Linux: it's the last word on AIX vs. Solaris

Linux: it's the last word on AIX vs. Solaris

Summary: With Linux on Power you get open source, you get a hot chipset, you get that IBM relationship, you get a clear future direction, you get a solid development community, you get access to lots of applications,you get a free any time escape to the cheaper Lintel world.

TOPICS: Linux, Open Source

If you look at the whole set of issues facing a data center manager researching Solaris on SPARC versus AIX on Power in response to some senior executive with an IBM agenda, the one thing that stands out is simply that most of the problems, including pricing, pertain to AIX and mainframe "headset" issues, not the Power architecture or IBM.

AIX 5L 5.3 is far behind - and the imitative Solaris 10 "features" in the current beta for 6L look more like complexity increasing paste ons than the kind of simplifying fundamental change characterizing Solaris 10. The Power6, on the other hand, is a pretty hot chip in its own right, a first step in a putative transition to Cell, and a credible test-by-customer for the higher gigahertz manufacturing processes needed for Cell.

Consider, however, Linux on Power: it isn't remotely in the "insanely great" category, but it works - and if Linux could eventually be tailored to Power the way Solaris is tailored to SPARC, the combination could work with IBM's business credibility to give Sun a real run for its customers.

Right now getting Linux code to run on Linux for PPC isn't very difficult - it's very inefficient, because of the endian change and the inappropriateness of the x86 specific ideas embedded in just about everything Linux, but it's relatively easy to do and it works. AIX code moves across to Linux fairly easily too - in part, I think, because the generational gap is much smaller than with Solaris: it's no coincidence that the Solaris to AIX porting materials on IBM's porting portal generally pertain Solaris 8 and earlier while the Linux to AIX stuff is reasonably current.

As a result there's an obvious recommendation for people either using AIX now or considering the value of a business relationship with IBM: think Linux on Power - and forget about AIX.

With Linux on Power you get open source, you get a hot chipset, you get that IBM relationship, you get a clear future direction, you get a solid development community, you get access to lots of applications, you get a free any time escape to the cheaper Lintel world - and you get Sun covering your bets because what runs on Linux, can usually run on Solaris too.

All, as the mole sisters would say, great stuff - and really, not so LinP at all.

Topics: Linux, Open Source

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • Linux on POWER too expensive

    Unless POWER hardware gets drastically cheap I don't see this
    scenario of yours working. I don't like Intel architecture but
    the future is Linux on cheap Intel and that is why Sun is changing
    their hardware line up. POWER technology is in all game consoles
    and is used more widely than SPARC. Until Sun/IBM make these chips
    a lot cheaper than they currently are than all UNIX'es are under

    Linux is not the best in all areas but it is good enough.
    Intel is not a great architecture but it is good enough.
    To run your Oracle DB on that combo is a lot less expensive
    and that is what really matters.

    In a world where technologies survived based on merit IA64 would
    have died and Alpha lived but as long as it's cheap and good enough than that is all that matters.
    • Well..

      Well there is a reason why companies do run Unix you know.
      In many cases it's actually cheaper to buy Sparc/Power/I2 box
      if you can save a software license, which often cost much
      more than the hardware box. Sure with CORE2 which is lightning
      fast the cases where this is true have become fewer.
      And if Linux is your OS of choice, then you should run that on
      the hardware. And it'll run well on all the 3 above mentioned

      You forget that chips is only a tiny part of the cost of a

      And if you do move around at customers, you'll notice that the
      4 x T5120 that gets installed might be replacing 2 big old SUN10K's. And the p560Q that is installed is replacing a p690.
      And the old V class server is replaced by a rx6600.
      The Unix hardware revenue is stagnant, so is the number of units
      sold. But the amount of raw UNIX power that gets installed and
      IMHO also the number of OS images, is growing.
      But no IDC or Gartner numbers will tell you that, cause they
      aren't measuring that.
      And despite a fierce rivalry between followers of the different
      UNIX religions, an AIX guy will rather see a HPUX box in his
      server room than a WININTEL abomination.

      // Jesper
  • Solaris on MIPS64 anyone?

    Attractive parts: low power, low cost, wide use in embedded systems.
    • And the dream combo: Solaris on Power

      There is a PPC port .. but I don't see IBM doing the right thing anytime soon.
      • Or Solaris on IBM OpenSPARC?

        Rumors are, that IBM is interested in building OpenSPARC CPUs...

        You heard it here first!
        • Since were talking rumors

          I heard that IBM wants to aquire SAP. THAT would be crazy eh? That would really give them the full stack.
          • Nahh..

