Slashdot Mirror
Build Your Own Blade Server
fw3 writes "Information week is reporting
that IBM and Intel are opening up the standards for the eServer BladeCenter. 'The companies will make available the design specifications for IBM's eServer BladeCenter product... hardware vendors can build "BladeCenter compatible" networking switches, blade adapter cards, and appliance and communications blades for enterprise networks.' Not really a new strategy for IBM, ISA of course was open from the start, IBM's technical references for the original PCs contained nearly all of the engineering data needed to build a PC. Looking further back I've been told by a reputable source that RCA was able to fully duplicate the System 360 System/360, mainframe working just a month behind IBM's own schedule by using IBM's published tech reports. (Of course IBM *didn't* share the details of OS/360, leaving RCA with a box but no OS.) See also stories from EETimes, CNN."
I would sure hate to be cat when the VPs read the heading I am sure something was kicked. Sun needs to read the writing on the wall, newspaper, toilet paper, everywhere - consumers are seeking alternatives from proprietary. Sun's Blade should have been the one in this heading yet they are happy chugging along while companies move forward. Sun is growing Dim.
As for IBM and the RCA scandal, where is the OS/360 today. I wonder if it would have had deeper market penetration if IBM had extended the OS to RCA? Could basically going proprietary with the OS been less successful rather than opening it?
I'm suprised there's no Wesley Snipes tie-in with this product. It would seem natural.
"Blades- the only thing between you, and the end of the world"
Just hope that HP and Sun follow lead and will make things a little easier.
Thus far you could somehow mix'n'match components for standard servers (rack mountable or not), but blades were like hacking a SOHO router...
Wonder how fast will the component manufacturers respond to this and start making parts available (i.e. - we will stop paying exuberant prices for replacement parts from the big guys...)
get a free ipod! This really works... 4 more GMail invites still available for signing up...
Looks like mainframes could be getting cheaper if more companies get their hands on manufacturing them. Looks like Microsoft will have to find a different way to inflate the TCO of running Linux than the current strategy: running Windows 2003 Server on an e-Machine versus Linux on Giant Fucking Mainframe 7000 on the single processor kernel.
It would be cool if it didn't suck.
Well this does seem to open up a new market for clone blade servers, but I'm just not sure who would actually purchase one.
Chances are, if you're going to be spending that kind of money on a server, you're probably going to want something from a reputed vendor, with good support, etc.
> IBM's technical references for the original PCs contained nearly all of the
> engineering data needed to build a PC
Yeah, after plenty of legal action!
IBM's technical references for the original PCs contained nearly all of the engineering data needed to build a PC.
Except for one of the key components to make a PC: the "Build your own BIOS" reference.
The ways of gods are mysteriously indistinguishable from chance.
Try running your OS/360 programs on your brand new zSeries box. Apparently they'll run fine. Of course in the mid-80s (2 decades after OS/360 appeared) IBM pretty much owned the mainframe market and most serious jobs needed a mainframe, so I don't think OS/360 did too badly.
In soviet russia stale jokes recycle you!
This was probably the same model they had in mind for the PC. They wanted to use commodity hardware and even encourage clone makers because they knew that would help allow them to match hardware prices of other high-volume competitors. They figured that they would maintain control of the platform through their proprietary BIOS, and that any clone manufacturers would have to license the BIOS from IBM.
Software vendors would write to the BIOS calls, and IBM would command a position akin to the present-day Microsoft, where they would be the arbiter of the standard interface between application software and hardware. That may explain why they outsourced the DOS OS to Microsoft; they may have thought of it as just a layer over the BIOS. They knew that versions of DOS that ran over other low-level APIs (of which there were a few examples) wouldn't be quite compatible enough to become popular, so they didn't bother to get exclusive control of DOS.
Unfortunately for IBM, the BIOS wasn't that hard to reverse engineer in a clean room environment, clones of the BIOS enabled Microsoft to sell 100% compatible versions of DOS to anyone, and the rest was history.
I guess the lesson to be learned is that if you're going to use software to maintain control over a commodity hardware market, make sure that the software is too crufty and complex to reverse engineer in a reasonable amount of time.
The IBM PC-AT spec opened the door to the commodity "PC" industry. The spec was detailed, and useful, enough for cloners to copy the PC, and the power of competition to drive the vast deployment of cheap PC hardware worldwide. Spawning not only Microsoft and Linux, but the Internet as we (think we) know it today. Especially in light of the obstacles to innovation domino effect we have today, like business process patents, domain name squatting, and every other "legal engineering" trick, IBM's PC-AT spec publication was a work of technology heroism.
