IBM has around 2,000 developers in Bengaluru working on the pSeries products (AIX, CPUs, firmware, compilers, etc.). In my experience they are pretty fantastic engineers with a tremendous amount of experience and problem-solving skills. Obviously AIX is on its way out, but I absolutely would not underestimate the people who work on it just because they’re in a different country. Many have gone on to do really cool systems and networking startups.
I’ve worked with – and in – several teams like these and I’m with the author on this prediction. I absolutely don’t doubt the expertise of the folks in the Bengaluru office – I was colleagues with some ex-IBM devs from that office, and knowing them – and hearing their stories – they are really good. However, IBM has a history of handling these teams very poorly and not capitalising on their expertise. They are under-financed, have poor access to the corporate grapevine, and have very little control over their direction. Half the reason why so many people from offices like that in Bengaluru go and do really cool systems and networking startups is that there are no internal pathways for them. The cool products their startups come up with could be IBM’s, but no one at IBM who can recognize or steer potential is ever around, and even if they were, they wouldn’t have the funding.
The reputation that these are the kind of teams where projects go to die is well-deserved. It’s not that there’s a shortage of people who could spin them around – there isn’t – it’s just corporate policy. Large enough companies, with high enough funds, can afford to hire really good people and still only send projects on life support their way.
I haven’t thought about AIX for over … um … two decades now? The last time I really used it was in the late 90s and my only good memories of it are of SMIT, The System Management Interface Tool. Yes, it was a menu driven program to admin an AIX system, but at any point, you could have it show you all the commands it will run. I found it very instructive, and wished more “control panel” type interfaces would do the same.
To me SMIT was barely usable. As soon as you ventured into anything more complex it would spew unintelligible mess of shell functions that were not that explanatory. I had better experience just reading the manual pages and learning to use the couple of commands I needed directly.
Someone here periodically posts about AIX because it does shared libraries in a different way to other ELF platforms, which apparently causes some excitement, so I’ve thought about it a few times over the past few years.
Solaris also has a similar thing for admin tasks, though I can’t remember its name (I do remember that it was written in Java and my UltraSPARC IIi with 512 MiB of RAM could run either that tool or a web browser to read the docs, but not both at once). A lot of cloud admin things do this, since they talk to a back end via JSON and can provide you the JSON to save / edit as an intermediate thing.
I’m curious what IBM’s plans are for existing AIX customers. I remember Novell spending a lot of time making NetWare run on Xen so that you could incrementally move to SuSE Linux (owned by Novell at the time) and keep buying things from them. Xinuos (the eventual owners of the SCO IP after SCO imploded after trying to sue IBM) did some integration work to make it easy to deploy OpenServer in bhyve on their FreeBSD-based platform. I talked to them about adding an OpenServer compat layer along the lines of the Linux one but their customers wanted 100% compatibility and if they needed to validate their code on an OpenServer ABI layer on FreeBSD they’d just as easily port it to FreeBSD in most cases. Now that IBM owns RedHat, I presume the goal is for everyone now running AIX to run RHEL, but I wonder how they will get there.
It’s worth noting that IBM i has a kind of dependency on AIX. IBM i has an AIX binary compatibility environment called AIX. Think of it like WSL, but for AIX binaries and on IBM i. Of course if they created this compatibility layer today they’d probably choose to make it compatible with Linux rather than AIX, but they chose AIX and now they’re stuck with it.
This means AIX, or at least the AIX ABIs as supported by IBM i, has to be kept alive for as long as IBM i is alive. So either this bodes badly for IBM i or they consider the amount of ongoing maintenance that PASE needs to be so small it can be handled on an ongoing basis by the i or the new skeleton AIX team. I suspect the latter rather than them canning IBM i though.
That said, I have acquaintances who LOVE it and tell me there is some seriously quality engineering going on under the hood there.
I have less than fond memories of AIX from my sysadmin days - doing multi-hour CATIA installs off of tape, the little SMIT administrative UI had an animation of a little running man. If he fell down, the install failed.
