Drink Beer and Geek Out

Linux vs. AIX

I’m the Director of Infrastructure for a Fortune 500 retail company.  As with many companies, the recent economy has been a tough one to survive – putting extra pressure on us (and other companies) to cut costs, while providing the same or very similar levels of service and up-time.

One area of interest has recently been peaked for me as I’ve watched other companies migrate from expensive platforms such as IBM’s Z-Series or P-Series systems to Linux (generally RedHat or SUSE) running on commodity Intel x86 (x64) hardware.  Verizon for example, recently migrated their SAP and PeopleSoft systems to Dell servers and RedHat Linux from IBM P-Series running AIX.

For example, a new IBM Power 6 system will cost you somewhere between $230,000 to a cool million bucks depending on the configuration you order (also depending on whether you pick refurbished or new – but a refurbished Power 6 may be hard to find right now).  A Dell R900/R910 will run you somewhere in the $15,000 to $35,000 range.  To top that off, from a pure horsepower perspective, the Dell R900/R910 platform with 32 cores and 512GB of RAM (at six to twenty times less money) will outperform a P570 system (I can hear the groans of AIX fanboys already).  I must clarify, there is more to horsepower in this equation.  IOPS, bus bandwidth, etc matter too – but my point remains the same.

Pros and Cons (AIX vs. Linux)

ISSUE/FEATURE WINNER
COST
SUPPORT
HDW INTEGRATION
FLEXIBILITY
MARKET SHARE
VENDOR DIVORCE
STAFF SALARIES
SECURITY FIXES
PERCEPTION
LINUX
AIX
AIX
LINUX
LINUX
LINUX
LINUX
LINUX
AIX

By now, you’re either a Linux fan, and you’re singing my praises, or you’re an AIX fan and you are cursing my existence!  Not so fast my friends.  Read on as we break these categories down.

COST – LINUX VS. AIX
I doubt anyone can argue with me on this point of contention.  Linux (as in the OS) is basically free.  Intel and AMD x86(x64) hardware is a commodity server that is both inexpensive and easy to find.  This is a clear WIN for Linux.

SUPPORT – LINUX VS. AIX
While it would be fairly easy to argue that Linux community support (especially when combined with a paid support contract from RedHat or other) could be considered better than AIX or other proprietary UNIX OSes, I think this will be a pretty simple concept to agree on.   Let’s first look at Apple Computer.  Apple runs OS X.  Although Apple originally derived OS X from NextSTEP, NeXT is based on the Mach Core, and BSD.  Over the many years, Apple in-house engineers have poured over and almost completely rewritten the entire source.  In the process OS X was born.   It would be pretty naive to think that anyone knows or could know the in-depths of an OS that they did not write themselves.  While the community can support Linux quite well, only the original coding team knows it the best.  I believe this confirms a WIN for AIX.

HARDWARE INTEGRATION – LINUX VS. AIX
Again, I like to use the Apple comparison.  Apple only has to support one hardware platform (or a small number in any case) – hardware platforms they designed, or designed to their specifications.   OS X does not need 2500 chipset drivers, 400 sound card drivers, 3000 DVD player drivers, etc, etc.   Linux faces the challenge (as does Microsoft Windows) of providing support for many different hardware platforms, with many different drivers written by different companies.  This is why OSes such as Windows have such low ratings in the stability arena (well, one big reason anyway) and companies like Apple can make rock solid reliable machines and software.   They can dedicate all of their resources to a smaller footprint of drivers and hardware platforms.  This is a clear WIN for AIX.

FLEXIBILITY – LINUX VS. AIX
Have you ever called your favorite vendor and asked for a new feature or option to be implemented in an application or operating system?  If you have, then you know they have a list a mile long from other customers.  The chances of your enhancement being picked or at least being implemented in the short term is next to zero.

With Linux, not only is there a huge community of developers eager to add new features, you have the source code at your disposal – lending your in-house development staff to make those modifications or feature enhancements.  After which, you can release those enhancements to the community for inclusion in future releases of Linux.  This in my mind is a WIN for Linux.

MARKET SHARE – LINUX VS. AIX
Although market share may not be important to you, it is to many IT shops.  This revolves around the rise of Linux and the fall of UNIX variants.  The lifespan of Linux is likely to be decades longer than AIX or others.  This must be considered in any IT long term strategy.  This is a WIN for LINUX.

