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.
Xserve Refresh?
- Thread starter MacRumors
- Start date
- Sort by reaction score
Apple does need, at the very least, a redundant option for power supplies. They aren't all that rare in 1U and they are a requirement for a real enterprise server. Since this is the only server Apple has, they need redundant power.
The hardware repair kits are nice but they won't keep you machine from actually dying.
What Apple doesn't need on the server is PCI-E. It may be the shiznit for the kiddes but PCI-X has more than enough bandwidth for the types of cards that get pulgged into a 1U server. PCI-X is the enterprise interconnect anyway. Go look at the multitude of RAID, Myrinet, FC cards that are available for PCI-E.. There aren't any (OK, literally a couple out of the hunderds).
The hardware repair kits are nice but they won't keep you machine from actually dying.
What Apple doesn't need on the server is PCI-E. It may be the shiznit for the kiddes but PCI-X has more than enough bandwidth for the types of cards that get pulgged into a 1U server. PCI-X is the enterprise interconnect anyway. Go look at the multitude of RAID, Myrinet, FC cards that are available for PCI-E.. There aren't any (OK, literally a couple out of the hunderds).
You guys are really missing the point.
Apple's experiment in the server market is really not much more than that.
They don't sell that many of them. I expect with the switch to INTEL they will sell even less.
Apple's experiment in the server market is really not much more than that.
They don't sell that many of them. I expect with the switch to INTEL they will sell even less.
Dude, you are misinformed. Xserves are premium enterprise-class machines. They run nuclear submarines and banks and small offices and major websites and supercomputer compute clusters and anything else. The amount of effort that has gone into their development is phenomenal, and their deployment seems to be expanding rapidly as high-end unix geeks realize that they are in fact very cheap boxes for what they deliver. I can't imagine classifying them as an experiment. That's like saying bluetooth is an experiment...~loserman~ said:
I also think Steve's stated reason's for the chip switch are BS.
He must've know the low-power, dual-core chips were coming soon.
Recently, Intel reported that their production is maxed out -- they can't produce enough chips to meet demand. So Steve's complaint about supply is moot also.
I'm guessing the REAL reason for the switch is to make Macs trojan horses. Create high-quality PC hardware with a free copy of OS X included, and hope people like it better than Windows.
He wants to compete directly with Dell.
I'd like the Xserves to continue to be PPC-based, but from what I've read, IBM is barely making money supplying current chips to Apple -- if they were supplying many less chips Apple would lose the volume discount and it might not make economic sense to use PPC chips at all.
He must've know the low-power, dual-core chips were coming soon.
Recently, Intel reported that their production is maxed out -- they can't produce enough chips to meet demand. So Steve's complaint about supply is moot also.
I'm guessing the REAL reason for the switch is to make Macs trojan horses. Create high-quality PC hardware with a free copy of OS X included, and hope people like it better than Windows.
He wants to compete directly with Dell.
I'd like the Xserves to continue to be PPC-based, but from what I've read, IBM is barely making money supplying current chips to Apple -- if they were supplying many less chips Apple would lose the volume discount and it might not make economic sense to use PPC chips at all.
Better architecture than PowerMac
The Xserve is designed from the ground up to be a SERVER. The ethernet controllers are better integrated into the system and have priority over the USB and FireWire bus. You can check out the architecture differences between the Xserve and PowerMac G5 at Apple's website.
http://www.apple.com/xserve/architecture.html
http://www.apple.com/powermac/architecture.html
Notice how far buried the single ethernet interface on the PowerMac G5 is? Also, the Xserve supports a serial port for UPS connection and management, has three bays for hot plug drives, and has a much better serviceability of the power supply and fans to replace them in case of failure.
Also, the OSX management apps such as Server Monitor are geared at montitoring all the subsystems and components of the Xserve--not the PowerMac. The PowerMac is designed as a personal computer--not a server.
If you needed to run a SMALL network and did not want the Xserve for what ever reason, I could see using a PowerMac G5, however I would really recommend against it. We have had our Xserve G5 for over a year and a half (along with Xserve RAID) and I would not trade them for a PowerMac any day.
macorama said:
The Xserve is designed from the ground up to be a SERVER. The ethernet controllers are better integrated into the system and have priority over the USB and FireWire bus. You can check out the architecture differences between the Xserve and PowerMac G5 at Apple's website.
http://www.apple.com/xserve/architecture.html
http://www.apple.com/powermac/architecture.html
Notice how far buried the single ethernet interface on the PowerMac G5 is? Also, the Xserve supports a serial port for UPS connection and management, has three bays for hot plug drives, and has a much better serviceability of the power supply and fans to replace them in case of failure.