I saw a LOT of that running man falling down in my time :)
Tangentially, I’m actually a bit worried about how Linux is turning into a bit of a monoculture. I’m glad the BSD community is still doing its thing.
The kernel is pretty nice from a variety of implementation perspectives. Most drive by analysis of AIX miss this. LVM, JFS2, extreme NUMA and SMP scalability.
The only several things that remain on AIX land are:
IBM TSM (backups)
IBM DB2 (database)
Most other databases and/or backup system do not even support AIX anymore.
PowerVM helps with its features like LPARs/partitioning but that will not change anything.
SAP never supported AIX with SAP HANA - You need SLES/RHEL on POWER for that.
But even with that other dinosaurs like HP-UX or Solaris often offer more features.
For example you can create several nodes HP Serviceguard cluster under HP-UX and put HP Virtual Machines as highly available services on that HP Serviceguard cluster. Its not possible in IBM POWER ecosystem as only LPAR Remote Restart with HMC is possible - but it is not a HA solution and can not be compared to that HP VMs under HP Serviceguard. Not to mention that you ALWAYS need to have HMC (physical or virtual) which also complicates things.
There once was POWER KVM project but I think it also died long time ago.
Today only OpenPOWER has any future with Linux or FreeBSD as OS.
Also AIX is deadly outdated - its package management is so PITA that only HP-UX can compete on that field :)
From old UNIX proprietary systems (HP-UX/AIX/Solaris) the Solaris had the brightest future (and features) but then Oracle happened and Solaris can be declared dead because of it. I hope that Illumos will not die.
Ah, I remember the time I have (on an IBM support position) asked the dear Indian AIX GPFS kernel devs to do something about the fact that they literally compare running process name with "samba" and then use a different (buggy) ioctl behavior based on that, which the customer running much newer Samba is complaining about. So, apparently what was “WONTFIX, works as designed”.
It seems that there are some IBM Canada folks working on AIX stuff: porting features to llvm-project for AIX. It feels “legacy” that many driver options work differently from common ELF operating systems and they update many tests to adapt AIX…
Didn’t know AIX had a present :) Also, a move to Bengaluru may mean a new, brighter future instead.
IBM has around 2,000 developers in Bengaluru working on the pSeries products (AIX, CPUs, firmware, compilers, etc.). In my experience they are pretty fantastic engineers with a tremendous amount of experience and problem-solving skills. Obviously AIX is on its way out, but I absolutely would not underestimate the people who work on it just because they’re in a different country. Many have gone on to do really cool systems and networking startups.
I’ve worked with – and in – several teams like these and I’m with the author on this prediction. I absolutely don’t doubt the expertise of the folks in the Bengaluru office – I was colleagues with some ex-IBM devs from that office, and knowing them – and hearing their stories – they are really good. However, IBM has a history of handling these teams very poorly and not capitalising on their expertise. They are under-financed, have poor access to the corporate grapevine, and have very little control over their direction. Half the reason why so many people from offices like that in Bengaluru go and do really cool systems and networking startups is that there are no internal pathways for them. The cool products their startups come up with could be IBM’s, but no one at IBM who can recognize or steer potential is ever around, and even if they were, they wouldn’t have the funding.
The reputation that these are the kind of teams where projects go to die is well-deserved. It’s not that there’s a shortage of people who could spin them around – there isn’t – it’s just corporate policy. Large enough companies, with high enough funds, can afford to hire really good people and still only send projects on life support their way.
I haven’t thought about AIX for over … um … two decades now? The last time I really used it was in the late 90s and my only good memories of it are of SMIT, The System Management Interface Tool. Yes, it was a menu driven program to admin an AIX system, but at any point, you could have it show you all the commands it will run. I found it very instructive, and wished more “control panel” type interfaces would do the same.
I also loved the little running person animation. You could watch your SP/2 fall over, in real time.
That’s a cool feature indeed - like “training wheels” for admins :)
And another thing I just remember—it was available via both the command line and the GUI. Again, I’ve never seen anything quite like it since.