VENDOR DIVORCE – LINUX VS. AIX
If you’ve spent any time in the IT industry you know that a divorce from a vendor or two along the path is inevitable.  What happens when you need to change hardware platforms because of issues with a vendor?  If you’re running AIX, you’re stuck with IBM.  When running Linux, your hardware platform is your choice.  This also gives you a great deal of flexibility when negotiating hardware prices.  This is clear WIN for Linux.

STAFF SALARIES – LINUX VS. AIX
Proprietary operating systems historically command higher salaries for support staff (Administrators, Programmers and the like).  Linux administrators are generally more plentiful and easier to find.  This has to be considered when choosing your OS.  This is a WIN for Linux.

SECURITY FIXES – LINUX VS. AIX
As you’re probably familiar, it’s not unheard of to wait weeks or months for your OS vendor to release a security fix.  With Linux, your likely to see a patch within hours, or at worst days.  WIN for Linux.

PERCEPTION – LINUX VS. AIX
Have you considered trying to convince executive management that Linux or any other Open Source solution is right for your business?  Linux in many circles is still considered a risky move.  Many CIOs and other executive IT Management are dead set on their idea that Linux is not enterprise ready.  Lucky, that school of thought is becoming more and more old school as giant corporations and web sites are turning to Linux.  Facebook, Google, Amazon.com and other massive web presences are all running on Linux successfully.  However, convincing your management to bet the farm on Linux can still be difficult.  Remember the old saying “No one ever got fired for hiring IBM.”  While that certainly is not true, it is still the perception in many shops.  This is a WIN for AIX.

17 comments

  1. Well, you can use a vendor distribution for Linux support (Red Hat, Novell, Canonical, just to name a few). They’ve written many parts of the operating system (especially the first two) and have top-notch support services for it.

    Well-loved. Like or Dislike: Thumb up 7 Thumb down 1

  2. Dave Airlie /

    I’m with Yaniv, to say Red Hat hasn’t got a developer capable of supporting nearly every corner of the Enterprise Linux distro is off by a long way. There is bound to be large amounts of code in AIX where the developers have left the company and nobody left has any idea how they work anymore.

    I won’t argue with the hw support though, untargetted hw requirements can always be a pain, but without that you’d just have the same vendor lockin problems as AIX.

    Rate this comment: Thumb up 4 Thumb down 1

    • Roedingers Cat /

      We use both in our company – AIX on 200 LPAR’s and 6 Big machines – and Linux on 80 servers (SLES).
      There is no problem in getting support 24×7 for Linux.

      Out greatest problem ist on one hand the hardware support (especially HP), where administrators spend most of there time in analyzing and waiting for response from the back level.

      On the other hand we got problems in analyzing crashes of applications on Linux. There is no such thing like an error-report (errpt) or a system dump that can be analyzed. IBM can do that. Novell and RedHat can not!
      So we get a fast problem analysis from IBM (always in the past 15 years) and a slower analysis from Novell.

      So everybody should consider what is more important. Saving money when buying hardware or saving money when having a crashdown.

      These are just my thoughts.

      Well-loved. Like or Dislike: Thumb up 10 Thumb down 4

  3. Linux Fan /

    The argument “Linux is not ready for prime time” has been debunked over and over and over. You ever fly? I do – weekly. The Federal Aviation Administration uses Red Hat Enterprise Linux to manage the 8000 planes in the air over the US at any given time. I’m not sure that workload could be any more “mission critical.” http://customers.redhat.com/2007/10/15/federal-aviation-administration/

    NYSE Euronext is running on Red Hat Enterprise Linux. $141 billion in trades per 6.5 hour day. That’s over $6 million per *second* – they didn’t choose RHEL because of cost, they chose it for reliability and support. http://customers.redhat.com/2008/05/12/nyse/

    Bottom line is, the old-school thinking “no one got fired for buying IBM” is just that – old school. Today, it’s all about collaboration, speed, reliability and support. If you can get that kind of performance and reliability for a fraction of the hardware and software cost, the question is why *wouldn’t* you?

    If you’re not adopting Open Source, you’re already way behind.

    Hot debate. What do you think? Thumb up 5 Thumb down 3

  4. StockCrank /

    –Insults removed by admin– Good luck with your Dell/RedHat combo. Come talk to me when it starts throwing random ACPI kernel panics. I know a solution that fixes it about 60% of the time.

