Skip to content


is future of AIX my future too?

Lately, I am a very busy person; I had to learn enough LINUX to install it on pSeries hardware, only to quickly realize that the level of support in relation to this platform is, let’s be honest, abysmal. On the other hand, if you do LINUX on INTEL/AMD and/or VMWARE the IBM experts are fast and the support you receive it up to what you expect.

After installing a pair of LINUX “hosts” in fully virtualized lpars I was sent for LINUX training. While being trained, I soon discovered the beauty and elegance of VMWARE. Upon my return from the training and with the “help of my friends” I set one RedHat quest on one WMWare cluster (DEV/TEST) and copied it to the other two (PRODUCTION clusters in two data centers). From now on, to build LINUX HTTP/TomCat box all that needs to be done is to clone the copy! By the way, before the copies were made, we disabled the networking services on the “gold” copy. One has to spend just a few minutes in front of a WMWARE to understand its power, elegance, and simplicity. Yes, here there is no more LINUX but on VMWARE.

Yes, I can hear it. Your voice is laud and clear. “How could I do that to AIX?” Why? Because my love is not blind; I want AIX to be the best OS there ever been. I do not want AIX to STAGNATE! I want to retire doing AIX! Is it enough for you?

To be honest, I always wondered why AIX persisted to stick around; each and every technical presentation I attended or read never really demonstrated the full potential of this OS. They never talk, nor exploit mksysb, file system snapshots, the fact that a file system can be increased/decreased while mounted, about smitty and hundred of other features of this OS.
Yes, AIX has many great futures like, for example, NFS4 ACLs or wpars but they are implemented and presented in such an old and arcane mode that most admins to not even bother to read and learn about them. Whose fault it is? Were these features implemented just to be able to say that AIX has them?

Virtualization is great and there is no way to neglect it or to avoid it. But why do you have to spend hours to read and learn before you can implement it? I did not read a simple manual but was able to use VMWARE – I got a five minute presentation from a WIN admins in our office. My time is expensive and in a short supply. I have family; my wife and children not to mention our dogs and cats – they all require attention and the time to interact with. My lawn needs trimming and I would like to have a free moment so I could go and do some fly fishing? Why do I need to spend hours learning to virtualize? Is it my fault (I might be not the smartest kid around) or it is the GUI that I can blame? Why I must separately select the slot for a VSCSI adapter on the client and VIOS server side? Why do I even need to be concerned with the slot? Tell me why? The only reason I will accept your answer is if you say that it is so to exercise my own mindfulness otherwise I think that this is just plain silly.

Maybe all what is required is to have HMC “GUI” designers to download a copy of Oracle’s “Virtual Box” or VMWare in order to get some inspiration? Maybe it is time to stop repeating how great AIX is and to return to the drafting board to make AIX and its components match the times (2012) we live in? It is not 1990 any more. I am with AIX since 3.2.5 and I have not seen any improvement in its access and control methods. Why? Are we perfect already? Who said so?

While I am at it, why AIX admin is restricted to virtualize only vertically? Why is he limited to virtualize only within one managed system/frame? Why is AIX virtualization not able to unify more then one managed systems? Being able to virtualize in both directions, why not to be able to detect a failing memory, disk or adapter and automatically provide its replacement with an appropriate element from another frame or frames? Too difficult, you say? Do not tell me that LINUX is simpler to use because it is smaller. If AIX is complex because it is big then hide its size and complexity so it is easier to use.

Today, I had to upgrade and install some patches on one AIX host and LINUX. Compare suma with yum, nope you just cannot compare these two -yum is better, no hassles no complications – a very straight forward process.
You say that LINUX does not have enough power for big databases.The existing INTEL/ADM hardware does not have enough horses but for how long? Additionally, if I can harness power of several smaller INTEL servers …….. why do I need one “big” iron? Indeed, the market is changing not just fast but in a very rapid pace. I remember people saying that SUN/SPARC (the darling of Wall Street) will never go away – such a powerful hardware! But where is it now? Have you seen SOLARIS lately?

