliant operating system that doesn't have all the core portions of the OS locked down. (I don't care about certification, I care about having 100% of the tools any stock UNIX system would have, which just isn't so with Mac OS X. Furthermore, any open standard unix OS is gonna have security issues, and we use proprietary code to lock down everything as tight as AIX, or even more so, but we don't have to pay for it, we do it ourselves.
However, if Apple works with IBM on *all* of these concerns, Apple could eventually do quite well in the low end servers market, and the high end workstations market, which would make IBM all the happier.
It's going to take time. However, if Apple builds it, they will be bought. Why? Visualize this a moment. If an IBM workstation is running Windows Server software and using an Intel chip, but IBM is making and selling it, are they going to make more money on that, or on the Apple high end workstation / low end server, especially if there are licensing issues for collaboration with OS?
"Hmm, let me get you over to Apple Enterprise sales. I'm sure they would be far better equipped to meet your needs than we can."
(Since IBM can only sell its servers to Corporations, not even small businesses, Apple would be poised to take on small and medium businesses, as well as the entire professional arena.)
Is this a pipe dream? I'm sure everyone will have an opinion. However, it would take specs like these in order for this company to consider using these machines with our high end servers. This is the most state of the art facility I've ever seen, so if Apple could schmooze it's way into being used here, no one would have a reason not to at least look at the Apple server platform.
Just wait and see.
Jaedreth
Well, here are a few things to keep in mind:
IBM's major cash cow besides enterprise server software and services, is enterprise servers, aka AIX and those wonderful high end servers that are not available to anyone but corporations (and then those that are the strict domain of the US Government). Trust me, a good friend of mine does Level 3 AIX support, Install/Restore.
Sure, IBM is getting *back* into Workstations. They had dropped them. But notice what their workstations are running on. They are actually Intel based. IBM is trying to keep up with the market while trying to *develop* a new market.
I'm sure IBM would love to have *nothing* to do with Intel, their competition for chips. I'm sure IBM would love to have *nothing* to do with Microsoft. And I'm sure they would want to help Apple make that happen.
Sure, use of IBM Smartdrives (which was consequently sold to Hitachi) showed some semblance of Apple's and IBM's in-bed-ed-ness, but not much. This new G5 architecture with Apple system controllers being fabed *by* IBM shows this to a greater detail.
Such rumors of IBM assisted high end workstations from Apple sounds like good business for Apple and IBM. Especially since Apple also needs help in the Servers market. These high end workstations should not be mistaken for servers. Sure these machines will be more awesome than XServe, by a good bit.
That means Apple needs to get *serious* about servers, and go to using 1U Blades especially in its cluster configurations. The main server that controls the cluster can be a souped up xraid, but Apple needs Blades, true to form.
Also Apple need a far better RAID solution. Apple needs to be able to put a lot more storage in one XRAID box. Apple needs to use smaller drives that hold more, and thus fit more in. Even if it takes up U4 or U5 to do so, if Apple wants to be a serious contender in the low end server market (competing with GUI based servers, not unix based, ie, with MS), each XRAID must be able to handle 10TB, and as a base configuration of fully stocked. (eg, all drives in place, but with the lowest drive config for XRAID, so you can still order an XRaid with less than 10TB by not filling it up, but you can also get much more by using larger drives.)
I know that sounds *extreme*, but we here where I work wouldn't consider them otherwise. As for the Blades, they would have to have a minimum of 8-processors per blade, and a minimum of 8 blades to a 1U. Because that's what we're using. (And no, you can't find them to buy publicly, only available to corporations like the one I work for. And they're made by Sun, and larger than a PCI card, but thin.) Then it would still have to be cheaper *and* for our software and OS to work on it. We are Unix based, but OS X is simply *not* a mature enough UNIX based OS for us to do with OS X what we do here.
That's why I'm so adamant that Apple must have *better* equipment than the competition for *less*, as well as a fully 64-bit and POSIX compliant operating system that doesn't have all the core portions of the OS locked down. (I don't care about certification, I care about having 100% of the tools any stock UNIX system would have, which just isn't so with Mac OS X. Furthermore, any open standard unix OS is gonna have security issues, and we use proprietary code to lock down everything as tight as AIX, or even more so, but we don't have to pay for it, we do it ourselves.
However, if Apple works with IBM on *all* of these concerns, Apple could eventually do quite well in the low end servers market, and the high end workstations market, which would make IBM all the happier.
It's going to take time. However, if Apple builds it, they will be bought. Why? Visualize this a moment. If an IBM workstation is running Windows Server software and using an Intel chip, but IBM is making and selling it, are they going to make more money on that, or on the Apple high end workstation / low end server, especially if there are licensing issues for collaboration with OS?
"Hmm, let me get you over to Apple Enterprise sales. I'm sure they would be far better equipped to meet your needs than we can."