    What a vast oversimplification. So go out and buy your Dell R900 for 35k. You going to run all your SAP tiers on one box? Of course not. So take SAP – ECC. You are going to need 2 of your sweet Dell boxes for DB redundancy. $70k. Oracle RAQ doesn’t work well with SAP, so what you going to use. Veritas Cluster/Storage foundations most likely. OK .. shell out another 15K per pair of boxes for that. 85k On top of it all your are going to need several several additional boxes PER TIER. A CI and several application servers. You’re looking at about 19K EACH. A Fortune 500 company would probably be needing 3-4 app servers minimum per SAP application to reduce load on the CI. A R710 would probably work nicely for that at about 15k ea (60k) Now multiply this by the number of SAP applications you are deploying. Lets be conservative: ECC, CRM, HR, SCM. 240k in app servers + 85k = $325k. That just bought you your production environment. Do it all over again for QA. You could probably skimp a bit on DEV. I’ve done this excercise — the RFP’s really aren’t that much different cost wise.

    Want to use virtualization to consolidate this horrendous hardware sprawl? Bzzzzt sorry. SAP doesn’t support any X86 virtualization technologies in anything more but a testing capacity.

    Sure you are paying 250k for something like a loaded P570 but you are usually also getting the full boat. PowerHA, DLPAR, PowerVM. And you can achieve far greater utilization and consolidation economies on the P gear.

    So go ahead and risk your job buying computers from Dell, operating systems from RedHat, Clustering from Veritas. I’ll just get a solution from IBM. (or Sun or HP for that matter)

    Mr Civikminded, or should I call you “Will Boege, IBM Global Services employee”? Ahh… the power of Google. I’ve deleted your insults as they are irrelevant for this conversation. But I do want to point some things out to you.

    1) Don’t come on here spouting insults. It’s a waste of everyone’s time.
    2) Arguments from IBM employees just drive home the fact you guys are scared of Linux.
    3) I never said we WERE switching, we’re evaluating. And that’s what scares IBM the most. Obviously we’re not idiots. Obviously we will test and evaluate this idea. Obviously we would never deploy a solution that is flawed, that’s what testing and evaluations are all about!
    4) I followed many of your posts on comp.x.x and zdnet, digg, etc. You have a problem with anything non-IBM and like to fly off the handle often don’t you? Geez dude. Why all the anger in your posts?

    - Mike

    Hot debate. What do you think? Thumb up 6 Thumb down 7

    • Stockcrank /

      A) I don’t think I was ever insulting. Sorry If I came across that way. Sometimes I get passionate.
      B) I think its pretty tacky/creepy that you release someones personal information on your site when they disagree with you and take the time to research them — Especially when the comment form I’m looking at right now says “Mail (will not be published”
      C) If you want to use google on me you’ll see an extensive history of contribution to the Gentoo community, Dshield.org, and various open source projects. So you think I’m some kind of AIX partisan, you’d be wrong.

      Well-loved. Like or Dislike: Thumb up 10 Thumb down 2

  5. UnixDude /

    I’ve been around AIX since 1993 and I’ve worked on as a hands on admin as recently as AIX 5.3 w/VIO on 64 way p595 hardware. I like IBM AIX/pSeries but it’s not bulletproof as IBM marketing will lead you to believe. I’ve experienced numerous AIX crashes (on 5.3+) that required kernel developer involvement and I’ve seen our expensive p595 hardware crash hard due to hardware/firmware failure of a non-redundant nature. Above all, it’s expensive.

    Linux/Intel Nehalem is a serious contender to proprietary Unix/Risc. I’ve seen the industry benchmarks and performed my own Oracle/Data Warehouse benchmarks between P6 and Nehalem. The performance gap has been closed with Intel QPI technology.

    Commodity hardware is finding its way into our data centers. The long term question is simply, “Do I want that hardware to run Windows or Linux?”.

    Rate this comment: Thumb up 4 Thumb down 2

    • AIX Guru /

      I too have worked with AIX since 1995, and I have NEVER had to have a kernel developer get involved.

      This is purely grandstanding by a Linux fan, trying to dissuade AIX users.

      We have Linux/Dell where I work and the pain can’t be anymore evident on the staff. Red Hat never has an answer why anything happens, hardware crashes regularly.