Also, the OSX management apps such as Server Monitor are geared at montitoring all the subsystems and components of the Xserve--not the PowerMac. The PowerMac is designed as a personal computer--not a server.
If you needed to run a SMALL network and did not want the Xserve for what ever reason, I could see using a PowerMac G5, however I would really recommend against it. We have had our Xserve G5 for over a year and a half (along with Xserve RAID) and I would not trade them for a PowerMac any day.
Who's using Itanium besides SGI (in its big servers)? Itanium is really hampered by its low clock rate (1.6 GHz) compared to current x86 offerings, and I don't see Intel putting much if any money into that line to bring it up to x86 manufacturing standards (3+ GHz clock) considering its comparatively low sales. It was not a failure, though, since all the manufacturers who got suckered into Itanium and Intel's empty promises are either dead or dying.Zappa said:Now, Itanium really is a great processor with many interesting features. What Intel needs is a hw-manufacturer with (relatively) little focus on backwards compability. Guess who fits that description... I for one would welcome a Itanium workstation, if it costs less than 10000$ that is
Because a cluster of Xserves costs a tiny fraction of the cost of a mainframe.bigwig said:
According to the designers of the MACH5 cluster of Xserves:
"Our research demonstrated that cluster technology could be purchased and maintained for a tenth the cost of traditional mainframe supercomputers. Superclusters, such as MACH5, will change the way industry competes and operates in the future faster and cheaper."
and
"MACH5 was fully designed and assembled for $5.8 million. NECs Earth Simulator in Japan, the worlds most powerful supercomputer, is a mainframe unit with a $350 million price tag. Considering that MACH5 achieves more than 60% of the Earth Simulators 40 teraflop theoretical performance, but costs less than 2% of the NEC supercomputer, the value of the Apple cluster system is considerable. "
On the Cheap...
While it would be great to have clustering and/or load balancing for every application it rarely happens in every situation. These servers were deployed (and upgraded) over several years. In my experience, most of the time you inherit another persons bad decisions.
As a consultant you recommend best case solutions, but the departmental managers decide how much to spend on a project or application.
Due to cost, only a few apps ran on clustered or load shared hardware. Most of the 'Category 1' applications ran on midrange or mainframe machines running UNIX or MVS.
Not to say that those Windows application owners still didn't scream like little babies when their box went down.
Not every organization gives out blank checks to deploy the best solutions. Most of the time you have to settle for the most bang for the buck. I just happen to want my 'good enough' solutions to have redundant power supplies. I had to change my pager battery enough as it was without adding more headaches.
Toe said:
While it would be great to have clustering and/or load balancing for every application it rarely happens in every situation. These servers were deployed (and upgraded) over several years. In my experience, most of the time you inherit another persons bad decisions.
As a consultant you recommend best case solutions, but the departmental managers decide how much to spend on a project or application.
Due to cost, only a few apps ran on clustered or load shared hardware. Most of the 'Category 1' applications ran on midrange or mainframe machines running UNIX or MVS.
Not to say that those Windows application owners still didn't scream like little babies when their box went down.
Not every organization gives out blank checks to deploy the best solutions. Most of the time you have to settle for the most bang for the buck. I just happen to want my 'good enough' solutions to have redundant power supplies. I had to change my pager battery enough as it was without adding more headaches.
While it would be great to have clustering and/or load balancing for every application it rarely happens in every situation. These servers were deployed (and upgraded) over several years. In my experience, most of the time you inherit another persons bad decisions.
As a consultant you recommend best case solutions, but the departmental managers decide how much to spend on a project or application.
Due to cost, only a few apps ran on clustered or load shared hardware. Most of the 'Category 1' applications ran on midrange or mainframe machines running UNIX or MVS.
Not to say that those Windows application owners still didn't scream like little babies when their box went down.
Not every organization gives out blank checks to deploy the best solutions. Most of the time you have to settle for the most bang for the buck. I just happen to want my 'good enough' solutions to have redundant power supplies. I had to change my pager battery enough as it was without adding more headaches.
After a couple of years, and an XServe update (with hot swap PSs), I thought that I would chime in.
The newest XServe should meet most requirements of a departmental server without question. Open Directory and LDAP along with free CALs make the XServe a viable Windows alternative.
Investigate the benefits and you will see why this makes sense in your environment.