IBM and Sun Microsystems are in the final stages of an 18-month-long project to adapt the OpenSolaris operating system to run optimally on Big Blue’s z System mainframes. David Boyes, president and chief technology officer of Sine Nomine Associates, a consultancy that handled most of the integration and migration duties, told eWEEK at the Gartner Data Center Conference 2007 that the new IBM-tuned version of OpenSolaris will be ready soon. He declined to be more specific.
At the end of the eWeek article is a comment about a TechTV video and a YouTube link. The video is broken up into 5 parts. To get the full picture as to what is being done to run OpenSolaris and Linux (both are discussed) on a z/Series machine, you have to watch all of the videos.
I know the folks in the AIX organization aren’t going to be too happy about this. It’s one thing for Solaris to run on System X. Sun already sells Solaris on x86 servers, so it’s in IBM’s best interest to win some hardware contracts for those who want Solaris.
But prior to this move, Sun didn’t have a mainframe offering. The only game in town is System Z, IBM’s crown jewel. It already offers two *nix options: RHEL and SLES. Why give Sun the mainframe solution they didn’t already have? If customers wanted a rugged big-iron UNIX on Z, then why not put your own flagship UNIX on your flagship platform?
It seems like IBM’s Systems & Technology Group is still feeling burned by Project Monterrey, the port of AIX to Itanium that never saw the light of day and ended up becoming a central pillar of the SCO lawsuit. There doesn’t seem to be any political will to expand AIX beyond System P. As Z becomes ever more suited to consolidating UNIX workloads and P attempts to become more like a pseudo mainframe, I think that both P and AIX are living on borrowed time.
I see smaller sub-enterprise Z models capitalizing on the resurgent popularity of workload consolidation and the newfound challenges of power and cooling in the datacenter to dominate the high-end. I see larger x86 NUMA boxes and clusters with multilevel caches, per-socket memory controllers, and serial point-to-point bus architectures (e.g. Barcelona and Nehalem) becoming the bread-and-butter platform for midrange IT and HPC.
I see AIX and HP-UX slowly dying out, Solaris being the only big-iron UNIX to successfully navigate the commoditization of IT. It really is a shame. AIX is an outstanding UNIX that continues to lead the market in innovative RAS and performance features. It’s trapped on a similarly outstanding SMP platform that’s suffering from an identity crisis and struggling to relate to a changing market.
Woe is AIX.
I’m not so sure. These mainframe efforts to bring in OSes are about consolidation. What they want to do with this is to take Sun’s hardware sales.
And they would probably prefer AIX on pSeries not to be consolidated onto mainframes.
IBM would never turn down a chance to move a customer from X/I/P up to Z. System Z is their highest margin platform, and once customers go Z, they’re almost certain to remain loyal customers for the long haul.
There has always been some degree of cannibalism at IBM between sales teams for the various platform, less today than there has been in the past, but once a customer implies any interest in Z, that becomes the pitch.
I don’t believe IBM would want to move a customer from pSeries to zSeries unless it was also consolidating solutions from other vendors onto that hardware as well.
If you think otherwise, then why don’t they make it easier?
“[…] they would probably prefer AIX on pSeries not to be consolidated onto mainframes.”
The next generation of mainframes are going to be based on POWER6, with emulation (similar to the AS/400 – RS/6000 merge). This will give them one hardware platform to support, with AIX for free.
“The next generation of mainframes are going to be based on POWER6, with emulation (similar to the AS/400 – RS/6000 merge). This will give them one hardware platform to support, with AIX for free.”
This is completely wrong. The eclipz project yielded 2 processors.
No, eclipz unites I, P, and Z. Hence the last three letters of the name. The POWER6-based processor for Z will be the last to come to market, but it will come.
“No, eclipz unites I, P, and Z. Hence the last three letters of the name. The POWER6-based processor for Z will be the last to come to market, but it will come.”
