If I'm a bit crazy I would attribute this fw port to the return of the fire-wire iPod.
Yes, but only because you'd even consider the possibility that the choice of a specific I/O port on the front of a server product might have anything whatsoever to do with a consumer media device.
Yes, but only because you'd even consider the possibility that the choice of a specific I/O port on the front of a server product might have anything whatsoever to do with a consumer media device.
The iPod is actually mentioned several times in the documentation regarding automatic setup of servers.
Xserve doesn't care of it's an iPod, FW disk, USB disk, USB keychain or something else though.
I'm a bit miffed that they've chosen software RAID. Yuck.
I'm also a bit miffed that the dual NIC don't seem to support TCP/IP acceleration.
Exactly correct. The Xserve update seems to be missing a few important features. Apple should have integrated hardware RAID support (with OS configutation support) and OS support for "teaming" or combining NICs according to the Cisco/Sun standards. Competing Sun boxes offer multiple NIC teaming, and heck, so does SUSE Linux running on datacenter style HP rackmount servers. Apple is behind the curve in this area.
And what's with the three hard disks? That's just bizarre; two or four make sense because they can be mirrored for the kind of reliability a server generally needs.
Apple should have designed this generation around the Small Form Factor drives. Seagate will be shipping the Savvio 10k 146GB drive mid to late september. Moving to SFF drives would have allowed Apple to plug in a 5 drive system which allows for RAID 5 and a Hot Spare. If people need more than 5 bays or large drives you have Xserve RAID
I agree you need multiple nics now with teaming functionality and acceleration. Accelerating TCP/IP unburdens your server by %10 or more Dell, HP and probably IBM will all be standardizing on these TOE NIC because Broadcom has a chip in wide use now that does TCP/IP, iSCSI and RDMA (fast transfer from server to server with little to no CPU involvement) and more. Perhaps it's in there and I'm just not seeing the right spec sheet.
However I must state they've done more right than wrong here. Better management tools, RPS, onboard video and fast procs will definitely get Xserves into racks.
Man I'd love to see what Apple could design regarding a Blade System. Since Apple is basing so much stuff around Grid computing and high end DCC I dream of the Blade System aimed at content creators and video production who need power on demand.
Ok so nobody has pointed out the fact it supports 32GB OF RAM! Yes thats right read the specs. All I want to know if it's the same setup essentially as the Mac Pro why the Mac Pro doesn't support up to 32GB of RAM.
I think this is a wonderful update and can't wait to get 1 or 2. I already have an XServe G4, XServe G5, & XServe RAID.....
What's the front firewire port for? There's no more fire-wire iPod to be used as target disk, Serial, or USB would be a much better choice now.
Granted, there's alot of 3rd party FW400 drives, FW800 or USB2 would be much better.
It probably would have been better if they had more ports, but FW400 is generally faster for real-world data transfer than USB 2 is. USB would be a useful feature, though I think mainly for the input devices.
Ok so nobody has pointed out the fact it supports 32GB OF RAM! Yes thats right read the specs. All I want to know if it's the same setup essentially as the Mac Pro why the Mac Pro doesn't support up to 32GB of RAM.
I think this is a wonderful update and can't wait to get 1 or 2. I already have an XServe G4, XServe G5, & XServe RAID.....
I'm pretty sure the Mac Pro will just as easily support 32GB but Apple doesn't seem to be pushing 4GB FB-DIMMs just yet.
I agree you need multiple nics now with teaming functionality and acceleration. Accelerating TCP/IP unburdens your server by %10 or more Dell, HP and probably IBM will all be standardizing on these TOE NIC because Broadcom has a chip in wide use now that does TCP/IP, iSCSI and RDMA (fast transfer from server to server with little to no CPU involvement) and more. Perhaps it's in there and I'm just not seeing the right spec sheet.
The Intel NICs support I/OAT, which accelerates TCP somewhat; it will remain to be seen whether Apple supports it in OS X. TOEs have shown some performance improvements (and some degradation), but only in Windows. I've yet to see any TOE speedups under any flavor of Unix.
Intel's going to have to market I/O Acceleration Technology a lot better. Broadcom's cleaning their clock with their chipset. I've viewed the I/OAT flash file a few times and I still cannot understand what I/OAT accomplishes that isn't already handled well enough by TOEs. They seem to have re-invented the wheel.
Funny you should mention the Unix TOE benches. Can't give specifics because of NDA but I'm looking at a powerpoint that's showing a 10G card improving %60 in Windows and over %200 in Linux. Obviously Windows 10G performance is going to have to improve muy rapido.
I thought many companies would be using Intel's NIC but Broadcom is cleaning house. Must be a feature and price thing. Here's hoping that I/OAT is supported in OS X Server.
I must say though I don't know how Apple gets the price of their hardware down. I almost feel wrong for complaining because it's pretty hard to build a solid 1U server with NOS and RPS for under $3k from a Tier 1 vendor. Plus the Xserves look sexy. I'll have to own one someday for my own needs.
Intel's going to have to market I/O Acceleration Technology a lot better. Broadcom's cleaning their clock with their chipset. I've viewed the I/OAT flash file a few times and I still cannot understand what I/OAT accomplishes that isn't already handled well enough by TOEs. They seem to have re-invented the wheel.
Intel is pitching it the opposite way: I/OAT provides the same benefit as TOE but it is simpler and costs less. Of course, we will have to wait for independent benchmarks to tell if that is true.
Comments
Hot Swap PS and hopefully fans will be welcome as well.
www.apple.com/xserve
Integrated video is nice.