            SAP was started by former IBM'ers that didn't like Evil Blue afaik.
            So I would guess that will be hostile takeover :)=

            // Jesper
          • No it wasn't

            SAZp was founded by a couple of IBM Europe Managers who fell in love with the Future System and realized how great it would be as an ERP host.

            Then IBM pulled the plug on it .. and that's what caused the rancor (rancour?) because they had their software not exactly ready to go, but getting there.
          • Well..

            I don't really have the facts to say if what you are
            saying is neither right or wrong.

            My point was that I figured that the 5 guys wasn't really to
            happy with their former company.


            So if IBM was to buy SAP it wouldn't be a happy wedding, but
            more like a shotgun one aka hostile takeover.

            // Jesper
        • I've not heard anything about that

          but if they made them for their own use (rather than on contract for Sun - which they've done before) it sure would be cool.
      • Just kidding...

        I think they evaluated that buying SUN wouldn't be worth it :)

        // Jesper just kidding
  • RE: Linux: it's the last word on AIX vs. Solaris

    I disagree with your AIX assesment, but then again you already know that. So for me the question is "Why Linux on SysP?" The only good reason I can come up with is to support IBM's consolodation strategy. Buying a licence for RHEL or SUSE is probably more expensive than an AIX licence so its not a cost thing, Linux has nowhere near the RAS features of Solaris or AIX, Linux still lacks the 'executive' credibility of Solaris or AIX so you'll never see it running a 3000 user + ERP installs anytime soon. Nevertheless IBM remains comitted to Linux. It will however run small to mid tertiary databases and applications with great effectiveness. (But so will Windows)

    I heard that IBM's VIO (virtual IO) image going forward will be be based on Linux and not AIX in the near future.
    • If I had to make a cost case..

      1 - buying support (aka "a license") for Linux is voluntary. So buy support for the pre-production test machine, but run on many.

      2 - applications are much cheaper and there's lots of freeware

      3 - Linux skills are more widely available than AIX skills.

      4 - there are integration values from using Linux on everything. Simplification - software portability - etc.
      • heh...

        "3 - Linux skills are more widely available than AIX skills."

        At least there are a lot more people who THINK they have Linux skills ;)
        • Amen (NT)

          • Yes :)=

            You should be able to get a vaccination against such people.

            // Jesper
  • Linux Isn't Ready to Depose AIX

    Regardless of the AIX's merits, the principal reason that IBM hasn't replaced it with Linux already is that the latter doesn't scale. That and the lack of enterprise featurss that datacenters expect.

    It's a commonplace in the AIX community: every couple of years, a rumor goes around that AIX is going to be replaced. IBM executives have in past said as much, but they underestimated the inherent difficulty and so AIX is still here. They have to use something to sell the pSeries.

    It wasn't until the 2.6 kernel that some serious memory and threading issues were addressed. And the notion that a "community supported" OS, ala The Bazaar concept falls on its face when the hardware starts to get serious. How many coders have a p570 in their basements to play with? Zip. Development at the high end ceases to be community driven and enters the province of corporate decision making.

    So what you're proposing won't be Linux, it'll be an IBM creation, heavily modified with the Linux moniker. More to the point, it's not here yet.

    • I agree - with qualifiers

      First, Linux on small Power machines works reasonably well. Not efficiently, but well enough.

      2) most RAS features enterprises insist on were developed in the days when hardware was inherently unreliable. That's not the case any more - most hardware failures now are human mediated: sticking the wrong board in, forgetting to reset p[ower filters, loading destructive software, running the thing over with a forklift etc.

      3) IBM claims to have hundreds of people working on Linux. I don't see them achieving anything - in part, I think, because the effort isn't focused on tech improvement. Focus that effort on Linux on Power and you could hope for the kind of hw/sw matching that works for well for Solaris/SPARC.
    • Linux is ready

      Linux, as provided by both Redhat and Suse, scales fine, runs nicely, and performs very well on all of the Power systems, from the low-end up through the 64-core 595 systems.

      The Linux community continues to drive lots of new serious features which are rapidly being incorporated for corporate use.

      The two Linux distros, like any operating system, nicely balance customer stability for production with the rapid evolution of features being integrated into mainline and the whole software stack. Both Redhat and SUSE understand how to support corporate level production environments with Linux, and on Power systems. It's all here today.
      • Nice Try...

        As soon as Linux supports all the DLPAR functionality provided by SystemP we'll talk. Hot memory removal/adding and hot processor addition/removal still is not supported. (AFAIK anyway)