Yeah, the one of the goals of eclipz was to unite, but it wasn’t very successful with the processors. The poster that I replied was hinting that AIX would soon run on mainframe hardware because the processors are the same. This rumor is not true, and was started by some tech journalist who wrote a story claiming that we were going to emulate the z architecture on power. That idea was entertained at one point, but the performance was terrible, and it was dropped.
Here are some differences:
– ISA
– cache hierarchy
– cache coherency
– page tables/supported page sizes
– pipeline
– packaging/pin layout/cores per chip
Here are some things that were common before:
– elastic interface
– gx bus
Here are some of the newly common things:
– nova memory controllers
– some high frequency circuits
– SRAM
– in order instruction processing
– binary and decimal floating point
Here are some things that p6 got from the mainframe world:
– r unit
– storage keys
None of the commonalities are going help anyone boot AIX on the mainframe. I think the z6 has much more in common with the bluefire z9 processor than p6.
I think porting Solaris to z/Series says a lot about (1) the demand for Solaris and (2) IBM’s inability to generate any excitement about their own products. While IBM carried the torch for Linux, it let AIX stagnate to the point where very few people saw any benefit in using it. HP also shot themselves in the foot in much the same manner. When I attend a meeting with HP Enterprise Sales and they talk more about Linux than HP-UX, that says a lot. Especially when they say that HP is more willing to put money into Linux than HP-UX, that tells me if I was an HP Enterprise customer to look elsewhere. And many people did exactly that, much to HP’s dismay.
Storage keys, live dump, concurrent update (kernel hot patch), checkpoint restart, runtime error checking, functional recovery, dynamic partitioning, micro partitioning, live migration, virtual I/O, component trace, lightweight memory trace, dynamic trace (ProbeVue), virtual real memory…
Surely we’re talking about a different AIX?
Innovation is not the problem. The problem is marketing and pricing.
Well, they let it stagnate from a marketing perspective, then. Instead they gave all of their marketing attention to Linux.
But I doubt that was a complete accident. Perhaps IBM is ready to give up the burden of having to maintain AIX indefinitely. They’ve enthusiastically supported and contributed Linux in the past; maybe in the near future they’ll jump on the Solaris bandwagon full-steam and start contributing there, too. Perhaps Solaris on System z is a sign that IBM is ready for a consolidation of the UNIX market behind one open-source front-runner. Were IBM to will it, Solaris on Power could shape up very quickly.
It would of course be a shame that AIX’s innovations would be left behind, but they wouldn’t have to be–IBM could always add them to Solaris, after all. Of course right now they would never admit that they’d give up AIX for Solaris, but they have after all embraced Linux wholeheartedly and surely Solaris provides most of the same benefits (cheap, maintenance burden is lifted from IBM) while adding an extra dose of “enterprise-readiness”. It makes even more sense for IBM to embrace Solaris in the future when you consider that they already partner with Sun on a variety of things (Java, OpenOffice), whereas I doubt IBM is very happy about Novell’s deal with arch-enemy Microsoft.
I’d say “partner” is a bit optimistic with regards to Java and OpenOffice; IBM has their own idea of Java’s direction, and their OpenOffice support is somewhat self-serving.
As for Microsoft, nobody sells more MS software than IBM; few companies use more MS software than IBM. Same holds true for Novell, IBM is their largest reseller and has made a financial investment. If anything, improved interoperability between MS and Novell benefits IBM more than any of the other players in the industry, and the community strife about patents versus GPL licensing wasn’t even a ripple in the water at the enterprise level that IBM serves.
I suspect this is purely an example of co-opetition. If IBM is going to run Solaris on their hardware, they are fully expecting to generate significant support revenue from doing so. Support and contract services are ultimately the motivator behind everything IBM does.
Sun will just have to gamble that IBM won’t be able to completely cut them out, since Sun is as heavily dependent on their service revenue as IBM is.