(Since IBM can only sell its servers to Corporations, not even small businesses, Apple would be poised to take on small and medium businesses, as well as the entire professional arena.)
Is this a pipe dream? I'm sure everyone will have an opinion. However, it would take specs like these in order for this company to consider using these machines with our high end servers. This is the most state of the art facility I've ever seen, so if Apple could schmooze it's way into being used here, no one would have a reason not to at least look at the Apple server platform.
Just wait and see.
Jaedreth
Re: Apple and IBM
**edit** choosing my words better after reading the 'don't pick on others' thread from arn **edit**
as you can see in this link, IBM still offers Unix Workstations running on hardware from the 604e to ones with the latest Power4+.
eWeek seems to think it's pretty nice in this article. They give it an Excellent in "capability" and the score it good in other categories even though they incorrectly believe that it doesn't operate with other hardware and OSes (Apple bragged about how nice it ran on Sun servers at WWDC). The only limitation I've seen to xServer RAID is the fact that it must be configured into to volumes if you use all the disks, one per FC port. Considering how incredibly low the cost is for such a nice 2.5TB RAID cabinet, I think this is entirely acceptable.
I think your critiques totally ignore the weak points of the xServe and they address non-issues. 2.5TB in 3U is NOT too little. If you knew more about RAID5, you would know that there are diminishing returns as you add more drives to one RAID 5 volume. Bandwidth becomes a big issue as the array size increases (as does controller power).
Your suggestion to use smaller (2.5"??) drives is kind of silly. Smaller drives are not available in high capacities. What's the biggest 2.5" drive? 80GB? The max speed is 7200 rpm and those are rare, the average 2.5GB drive is 5400 while many are still 4500rpm. So, if you want a very very slow, very very expensive RAID, I'm sure smaller form drives are the way to go.
Apple needs 10+ TB RAIDs to compete in the LOW END SERVER MARKET?
Huh???
The lowest price 10TB RAID case I could find runs just under $50,000 and they don't even post a picture of the thing.. they have a cheap-o rendering on the web page. This is EXACTLY what the low end server market is looking for... more crappy $50,000 drive arrays. BTW, do you have any clue what it would take to back that up?
I'm surprised Sun doesn't sell these to other companies.. they'd sell like hot cakes. That seems like REALLY bad business to me.
Apple NEEDS a 64bit OS, it needs to be POSIX compliant, and it can't have it's core locked down (proprietary, closed source).
So, in light of these pearls of wisdom you tell us all that you don't care if the OS has certifications (though this IS what POSIX is, a certification of compliance). You also tell us that your organization is so great because you practice security through obscurity (you hide your flaws).
Oh... BTW. the entire core of OS X IS open source already, it's called DARWIN. It's only been available for like 3 years now, it even runs on x86.
**edit** choosing my words better after reading the 'don't pick on others' thread from arn **edit**
When did IBM get out of the workstation market? From what I know, they've always offered Power and PowerPC workstations. They may have lagged in performance (you can only stretch a Power3 so far) but I'm pretty sure they've always been there.
as you can see in this link, IBM still offers Unix Workstations running on hardware from the 604e to ones with the latest Power4+.
IBM shipped intel chips in their desktop machine when they could have shipped AMD. IBM makes a LOT of money offering enterprise support and consultation for Windows clients. I'm not sure if I agree with your take on this. IBM is interested in making money on whatever pays... just like MS will keep making Mac products like Office as long as they turn a good profit
Who sells 1U blades? Blades are typically 4u, 7u... they typically stand up in a an enclosure (backplane). They don't achieve high computational density by packing a lot of cpus in 1U, they do it by packing a lot of nodes vertically in a multiple U rack unit.
Huh?
eWeek seems to think it's pretty nice in this article. They give it an Excellent in "capability" and the score it good in other categories even though they incorrectly believe that it doesn't operate with other hardware and OSes (Apple bragged about how nice it ran on Sun servers at WWDC). The only limitation I've seen to xServer RAID is the fact that it must be configured into to volumes if you use all the disks, one per FC port. Considering how incredibly low the cost is for such a nice 2.5TB RAID cabinet, I think this is entirely acceptable.
I think your critiques totally ignore the weak points of the xServe and they address non-issues. 2.5TB in 3U is NOT too little. If you knew more about RAID5, you would know that there are diminishing returns as you add more drives to one RAID 5 volume. Bandwidth becomes a big issue as the array size increases (as does controller power).
Your suggestion to use smaller (2.5"??) drives is kind of silly. Smaller drives are not available in high capacities. What's the biggest 2.5" drive? 80GB? The max speed is 7200 rpm and those are rare, the average 2.5GB drive is 5400 while many are still 4500rpm. So, if you want a very very slow, very very expensive RAID, I'm sure smaller form drives are the way to go.
Um, yea.
Apple needs 10+ TB RAIDs to compete in the LOW END SERVER MARKET?