SuSE had YaST.
To me SMIT was barely usable. As soon as you ventured into anything more complex it would spew unintelligible mess of shell functions that were not that explanatory. I had better experience just reading the manual pages and learning to use the couple of commands I needed directly.
s/had/has/
https://en.opensuse.org/Portal:YaST
HP-UX SAM could do this.
Someone here periodically posts about AIX because it does shared libraries in a different way to other ELF platforms, which apparently causes some excitement, so I’ve thought about it a few times over the past few years.
Solaris also has a similar thing for admin tasks, though I can’t remember its name (I do remember that it was written in Java and my UltraSPARC IIi with 512 MiB of RAM could run either that tool or a web browser to read the docs, but not both at once). A lot of cloud admin things do this, since they talk to a back end via JSON and can provide you the JSON to save / edit as an intermediate thing.
I’m curious what IBM’s plans are for existing AIX customers. I remember Novell spending a lot of time making NetWare run on Xen so that you could incrementally move to SuSE Linux (owned by Novell at the time) and keep buying things from them. Xinuos (the eventual owners of the SCO IP after SCO imploded after trying to sue IBM) did some integration work to make it easy to deploy OpenServer in bhyve on their FreeBSD-based platform. I talked to them about adding an OpenServer compat layer along the lines of the Linux one but their customers wanted 100% compatibility and if they needed to validate their code on an OpenServer ABI layer on FreeBSD they’d just as easily port it to FreeBSD in most cases. Now that IBM owns RedHat, I presume the goal is for everyone now running AIX to run RHEL, but I wonder how they will get there.
I loved that interface, at least as an attempt to make command line tools more discoverable.
From hackernews:
We always called it Aches :)
That said, I have acquaintances who LOVE it and tell me there is some seriously quality engineering going on under the hood there.
I have less than fond memories of AIX from my sysadmin days - doing multi-hour CATIA installs off of tape, the little SMIT administrative UI had an animation of a little running man. If he fell down, the install failed.
I saw a LOT of that running man falling down in my time :)
Tangentially, I’m actually a bit worried about how Linux is turning into a bit of a monoculture. I’m glad the BSD community is still doing its thing.
The kernel is pretty nice from a variety of implementation perspectives. Most drive by analysis of AIX miss this. LVM, JFS2, extreme NUMA and SMP scalability.
The only several things that remain on AIX land are:
Most other databases and/or backup system do not even support AIX anymore.
PowerVM helps with its features like LPARs/partitioning but that will not change anything.
SAP never supported AIX with SAP HANA - You need SLES/RHEL on POWER for that.
But even with that other dinosaurs like HP-UX or Solaris often offer more features.
For example you can create several nodes HP Serviceguard cluster under HP-UX and put HP Virtual Machines as highly available services on that HP Serviceguard cluster. Its not possible in IBM POWER ecosystem as only LPAR Remote Restart with HMC is possible - but it is not a HA solution and can not be compared to that HP VMs under HP Serviceguard. Not to mention that you ALWAYS need to have HMC (physical or virtual) which also complicates things.
There once was POWER KVM project but I think it also died long time ago.
Today only OpenPOWER has any future with Linux or FreeBSD as OS.
Also AIX is deadly outdated - its package management is so PITA that only HP-UX can compete on that field :)
From old UNIX proprietary systems (HP-UX/AIX/Solaris) the Solaris had the brightest future (and features) but then Oracle happened and Solaris can be declared dead because of it. I hope that Illumos will not die.
Regards.
Ah, I remember the time I have (on an IBM support position) asked the dear Indian AIX GPFS kernel devs to do something about the fact that they literally compare running process name with
"samba"and then use a different (buggy)ioctlbehavior based on that, which the customer running much newer Samba is complaining about. So, apparently what was “WONTFIX, works as designed”.It seems that there are some IBM Canada folks working on AIX stuff: porting features to llvm-project for AIX. It feels “legacy” that many driver options work differently from common ELF operating systems and they update many tests to adapt AIX…