Have you ever tried to integrate your own menus with smitty? How many of you read its customization manual and did not experience a form of a mental breakdown? It is really to difficult to revise it to make it more accessible and easier to use? A lot of programming jobs have been outsourced overseas, cannot one or two of these more cost effective programmers be used to write a new smitty customization module with a modern interface. I am not kidding, I really would love to be able to easily incorporate my scripts and other tools in smitty menus and I think there is more people interested in an easy smitty customization. Could you help us?

Is it possible to turn the existing firmware upgrade process into one single step including an automatic download of code for the selected hardware component?

A the standard answers go, this one is often used – “well, there is just not enough demand for this so we decided to invest our resources somewhere else”….. sorry, I do not buy it. People lose interest in something they cannot understand in a day or two aka they lose it quickly; there are deadlines to meet, work orders to close, …. there is never enough time. So, it is got to be easy to “COMPREHEND” in order to be used – in order to be in “DEMAND”! It is the fault of the maker if there is no or very little demand for his product! So if you make it easy to use it will be used, I bet.

As for my retirement as AIX administrator – only the time will show …. . For now, having Darwin’s theory of evolution in mind, to survive as UNIX admin I have to keep adjusting to the ever changing environment around me with no attachments nor preferences. LINUX is now part of this blog too.

It is the question of a passion and a vision that strengthens one and lack thereof weakens another. Not to mention that blind applause is poisonous.

Carry on AIX admins!

Posted in Linux, Real life AIX.

Tagged with , , .


12 Responses

Stay in touch with the conversation, subscribe to the RSS feed for comments on this post.

  1. Hajo Ehlers says

    > is future of AIX my future too?
    i have been working with AIX since 4.1.5 and i still see it as one of the best unix systems around …. but i just hope to survive the next 10 years and then retire since imho AIX is on its way down.

    Its not because AIX is bad or is not up to date. Its because AIX is an oldfashion OS in a world changing into “clouds” ( In AIX terms PSSP :-) . In these environment it is important to easly manage your VM. But as i said ” To manage”. The VM itself is just an container for an appllication. So the view is changing from the OS – which simply must work – to the management of these VMs.

    But do we have anything in the IBM world that allows us “easily” to manage and virtualize AIX or other power based systems ?
    IMHO there is NOTHING .

    The VIO: Its a joke in terms of management and configuration, Our ESX hosts using shared storage since we started with VMware . The installation of an ESX takes 5 minutes and the customization is a 10 liner script. I never do a backup of a ESX server. There is no reason.
    For managing there is a very good GUI and very good command line tools. Reminds me on AIX – SMIT and CLI.

    So compare the HANDLING of a VIO and an ESX system. The VIO will loose . On a VIO you will even mentaly suffer. Meaning: If i have to work with the VIO i get paid for my suffering

    Note: The lastest update for the VIOS ( 2.2.1.4 ) mentioned that you have to shutdown the WHOLE VIOS cluster for the update of a single VIOS in case you use shared storage. In the vmware world this would be joke.

    SYSTEM DIRECTOR:
    One upon a time i got a DVD with system director ( i think some 6.x version ) and the recommandation to download a pdf for installation and how to use it. After i got aware that the PDF had 800 pages i decided to stop even installing SD.

    Month later during a DS session i had been told about the latest feature of the SD – they are able to deploy images. As i asked if they support installation from scratch i simply got an “i do not understand” look at me.
    I am not quite sure but i think the hardware resource to run a SD where close to a 6 core p740 with 128GB of ram.

    SOFTWARE
    Have you ever tried to get opensource software running on AIX. For example python 3.x with all the dependencies ? I have got even in touch with IBM asking if they could not support the OpenSuse Buildservice to extend it for the power platform and aix. Suse was already interested.
    The answer was simply: We do not think we will make money .
    Beware that we are talking about a maximum of 100k $ to get this working.

    The result is that i suggest to my customer to use linux since there we get what we need.
    And to IBM: You are right – you not do make money. You even lose money since we are moving away from AIX.