      Linux doesn’t have an good LVM like AIX, disk management in Linux is shoddy for wanting to be a production environment OS, clustering in Linux is a joke at best (daily problems here, with no help from Red Hat), virtualization is clownish at best. Using the Satellite server one has to pay hundreds per server for provisioning licenses, hundreds more per server for management licenses, hundreds more per VM for management; so when you have 1000 servers, you have a serious annual outlay, just to use a Satellite server (that is free with AIX NIM).

      With AIX you get Workload Manager, Kernel Recovery, Storage Protection Keys, Advanced First Failure Data Capture, ProbeVue, cluster systems management. HA built into the kernel now, DLPARs, workload partitions, live partition mobility.

      All integrated nicely and works together with AIX. Plus with IBM support, you don’t get an answer that basically says “oh, well. Live with it.” And you don’t have a ticket open for 9 months that support at Red Hat hasn’t even done anything with.

      AIX is the only way to go for mission critical workloads.

      Well-loved. Like or Dislike: Thumb up 20 Thumb down 2

  6. Hidden due to low comment rating. Click here to see.

    Poorly-rated. Like or Dislike: Thumb up 4 Thumb down 7

  7. And one last thing. Whether IBM’s premium price is justifiable or not pales into insignificance compared to the heap of sh1t that is SAP.

    Its architecture is from the late 1970’s and is just about the best way to ensure you need to pay IBM a lot more than you first thought.

    The SAP marketing message that it “Solves All Problems” also fools only those gullible enough to look no further than the first paragraph – so usually the IT director then. Once you’re in, you have the widest scope application system on the planet, almost none of which will fit the business functions you need – and you’re fully locked in with the only resource able to help you a limited pool of very expensive independent consultants and an even more expensive group of consultancy firms who really should know better by now.

    Hot debate. What do you think? Thumb up 6 Thumb down 4

  8. Jeff Collins /

    Look at IBM’s new p7 750′s and the economics become much more blurred.

    Well-loved. Like or Dislike: Thumb up 9 Thumb down 4

  9. Rakesh Singh /

    I’ve been working on AIX for about 9 years and at the last company, we had over 900 LPARS. On my personal machines, I run Linux and have been doing so for about 7 years. Unfortunately, as other mentioned, Linux never gained momentum as the powers that be had the ‘warm fuzzy feeling’ with Solaris and AIX (and even Windows).
    Personally, from a hardware perspective, I love the virtualization options on the p-Series with its micropartitioning, shared processing & memory pools etc.
    On Linux, I miss AIX’s LVM, NIM, hardware management, WPARS, Live app mobility etc.
    Maybe I am fortunate, but we never had any major issues on AIX. Neither have I had major issues on Linux or FreeBSD.
    The biggest claim is that the Power6/7 chips are the fastest cpus around, but I’ve heard otherwise from those with new Intel cpus. I’d like to see independant benchmarks besided TPC results etc.
    If the performance of Intel comes close to, or betters the Power cpu, and one does need the ‘bells and whistles’ of the p-Series/AIX combo, then the Intel/Linux combo makes the most sense.
    Question is, would you then run Linux or perhaps Solaris/OpenSolaris & benefit with zones, ZFS etc. ? :-)

    Well-loved. Like or Dislike: Thumb up 12 Thumb down 0

    • Allan /

      I’ve worked with AIX systems for 20 years now. Fantastic platform. Expensive platform.

      While IBM has reduced the hardware pricing with the new p7s to the point where it’s actually quite competitive, all things being equal, it’s the maintenance pricing that’s the killer; over a 3-year period, the software and hardware maintenance pricing will exceed the purchase price of the system.

      Strangely, the majority of the maintenance price is the software maintenance on AIX, not the hardware. Annoying in the extreme; in my case, I’d prefer to run on AIX, but the maintenance costs make it impossible to argue with an ESX setup on commodity hardware.

      Well-loved. Like or Dislike: Thumb up 17 Thumb down 0

  10. thanks buddy for such nice post

    Anuj

    Rate this comment: Thumb up 3 Thumb down 0

  11. Allan C /

    Just to be fair on the topic of Vendor divorce. You can leave IBM and get hardware from Bull which runs AIX.

    Rate this comment: Thumb up 1 Thumb down 0

  12. chris Carter /

    I am talking with some folks on taking the AIX / SAP ecc6/ Pseries and keeping the Production on it BUT taking everything else off and moving to x86 with Linux..so all Dev, QA, test & training would then flow into it rather then all the cost’s on the AIX & P-series.