I’d be very surprised to see IBM making any code or development contributions to Solaris, though, unless Sun was willing to license it. IBM is very protective about their IP, and contributing code to the linux kernel while still retaining ownership of that code is much different from having to assign copyright to a competitor in order to have code accepted.
Will be interesting, regardless, to see how this all shakes out.
And the majority of this can be done with Solarisas well.
While I haven’t used AIX since 5L 5.2, pricing for both hardware and software made the decision for many to go to either Linux, Windows or Solaris much easier. And IBM’s marketing, well I haven’t talked to anyone from IBM in almost 4 years, since I left the last project that was virtually “owned” by IBM.
And I think where I work at would be real high on some marketing exec’s radar, it is for Sun, Dell and HP.
“Storage keys, live dump, concurrent update (kernel hot patch), checkpoint restart, runtime error checking, functional recovery, dynamic partitioning, micro partitioning, live migration, virtual I/O, component trace, lightweight memory trace, dynamic trace (ProbeVue), virtual real memory…
Surely we’re talking about a different AIX?
Innovation is not the problem. The problem is marketing and pricing.”
How about a little modesty? Are these really innovative? Sure it took a lot of hard work to implement them in AIX, but almost all of them were originally implemented by others.
I think porting Solaris to z/Series says a lot about (1) the demand for Solaris and (2) IBM’s inability to generate any excitement about their own products.
Not really. It’s more related to point 2:
1. Sun is running around trying to get people to run Solaris. As long as Sun are willing to put in some time and money, most are OK with it.
2. Solaris on Sun’s own hardware is no longer enough for them if Solaris is to remain relevant to people.
3. IBM don’t give a toss what is running on any system such as z or p, as long as it generates margins as wide as the Grand Canyon.
From the article “Boyes said he knew of at least “30 to 40 customers that are lined up and waiting for this.” I somehow don’t think people would go through the effort (which by their own admission is not trivial) if there was no demand for it. This isn’t a couple of geeks trying to get OpenSolaris to run on an XBox, the cool points wouldn’t be worth the effort.
And why wouldn’t Sun want Solaris/OpenSolaris to run on a z/Series machine? Choice is good, right?
IBM’s margins is the stuff of legend, and let’s not even get into IBM Global Services.
Sun has done a really good job modernising Solaris to bring its performance up to par to Linux for small systems and world leading scalability on very large systems on both x86 and SPARC.
It only seems natural to extend the reach of Solaris to Z and P series, with Itanium platforms certain to become victims too eventually. What will become of AIX and HP-UX when Solaris runs on their native architectures too?
I’ve ported the most recent releases of my Linux distribution to x86_64, MIPS and SPARC and Solaris seems to be the most prominent platform that I’ll have to contend with since both aforementioned OSes are seemingly going away.
I’d love to support Z series, but as a small distribution and software developer the requisite hardware seems completely out of reach for me. Any ideas how to get access to one?
IIRC you can contact IBM and I think that you can either get access to or rent out time on a mainframe – a university might provide access, I am unsure. I do know that at one point they were running a programme to encourage more developers and IT people to consider a career in mainframes.
Regarding Solaris on the mainframe, this is great news, it’ll be interesting whether HP eventually sucks in their pride and works with Sun to bring Solaris to Itanium – it shouldn’t be too hard given that it was in ‘early access’ a couple of years ago, and IIRC the compilers were partially done too.
I think there is also a move to bring Solaris to POWER as well, which should make things interesting – the question now is whether Sun has the skills to be able to support those new platforms by way of services and consulting or risk losing those profits to IBM.
As a Linux user and advocate, I am pleased to see another Open OS running on the Z. As much as I like Linux, it is not the be all and end all of Open operating systems. Multiple Open OSes running on IBM’s mainframes is a sign of health for Open Source, and for competition between competing ideas in that marketplace, as well.
This is interesting, but I have to admit that I have not yet been able to convince myself that the mainframe for unix consolidation is more than an IBM attempt to move more boxes. Does anyone here have any first hand experience?