    So what is my conclusion. AIX will survive but will be slowly dying. I just hope that it will not die before i retire.

    Its sad to see this happen and i would be happy to be wrong. But i do not know why IBM should behave differently then they did the last 15 years.

    Hajo

  2. MarkD:-) says

    Sebastain, thanks for your comment! The System Directory is another Sticly Point for me too. I tried it and I have the same experience as you. I wasted my time and effort and I am not about to try it again. What a pity.
    By the way, why this has to be a separate product? Could HMC be made to do all what System Director may eventually do? From what I heard, HMC will always be around aka System Director will not replace it.

    MarkD:-)

  3. Sebastian says

    Hi, I love AIX since I started with version 4.3.2 13 years ago. It was my fist UNIX OS and thoght its ok. Then I looked around my colleagues on HP-UX and Solaris and loved AIX for all the cool features it had. With Power4 Lpars I liked it even more until I had dlpar problems, that were anoying in power 4 times. Power 5 and 6 came, got the dlpar problem solved and introduced the not so easy to configure VIO servers. LPM got added, really beautiful things but hard to configure them. I love the AIX SW/HW stability but the Administration of the LPAR got way complicated after power4. Even when you manage to get it all right with dual vio, mapped storage and even a clusted lpar on top if you did not document each step you did carefully, youre lost! Neither the VIO nor the HMC can tell you “quickly” which lpars are affected when you loose one of the IO adapters on the VIO. The HMC is nice and ok for their tasks and powerful on commandline but it lacks a intuitive interface like I find in my LINUX VMware Vcenters. They tried to force us to use Systems Director. I installed it twice first time 6.0 and second time 6.1 but it’s still a joke compared to a Vcenter. I did’nt even manage to connect to all my clients. IBM builds great HW but their Managment GUIs are a mess and hard to understand. On the last TechConf I visited (Denmark) we had a meet the experts round and we were asked, who uses the SD for AIX Productive, I think we were somewere arround 200 people and only 1 Customer in europe uses this “thing” really to administrate their LPARS. It was no surprise to me, but hopefully for the developers. It seems to me that no single Admin has influence on how Managment GUIs within the IBM AIX devision are designed. See for example this brilliant script from a good perl programming AIX admin: https://www.ibm.com/developerworks/mydeveloperworks/blogs/brian/entry/aix_vio_mapping_virtual_scsi_disks_on_aix_to_physical_vio_disks6?lang=en why is IBM not able to put a similar thing into their gui and make VIO disk mapping a lot easier?
    His entry on generating mapping informations is brilliant as well: https://www.ibm.com/developerworks/mydeveloperworks/blogs/brian/entry/vio_virtual_scsi_disk_mapping_update6?lang=en
    Years after years we are waiting for tools for administration and documentation from IBM but I think the’ll never come. Look at AMS (Active Memory Sharing) together with Memory deduplication, can it get even complicated to configure? IBM has really nice features but the inital configuration does take a lot of time…
    I hope this changes in the future!
    Cheers Sebastian

  4. James Carstensen says

    Jeff, Have you been making a real go with System Director? We bought SDMCs just for dual VIOS on Power blades, but they were discontinued earlier this year. But they were less straightforward than the HMCs, which while a mighty and reliable workhorse MarkD describes well in terms of high complexity.

  5. MarkD:-) says

    :-)

  6. Chris Gibson says

    Got your sights on VMWARE (or LINUX) Champion 2013 perhaps? :-P

  7. MarkD:-) says

    crossing the line and joining the “Dark Side”… it is not my choice.
    :-(

  8. Chris Gibson says

    They got to you, didn’t they…..? ;-)

  9. MarkD:-) says

    Anthony, thanks for you input! I agree with you 100% and more.
    On another note, could you elaborate more on your latest project?