But of course, every silver lining has its cloud. For example, the PC-AT spec didn't specify exactly where the motherboard screw holes must appear. So not only were there incompatible motherboard/chassis combinations, but the kluges to accomodate the differences made many cheap boxes significantly more expensive for manufacturers on a volume basis. Just an example of how the 80% solution can spawn its own problems, that require 80% more time to solve. Let's hope we've learned from the last watershed spec publication, and get all the details in the new blade server specs. Especially if we're all going to use them.
--
make install -not war
or do, see for yourself
When it left, it went from Michigan to Georgia, then on boat to Taiwan, where it's probably polluting groundwater to this day.
IIRC RCA wasn't the only company to mimic IBM's systems as I thought that was the business model for Amdahl.
A feeling of having made the same mistake before: Deja Foobar
IBM has historically been a friend of open hardware standards?
If they're trying to make that point.. well, it's just historical revisionism.
Yes, ISA was open. That's why IBM tried to push the MicroChannel bus architecture.
As for mainframes.. IBM invented what we now call FUD to battle Honeywell and Amdahl and the like.
And I'd like to see someone try and build a mainframe clone today. IBM has some seriously secret stuff in those boxes. My father is a mainframe veteran, and he knows some of this stuff. He can't say what, though, because he's under an NDA.
So if you're trying to float the idea that IBM builds hardware to open specifications and always has.. you're just wrong.
Actually OS/360 was "open source". In fact, it was never even copyrighted. I have the entire source on CD-ROM. Anybody can get it at http://www.cbttape.org.
OS/360 had two flavors. MFT and MVT. MVT became SVS when it had virtual storage added on. MVS was in parallel development and once stable replaced SVS. Again, virtual storage. MVS was replaced by MVS/XA when the addressing scheme was changed from 24 bit to 31 bit. MVS/XA was replaced by MVS/ESA along with changes in the I/O architecture. MVS/ESA was replaced by OS/390. OS/390 is in the process of being replaced by z/OS. z/OS is the "flagship" decendant of OS/360. Most programs written for OS/360 will still run on z/OS today. z/OS also has an integrated "UNIX" personality so that it is possible (but not easy) to port UNIX code to z/OS. I have done this with GNU make, gzip, and bzip2 myself. And I'm not an expert in C either! z/OS is fully 64 bit capable on the latest eServer zSeries mainframes. That's 64 bits of addressing and data in registers. At the same time it is fully backward with the older 31 and 24 bit address and 31 bit data registers.
Unfortunately, unlike the original OS/360, z/OS is almost 100% "closed" source. It is even written in a proprietary language with IBM does not license to the general public.
Troll? I was mis-informed. I knew that Compaq had to reverse-engineer the BIOS to make their PC's IBM compatible; I was not aware that they needed to do this despite the source code being available.
Mod parent overrated I agree. But troll?
The ways of gods are mysteriously indistinguishable from chance.
IBM's BIOS source was copyrighted. If Compaq had simply compiled the source code listing, they would have faced a lawsuit (and a unfriendly precedent in the form of Apple v. Franklin).
this is my blade server.
free online diet tracking.
The truth is Sun just does whatever the hell it feels like. Their engineers have two bosses, themselves, and whatever the current customer base asks for -- the market and analysts be damned. It's like here's everybody doing their thing, chasing after each other, and over here, is Sun. It's in its own little world. Oracle is a lot like that too.
And that can be both good, and bad.
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
Back in the mainframe days - which is when I got my start in the biz - OS/360 and OS/370 (up to OS/MVS V7) were open source. The source was distributed on microfiche, and system programmers were encouraged to modify the code to make the whole thing run better. There was a user organization called the Society to Help Avoid Redundant Effort (SHARE) at which system programmers shared their code modifications with each other, and with the IBM developers. Some of the good stuff made its way back into the standard "distro" - although we didn't call it that back then.
Similarly, the hardware diagrams were standard manuals that existed in every datacentre. I remember browsing through them shortly after I finished school (a hundred years ago or so) and thinking, "there really isn't much to these mainframe computers; nothing much more than the final exam in electronics." But based on those diagrams, and other info, our datacentre was the first in the world to put the 9th megabyte on an S 370/168!
And yes, at the time, I did get questioned about how on earth we could have so much work that we needed a 9th megabyte on a 168.