    It is a great strategy in my opinion and from a few folks I have spoken to it works well…has anyone got an opinion? I am trying to get as much research on this as possible…

    cmc@hiln-solutions is my email with any concerns or thoughts..Thanks all!

    Thank you
    Chris

    Rate this comment: Thumb up 1 Thumb down 0

  13. Interesting note or omission.

    There aren’t many “pure” flavors of Unix. For the first handful of years, OS X and Linux were among them, not pure – not certified, not in the class of AIX, HPUX, Solaris, etc.

    But fully 2.5 years before this thread started, Mac OS X has been “pure” Unix. So one isn’t really comparing “Apples to Apples” (forgive the pun) in this instance.

    And while there clearly is no necessity for an OS to BE “pure” – the set of standards by which one particular OS establishes for themselves, from a Developer perspective ( “developers Developers DEVELOPERS!” ) – that’s something that can’t be said of Linux, let alone any particular flavor of Linux.

    That being said (and here’s the certification to prove the above — http://www.opengroup.org/openbrand/certificates/1190p.pdf ) – the FAA example is indeed a good one. So too, however, is the DOD use of 1760 Sony Playstations, using the IBM “Cell” CPU, to run Linux – as the single most powerful computer under the Dept of Defense umbrella!

    That “half petaFlop” computer came in, (rough recollection) at 9% of cost for off-the-shelf purchases vs. a “Big” machine (be it IBM, Sun, HP, or for that matter, Dell’s server division or even, until they were cancelled by Steve, the 2U version of the Mac OS X Server – with twin Xeons (granted – not in the 10 core, 8 socket realm of the latest Xeons, but not meant to have been, but just as scalable) – as well as using 1/9th the ENERGY.

    You can find the details of the system via Google – and start here. http://blog.cleveland.com/metro/2010/11/defense_department_discusses_n.html

    The info is already 1 year old… but considering the ever expanding power of the CPUs (when Power8 hits the shelves soon, with 16 cores, 64 threads, and 600gFlops per CPU, it changes everything, despite Intel ~50 core announcement of recent) – the key will simply be matching the CPU capabilities with the OS and application to achieve things on a scale we really couldn’t have predicted accurately enough in 2001.

    Power8 is just around the corner for 2012. Intel is making great strides as well, but not quite like the Power 8 for production.

    And don’t forget — while Apple threw up its hands and moved to Intel not long ago, the “G6″ that Apple was waiting for Freescale (formerly Motorola) finally appeared this quarter. A 12 core CPU with 2 threads per core (not 1, but not 4, like the Power 7) complete with a new, advanced Altivec / SIMD code, that clocks in at just under 200gFlops, and you can scale them on the order of 600+.

    In that sense, when you look at what is inside the F-22 and F-35 (apparently no better than a dual core “e600″ – aka the G4 processor) still at 90um, compared to the newer (and newer Power Architecture standard, rather than the original, which the F-22 called for, to take advantage of improvements downline) – the new Freescale CPU is 50x the power in the same space… and whatever it is that makes that plane so superior, clearly isn’t linked to just the DSP capabilities of a 15 year old CPU.

    It’s sad, however, that we are stuck with, at $369,000,000 each, (after factoring in the costs of only 187 planes, vs. 800) that each is running on the same stockpile of avionics that run on Linux and proprietary aircraft systems, and a “G4″ – that has been sold to Russia since May of 2000 for their Tupelov business jets.

    And it only gets worse…

    The original PowerPC G3 — is still available at 1/3rd its Macintosh clock speed (from the original iMac) TODAY, radiation hardened, and screaming at 100MHz, for just $200,000 per CPU.

    Makes you wonder if that old iMac you just threw away might be worth a bit more with some foil over the top of the G3 !!

    You’ll find at least 15 of these CPUs (Google it) – in use, on deep space probes. And several that are still on the ground and may never be launched, given the demise of the Space Shuttle.

    Which, by the way, was given an upgraded 80386 at one point, across the entire fleet, which is the primary reason that virtually all of the occupants of the shuttle had their own MUCH MORE POWERFUL HP calculators with them…

    Clearly the combination of CPU, OS and application cannot be fully evaluated on the basis of just one leg of the triad.

    Rate this comment: Thumb up 0 Thumb down 0

Leave a Reply