Granted, there's alot of 3rd party FW400 drives, FW800 or USB2 would be much better.
If I'm a bit crazy I would attribute this fw port to the return of the fire-wire iPod.
But I'll conclude that it's a remnants from the original XServe
If I'm a bit crazy I would attribute this fw port to the return of the fire-wire iPod.
Yes, but only because you'd even consider the possibility that the choice of a specific I/O port on the front of a server product might have anything whatsoever to do with a consumer media device.
I'm a bit miffed that they've chosen software RAID. Yuck.
I'm also a bit miffed that the dual NIC don't seem to support TCP/IP acceleration.
They can rectify this by working diligently with Parallels and VMware on server virtualization.
I'm assuming that now that we have SAS support a new Xserve RAID is due with 4Gb Fibre and 14 SAS/SATA drive bays.
Next we'll need XBlade
Up to 8 servers in a high density enclosures running Woodcrest Blades, Gigabit and Fibre interconnects with Xsan and Xgrid support.
Mac Pro front end. Xserve or Xblade render/HPC clusters and Xserve RAID.
Also they might want to start converging the management of Xserve RAID and Xserve into one console app with SMI-S support.
Yes, but only because you'd even consider the possibility that the choice of a specific I/O port on the front of a server product might have anything whatsoever to do with a consumer media device.
The iPod is actually mentioned several times in the documentation regarding automatic setup of servers.
Xserve doesn't care of it's an iPod, FW disk, USB disk, USB keychain or something else though.
Now for what isn't so hot.
I'm a bit miffed that they've chosen software RAID. Yuck.
I'm also a bit miffed that the dual NIC don't seem to support TCP/IP acceleration.
Exactly correct. The Xserve update seems to be missing a few important features. Apple should have integrated hardware RAID support (with OS configutation support) and OS support for "teaming" or combining NICs according to the Cisco/Sun standards. Competing Sun boxes offer multiple NIC teaming, and heck, so does SUSE Linux running on datacenter style HP rackmount servers. Apple is behind the curve in this area.
And what's with the three hard disks? That's just bizarre; two or four make sense because they can be mirrored for the kind of reliability a server generally needs.
I agree you need multiple nics now with teaming functionality and acceleration. Accelerating TCP/IP unburdens your server by %10 or more Dell, HP and probably IBM will all be standardizing on these TOE NIC because Broadcom has a chip in wide use now that does TCP/IP, iSCSI and RDMA (fast transfer from server to server with little to no CPU involvement) and more. Perhaps it's in there and I'm just not seeing the right spec sheet.
However I must state they've done more right than wrong here. Better management tools, RPS, onboard video and fast procs will definitely get Xserves into racks.
Man I'd love to see what Apple could design regarding a Blade System. Since Apple is basing so much stuff around Grid computing and high end DCC I dream of the Blade System aimed at content creators and video production who need power on demand.
I think this is a wonderful update and can't wait to get 1 or 2. I already have an XServe G4, XServe G5, & XServe RAID.....
What's the front firewire port for? There's no more fire-wire iPod to be used as target disk, Serial, or USB would be a much better choice now.
Granted, there's alot of 3rd party FW400 drives, FW800 or USB2 would be much better.
It probably would have been better if they had more ports, but FW400 is generally faster for real-world data transfer than USB 2 is. USB would be a useful feature, though I think mainly for the input devices.
Ok so nobody has pointed out the fact it supports 32GB OF RAM! Yes thats right read the specs. All I want to know if it's the same setup essentially as the Mac Pro why the Mac Pro doesn't support up to 32GB of RAM.
I think this is a wonderful update and can't wait to get 1 or 2. I already have an XServe G4, XServe G5, & XServe RAID.....
I'm pretty sure the Mac Pro will just as easily support 32GB but Apple doesn't seem to be pushing 4GB FB-DIMMs just yet.
I agree you need multiple nics now with teaming functionality and acceleration. Accelerating TCP/IP unburdens your server by %10 or more Dell, HP and probably IBM will all be standardizing on these TOE NIC because Broadcom has a chip in wide use now that does TCP/IP, iSCSI and RDMA (fast transfer from server to server with little to no CPU involvement) and more. Perhaps it's in there and I'm just not seeing the right spec sheet.
The Intel NICs support I/OAT, which accelerates TCP somewhat; it will remain to be seen whether Apple supports it in OS X. TOEs have shown some performance improvements (and some degradation), but only in Windows. I've yet to see any TOE speedups under any flavor of Unix.
Funny you should mention the Unix TOE benches. Can't give specifics because of NDA but I'm looking at a powerpoint that's showing a 10G card improving %60 in Windows and over %200 in Linux. Obviously Windows 10G performance is going to have to improve muy rapido.
I thought many companies would be using Intel's NIC but Broadcom is cleaning house. Must be a feature and price thing. Here's hoping that I/OAT is supported in OS X Server.
I must say though I don't know how Apple gets the price of their hardware down. I almost feel wrong for complaining because it's pretty hard to build a solid 1U server with NOS and RPS for under $3k from a Tier 1 vendor. Plus the Xserves look sexy. I'll have to own one someday for my own needs.
Intel's going to have to market I/O Acceleration Technology a lot better. Broadcom's cleaning their clock with their chipset. I've viewed the I/OAT flash file a few times and I still cannot understand what I/OAT accomplishes that isn't already handled well enough by TOEs. They seem to have re-invented the wheel.
Intel is pitching it the opposite way: I/OAT provides the same benefit as TOE but it is simpler and costs less. Of course, we will have to wait for independent benchmarks to tell if that is true.