Huh???
The lowest price 10TB RAID case I could find runs just under $50,000 and they don't even post a picture of the thing.. they have a cheap-o rendering on the web page. This is EXACTLY what the low end server market is looking for... more crappy $50,000 drive arrays. BTW, do you have any clue what it would take to back that up?
Can't you post a link to these mysterious Sun blade servers that Sun only makes for your company... the only Sun 1U blade server, the only one that has 8 cpus per blade, 64 cpus per 1U??
I'm surprised Sun doesn't sell these to other companies.. they'd sell like hot cakes. That seems like REALLY bad business to me.
so.. wait, let me follow this if I can. I know I don't know as much about computers as you do.
Apple NEEDS a 64bit OS, it needs to be POSIX compliant, and it can't have it's core locked down (proprietary, closed source).
So, in light of these pearls of wisdom you tell us all that you don't care if the OS has certifications (though this IS what POSIX is, a certification of compliance). You also tell us that your organization is so great because you practice security through obscurity (you hide your flaws).
Oh... BTW. the entire core of OS X IS open source already, it's called DARWIN. It's only been available for like 3 years now, it even runs on x86.
If Apple tried to follow your lead, they'd be out of business in 2 years.
yes
Re: Top price is half a mil
I would not under estimate a corporations ability to spend money on a whim. You would be absolutely surprised with the way SOME companies will spend the share holders money.
In any event I do not think this is the issue with Apple. First if we accept that a machine with 64 processors exists at Apple, there can be a number of reasons for those machines and a number of possible configurations.
The most obvious reason is for research and developement of new machine concepts. This would be one reason for the machine, it may have no future at all as a saleable machine. That doesn't mean that technology won't be developed on the mystery machine for future Macs.
At this moment I believe that the 970 is just to new for anybody to be release massive machines based on it. Maybe in 6 months to a year we may see large multi processors but I think maybe Apple has a lot of work to do on their core systems this year. Stabalizations of and expansion of sales ought ot be job one right now after the recent results posted by the company. Besides I have this suspicion that maybe GPUL2 is a more highly integreated chip, such a device could serve Apple well in several of its lines. I'm not an Apple engineer but if I was that is what I'd be shooting for, a 970 that needs little in the way of support chips for blades, IMacs and portables.
Dave
I would not under estimate a corporations ability to spend money on a whim. You would be absolutely surprised with the way SOME companies will spend the share holders money.
In any event I do not think this is the issue with Apple. First if we accept that a machine with 64 processors exists at Apple, there can be a number of reasons for those machines and a number of possible configurations.
The most obvious reason is for research and developement of new machine concepts. This would be one reason for the machine, it may have no future at all as a saleable machine. That doesn't mean that technology won't be developed on the mystery machine for future Macs.
At this moment I believe that the 970 is just to new for anybody to be release massive machines based on it. Maybe in 6 months to a year we may see large multi processors but I think maybe Apple has a lot of work to do on their core systems this year. Stabalizations of and expansion of sales ought ot be job one right now after the recent results posted by the company. Besides I have this suspicion that maybe GPUL2 is a more highly integreated chip, such a device could serve Apple well in several of its lines. I'm not an Apple engineer but if I was that is what I'd be shooting for, a 970 that needs little in the way of support chips for blades, IMacs and portables.
Dave
Sorry that this is a late entry but there is a 7 1/2 hour time change here.
Canberra ? You are not serious not knowing WHAT Canberra is ?
😛
It's unbelievable! It is like someone outside the US saying they think New York is the capital of the USA. The seat of government was created just like Wash DC out of bits of land from two states and thus sits on the edge of Victoria and New South Wales. The city's layout was designed by an American and an artificial lake was named after him: Lake Burley Griffin!
No wonder Wintel was successful in the US!!
Tears! tears! tears! 'nuff said.
Apple has tested the waters with XServe... they imply that they are happy with the response to the XServe. Industry and reviews are generally positive toward them... good business says get moving further into this sector of the enterprise market.
And it is highly unlikely that Apple will make way-behind-state-of-the-art servers/blades or whatever devices they chose to make with the PPC line up ahead.
Canberra ? You are not serious not knowing WHAT Canberra is ?
😛
It's unbelievable! It is like someone outside the US saying they think New York is the capital of the USA. The seat of government was created just like Wash DC out of bits of land from two states and thus sits on the edge of Victoria and New South Wales. The city's layout was designed by an American and an artificial lake was named after him: Lake Burley Griffin!
No wonder Wintel was successful in the US!!
Tears! tears! tears! 'nuff said.
Apple has tested the waters with XServe... they imply that they are happy with the response to the XServe. Industry and reviews are generally positive toward them... good business says get moving further into this sector of the enterprise market.
And it is highly unlikely that Apple will make way-behind-state-of-the-art servers/blades or whatever devices they chose to make with the PPC line up ahead.