    All the best,

    MarkD:-)

  10. Anthony English says

    Time is a very important reason for simplicity, but there’s another one as well. Younger people are not too interested in IT, because if it’s not point and click, if it can’t work on an iPhone, for example, who is going to maintain it? AIX people in the west are generally in their 30s (at the youngest) or more likely their 40s or 50s. As the technical skills become rarer, and more people want to be able to master the platform in a tweet and a 5 minute youtube video, there simply has to be a simpler alternative for administration, or else companies will left with expensive cars that no one knows how to drive.

    I worked on a small V7000 implementation using EasyTier recently. No need to worry about LUNs, RAID levels, ranks, arrays, storage complexes. The common things should be simple to do, with an easy interface. Maybe PureSystems will be the answer.

  11. MarkD:-) says

    Jeff,

    thanks for your comment! SVC? We started our SAN buying two SVC’s. A few years later they are gone – just too expensive, IBM is charging too much. So we have HDS and some other ones. What luck do you have with System Director? I looked at it two years ago and after realizing that I cannot get much support without buying maintenance I dropped it.

    MarkD:-)

  12. Jeff Q. says

    bravo! I’d like just to say, as a die hard AIX/Power zealot for the past ~14 years of my career I agree with you 110%

    I recently attended an IBM customer feedback meeting where we’re presented with some futures and asked about direction on various topics and more so this year and in the past I hammered on simplicity.. why is vmware/linux taking over in our DC’s? not because it’s more flexible than aix/power.. it’s because it does the important things really easily and I think that AIX people often lose sight of this.. one conversation we had was around RAS in a single frame vs LPM. if you had to do maintenance on a frame would you prefer to rely on RAS features to get you through this maintenance without downtime or leverage LPM? well my answer, for the majority of our environment.. if i truly had the same DRS/vmotion capabilities without all of this headache.. i’d just be relying on LPM, but of course if i buy a 795 it’s because i’ve got partitions i’m not likely to schlepping around to other frames and need RAS.. I want both, but I want IBM to focus on the simplicity.. many other’s in the room were very fixated on RAS features, I think quite a few don’t realize how little their x86 counterparts worry about taking down a commodity box for firmware updates in an ESX cluster. evacuation occurs without much planning or thought assuming your cluster design is solid. the response I get is “but those are all non-critical, small workloads!”.. eh, i think that is becoming less true. at least in our environment..

    the old school admin in me says I have to have root shell access to VIOS, i need to be able to control my slot numbering, my VTD naming scheme, every VM should always be lun backed blah blah blah but the forward thinker in me says make it hidden, i don’t care so long as it works reliably and predictably.

    this is why I put so much focus on Systems Director over the past two years and constantly pounding on IBM to make it better. I want that single pane of glass, i want to be able to truly look at frames as a pool of resources that, for the most part.. balance themselves out, we’ll always have some parts of our infrastructure that are too sensitive to manage this way but if IBM doesn’t really focus on simplicity.. that infrastructure will be the only IBM infrastructure we have. yes i know, IBM has resource pools today but how many are using them anywhere near the same way x86/vmware does it? very few i’d wager. my shop at least has the added headache of non-ibm storage.. good luck making even 10% of those features work for you without hearing the phrase “you should buy SVC”. yeah, sell that to management who just asks, why don’t we need that for our ESX farms?

    the good news is, I think they do understand this.. I think they are working on the problem, I think.. once SSP’s on VIOS go through their next release cycle they’ll be ready for more widescale adoption, not just sandbox playgrounds. I think then you’ll start to see more doors opening up.

    I realize many admins will read this and go “yeah, but!” through every single sentence.. there’s always the corner case, there’s always the “my organization has to have it this way”.. yup, you’re all right. but let me say this:I think that approach has contributed to where IBM is at today, big customers accustomed to doing it their way…. they like it when ibm adds knobs for them but don’t you dare take away the old knobs.. i need those too! it contributes to this very complex, feature rich architecture that requires much more effort to manage and plan than the alternative. it does it lend itself well to commodity management.

    haha, thanks for opening the door for my vent, Mark!



Some HTML is OK

or, reply to this post via trackback.



© 2008-2013 www.wmduszyk.com - best viewed with your eyes.