Originally posted by: Edward_R86

I wasn't sure if this was the best forum to ask this question but couldn't really find one specifically for this sort of hardware and what to do with it. 

Basically, I have this IBM Power Systems 710 Express Machine which might be loaded with AIX and if not I'll load it with PowerLinux of some kind, but anyway I've gathered from my research that I can do various administrative tasks via ASMI by setting up an ASCII terminal since I do not have a graphics card and can't use a monitor of any kind and it seems there's even a web interface for it as well but the part where I get lost is how do I manage the OS itself after the system has been booted up? 

Apparently this ASMI approach no longer becomes useful when I want to issues commands in AIX or PowerLinux since it's not available during IPL or runtime. So what options are left when someone does not have a monitor and needs to do stuff with the OS itself?

I'm new to these type of machines and looking for guide and swimming through mountains of documentation for a very specific thing isn't easy.

Originally posted by: Wouter Liefting

Your Power 710 will have a "Hypervisor" running on it. This runs on the Flexible Service Processor (FSP) and is a kind of mini-OS intended for virtualization. In the Intel world, this can be compared to VMWare ESXi. Apart from quite a few technical differences, another difference is that you cannot remove this hypervisor. (*)

When the machine is physically switched on, the hypervisor starts running. It may take a minute to fully boot. Once it's running you can communicate with the hypervisor three ways:

1. Via a serial console (or PC with a nullmodem cable and emulation software such as minicom).

2. Via the ASMI interface (https protocol)

3. Via the HMC - which is a separate machine running specialized software and useful if you manage multiple Power machines. It can roughly be compared to VMWare vCenter.

Options (1) and (2) allow you to start and stop the machine, which effectively means starting and stopping its single partition or main (IVM) partition. Option (3) allows you to create and delete partitions (similar to VMs on VMWare), start & stop them and so forth.

In your case, it seems you don't have an HMC. Most likely your machine will be in the "factory default" configuration of having one big partition which has access to all hardware. (You can reset the machine to factory default from the ASMI menu - options 1 and 2 above.) You can indeed install AIX or Linux on Power on here (note that "PowerLinux" is actually referring to something else (**) ). Since you don't have graphical hardware, if you do such an install in the factory default configuration, the installer and the OS will use the first serial port for the console. So you need to use the same method as option (1) above. In fact, as soon as the partition OS starts, it kicks the hypervisor off the first serial port, so that that serial port can be used for the OS console.

AIX can be installed completely using this serial console. For Linux on Power, it can be done but is quite awkward. The easiest solution is to append the "vnc" parameter to the boot prompt of the Linux installer. This starts a VNC server as soon as your network has been configured, so you can continue the install via VNC. And once you have installed AIX or Linux on Power, the serial port will still host the console.

Another option is to install a "Virtual I/O Server" on this factory default configuration. If the VIO server detects this and does not detect an HMC, it will run a dumbed down version of the HMC code on the VIO server itself. You can then https to the VIO server and create/delete/start/stop partitions from there. This solution is called IVM - Integrated Virtualization Manager. Even on a 710 it's perfectly possible to run a dozen or so partitions simultaneously, provided you have enough memory and disk space for them.

(*) At least not on a Power7 server. On a Power8 you can upload other firmware which essentially reduces the hypervisor to next to nothing, and thereby enabling KVM on Linux on Power, AKA PowerKVM.

(**) PowerLinux is the IBM marketing term for Power hardware that is sold exclusively for Linux workloads. They have their firmware modified so that you cannot run AIX or IBM i, and are sold at a price point that makes the TCO comparable to an Intel/VMWare/Linux solution.

Originally posted by: Edward_R86

Ah, very interesting! Yeah, that's starting to make much more sense. My next step I suppose is to figure how best to utilize these virtualization capabilities as it was my main goal for this machine. Yeah, I was looking into IVM and the VIOS and got a bit disconcerted when I read that it's very limited in it's capabilities if it hasn't been activated with a VET code. Something about being stuck with only 3 VMs at 4GB of RAM per VM. Seems like a waste of hardware if the machine has very high specs.

Someone mentioned WPARs for AIX which might help serve as VMs yet again I don't know how far I can go with those or if it's even feasible. Or maybe install more OSes across individual LPARs? Not even sure if that's possible for them to all be running at once so I can manage each individual one.

Thanks for the input! Definitely helped.

Originally posted by: Wouter Liefting

WPARs is a totally different technology. Very briefly, what it comes down to is that you fence off a series of processes and filesystems within AIX. This fenced off area then has its own /etc filesystem with its own root password, IP address, services and whatnot. It can be compared to "zones" or "containers" in other UNIX OSs. But all WPARs share the same kernel and a few other resources.

WPARs are not dependent on the underlying hardware platform or hardware/PowerVM license. Install AIX 6 or 7 on any supported platform (Power4 and up) and you can create as many WPARs as you like until you run out of resources. It's also not dependent on specific licenses: Basic WPARs are included in the base AIX license. (Versioned WPARs do require a separate license though.)

IVM is indeed limited. But that's not a problem, as IVM is intended for low-end Power hardware where those limitations are hardly relevant. However, in those environments the cost of an HMC would be prohibitive. I use IVM on an old Power5 p520 with no issues whatsoever.

As far as the VET codes is concerned: Have you checked that your machine doesn't already have an upgrade applied against it? In any case, even with PowerVM Express edition you can install a VIO server w/IVM and play with a few LPARs, to see if you like it.

Originally posted by: Edward_R86

Yeah, most definitely doesn't have a VET code or anything. After running into so many hurdles about what to do with this machine as far as visualization goes I think what I'm looking to get out of it is beyond the scope of it's intended use-case. I'm basically looking for a solution which can create many VMs with minimal limitations per guest much like ESXi does which we have however that I believe is only supported on x86 based systems. Mostly likely I will have to go with something else as I have no idea what else I could do with it. 

Anyway, thanks for the help!