It just seems like with the cost of the machine (100k entry level with one cpu, last I saw), there is no way you could run enough VMs to make the cost numbers work, especially considering you probably have the cpu performance of an old POWER4 or a single core opteron. Some may claim that it’s all about IO, but you still need some cpu here and there.
Now I know some people will claim RAS is that much better, but first of all, Linux especially is developed with a way different mindset than, say, z/OS, negating some of the traditional reliability, and secondly, for the price, you could have a multi-node VMware cluster between buildings and still have money left over.
Anyone know what I don’t know? And by know, I mean, have actually done it.
Well, Solaris is a little late. zLinux runs, and runs, and runs there already – for 8 years now
SCNR
Well, Solaris is a little late. zLinux runs, and runs, and runs there already – for 8 years now
SCNR
In mainframe terms, what’s 8 years? Linux was a little late, seeing as OS/360 already ran there for, oh, 40 years or so.
Someone mentioned getting access to a zSeries for development purposes, you can run Hercules and emulate one, but it won’t be especially fast.
IBM and HP are slowly moving away from their proprietary unixes, they were moving towards Linux but now Opensolaris satisfies the same requirements too…
With their own proprietary unix they have to pay for all of the development themselves, and worry about compatibility with other systems their customers might be using in parallel or migrating from. With an open OS the development effort is shared, and compatibility with other vendors’ kit becomes much easier, especially if the other vendor is running the same OS.
IBM and HP are hardware and support companies, software doesn’t make them any money it’s merely necessary to facilitate sales of their hardware and consultancy services. They could just as easily sell someone else’s proprietary OS, and do none of the development themselves, but then they would always be a second tier support provider which isn’t an ideal situation.
In the not so distant future i can see AIX and HPUX being killed off, with IBM/HP respectively merging their unique features into Linux or Solaris.
And talking of dead proprietary unix, Tru64 was arguably a lot better than HPUX, but that’s gone now… I wonder if these vendors could be convinced to open up the source to their unixes (especially unused ones like Tru64) so that things can be ported across.
this reminds me of Netscape’s business model in 1999. ‘We will port our browser to all platforms, regardless of whether users will need a browser on that platform and whether we will ever provider a 2.0 version.’
How does Sun make any money on like 10 installations of Solaris on a Mainframe? There’s precious little economy of scale here.
I just don’t see the financial benefit to Sun here. Where is the bottom line benefit??
this is about marketing, not money.
In the old Enterprise environment I support 3 Z Series running Z/os and I am glad I don’t have to anymore.
In all honesty the cost of a IBM Z-series you can purchase a complete Dell server farm and buy Gold support. IBM has some good products but the support is very expensive (hardware/software) and the other vendors have really caught up and by passed big blue.
I remember they installed a Linux distro I think maybe SuSE on an L-Par and they tested it for a little while but that is as far as it went. The new ‘smaller’ company I am with I support ONLY Red Hat Enterprise Linux, JBOSS, VMWare, some Oracle and NetApp Filers. I do not miss the IBM mainframes at all, nor do I miss the Z/os I would rather really have teeth pulled than go back into the mainframe world.
Sun has a solid OS and the I/O it can achieve on their hardware is very impressive but other vendors are hungry the competition is greater than what it was in the past, also new technologies are on the scene.
Virtualization right now is where it is at and until something else comes along the big expensive hardware platforms will die off a little at a time.
However, the mainframe is NOT going away in some shops with CICS, batchprocessing, and uptime that is really hard to beat it will be very difficult for a place to replace the big iron box…
http://realworldtech.com/forums/index.cfm?action=detail&id=83745&th…
The Z6 mainframe CPU shares some come features (cache etc.) but is not a version of the Power6.
Here’s the link to the IBM pdf on the matter:
http://www2.hursley.ibm.com/decimal/IBM-z6-mainframe-microprocessor…