Slashdot Mirror
Linux-only POWER5 server From IBM
vaporland writes "This story from Infoworld.com talks about IBM's new low priced POWER5 based servers which will ship with Red Hat or Suse Linux, but not IBM's AIX.
My question is, will it boot up Apple's OSX Server?"
Highly doubtful, OSX requires special roms... Maybe in Mac-On-Linux though.
IBM wants your apps. They'll give you a free laptop bag if you develop an app for Linux on their Power platform. I think the most disturbing thing about this is the picture of Tux in a blue tux. He looks like he's ready to sing in a lounge somewhere.
But (somewhat) seriously, I could make room a four-way 64 bit Linux box on my desk. Chances are you've probably not seen my desk -- not an easy task.
Recipes for geeks -- no meatloaf, we promise.
My question is, will it boot up Apple's OSX Server?"
No, because OSX does checks to look for Apple-specific hardware on boot.
I'm out of my mind right now, but feel free to leave a message.....
lack of device drivers?
Possibly the cost of an AIX license? (though thats not a silicon limitation)
In the future, I would want to not be isolated from my friends in the Space Station.
Well, linux on PPC has certainly been improved a whole lot.
GCC for PPC used to produce ungodly slow code. However Apple and IBM have been doing alot of optimisation work on gcc in the last few years, so now it's almost as good as the x68 target.
Because the current RedHat offering, RHEL, is a BUSINESS application, and in business the mantra "If it ain't broke, don't fix it" applies VERY strongly.
If you want 2.6, just use Fedora instead.
This is much more likely than it running OS X. And probably a lot more useful too.
/. anyhow? I guess maybe they want to encourage discussion, but in this case, they've encouraged discussion that is off topic. IBM is releasing a POWER 5 Linux machine - lets talk about IBM, Power, and Linux, not Apple.
Why is it you have to ask a really dumb question to get a submission posted to
Well, their XLF is allready out for linux, I expect that XLC/C++ will get a linux release too sooner or later.
Here is the answer...
http://www.redhat.com/software/rhel/kernel26/
because most of the features important to Redhat has already been backported to their "2.4"
The thing is that Redhat isn't targetting the same market, as say Fedora or Suse.
That's why they created Fedora, otherwise all you guys would be bitching about redhat using 2.4 on your desktops.
They have their OS certified by many 3rd party makers of propriatory software, such as Oracle. Stuff like that takes a lot more work then just slapping 2.6 into a OS and making sure that everything works.
Everything has to work the way Oracle (for example) expects it to work, and Oracle is only going to tell Redhat the minimal it is needed to get it running, since it's closed source.
Don't worry, Redhat isn't dropping behind, it's just that they've adopted a much more long term-style revision policy compared to other distros. When they switch to 2.6 it will be a relatively cutting edge version of 2.6.
Which should be pretty soon. Also PPC developement is going to be falling a bit behind x86 developement due to the relative popularity of the different platforms.
Are you suggesting the fortune 500 companies spending thousands of dollars on RHEL 3* deployments should have either expected a kernel less then 6 months old in the intial RHEL 3 release? Or perhaps that their shiny new product certified to keep component version numbers stable for 5 years suddenly do a major version kernel upgrade 6 months into the product life cycle?
* RHEL 3 is the only Red Hat distribution you could be talking about. All other RHEL versions came out before kernel 2.6, and FC2 and FC3 do use kernel 2.6. FC1 won't get it because FC2 is halfway through it's own life cycle (which means FC1 is old & busted).
for quite a while (year +) on the iSeries (old AS/400) and now the i5 (rebranded, again AS/400) boxes
n ux /dist.html
http://www-1.ibm.com/servers/eserver/iseries/li
They started serious power development in Redhat 7.1 . Right before RHEL came out. It was always 64-bit though. Intented for pSeries and not Macs. That is why it was not well known.
Flexible bare-metal recovery for Linux/UNIX
Good thing the kernel's open source, then. There are things like XPostFacto which make OS X work on older Macs by providing the necessary drivers/etc and faking things as necessary, for instance.
http://freshmeat.net/projects/mol/
http://www.maconlinux.org
Sorry for the AC post
PowerPC RISC architecture was developed by IBM in the late 80s / early 90s. Apple, IBM and Motorola formed a triumverate to further develop the architecture. IBM came into the deal with the 601 series processor already in development. When Somerset Park was shut down, Motorola inherited most of the physical assets, while all three companies retained ownership over the intellectual properties. The last processor series developed in concert between the three companies was the PowerPC G3.
Motorola and IBM diverged on development after the G3. Motorola was very interested in further developing vector processing, while IBM focused on better enhancing the core of the processor. Hence, then next iteration of the series from Motorola was known as the G4, while IBM went on to develop the Power 4, using similar underlying core architectures, though Motorola was stymied by clocking problems that didn't seem to be as much of an issue for the Power 4 processors.
The G5 processor that Apple is using now is actually a variant of the Power series from IBM that has included vector units. The Power 5 is also developed from the same core. Call them cousins if you will. So it would not be beyond the scope for someone to find a way to run Mac OS X Server to run on a Power 5 architecture. In fact, about the only thing stopping that development, I bet, is the threat of litigation from Apple.
This was mentioned in an article in the Wall Street Journal today. The article is regarding vendor-backing of LSB2. Near the end, the WSJ stated this product is meant to compete with Sun and HP workstations. Link to related story, as WSJ's requires subscription services.
Linux: The world's best text-adventure game.
PowerPC IS a subset of POWER, but with Altivec added. Altivec is apparently pretty much motorola's game, but because they got in bed with IBM, IBM got it too. (Partial custody?) The latest PowerPC processor is not either-endian, support for the either-endianness of many instructions was dropped in the G5. So it's sometimes-either-endian. This change has caused problems for people who write emulation software of the x86. PPC601 was the only PowerPC processor which had a full POWER instruction set, after that they've all had less instructions, with the obvious exception of the G3 and up having altivec/SIMD.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
ECC/Registered memory, PCI-X, SCSI, RAID (I mean $600+ PCI-X cards here, not the crappy little software RAID you get with your on-board ATA controller), rackmountable cases with redundant PSU's and insanely loud 15kRPM Delta fans, SMP/NUMA, lots of IO-APIC's (= lots of IRQ's), serial console even in BIOS and POST, and generally higher quality components, with on-site warranties with response time guarantees and things like that.
That answer your question?
I can name you one...
On a p690, you can run 32 processor jobs on AIX. Linux on a p690 won't get you farther than 26 processor jobs before the box crumbles.