IBM Moves HMC Management To Native Power Server, LPARs
August 14, 2017 Timothy Prickett Morgan
Out of band management is not a new thing in the IT sector, and many of the best and most sophisticated pieces of software in the world have a distinct management console of some sort that gathers up the state of a machine or collection of machines and uses it to initially configure those devices and to coerce them to behave themselves despite their nature for electronic mischief.
The Hardware Management Console, or HMC, has been around for so long in the AS/400, iSeries, System i, and IBM i line for so long we can’t remember when people were not complaining about it. (Or its partner in crime, the Virtual I/O Server, which as the name suggests, provides a layer of system software to abstract and virtualize access to peripheral devices on Power-based systems.) Some have called it the Hardware Mangler Console, and others have called it some things that we can’t print here. Everyone has called the HMC too damned expensive, and they were always right about that. But it is the nature of providers of any kind of system or framework to charge big bucks for the management console, and all things considered, given the integrated form of the IBM i platform, paying two or three times the value of a server that is equipped with the HMC firmware so it can manage one or more IBM i boxes is not as bad as it seems. vCenter Server is no walk in the park, either, and it sure ain’t cheap.
One gripe that companies had about the HMC is that its functionality should have been included with the OS/400 and IBM i stack itself, especially since the whole point of IBM i is that it is an integrated platform, with all of the bells and whistles included and merged together in an easy-to-use, easier-to-buy fashion. We have also complained that in a world where IBM is trying to promote the Power architecture, it seemed counter-productive as well as counter-intuitive to have the HMC management stack for Power iron running on X86 servers. This is particularly true given that IBM has had single-socket Power iron capable of running the HMC stack for a long time, and even more odd given that OS/400 had logical partitioning back in 1998 and Linux was supported on Power logical partitions, or LPARs, since 2000.
There are a lot of HMCs out there in the world, and IBM’s current support page for the HMC is still active with Version 3 of the stack, which was created to configure and control Power4-based systems way back in 2001. HCM Version 8 is able to manage machines using Power6, Power6+, Power7, Power7+, and Power8 processors and is the current version. In October 2015, in announcement letter 215-390, IBM offered the HMC stack in a virtual appliance format that ran atop either Red Hat’s KVM hypervisor or VMware’s ESXi hypervisor and that could be deployed on any X86 server that customers chose. The server had to have an X86 processor that support Intel VT-x or AMD-V hardware virtualization extensions and required at least four cores, 8 GB of memory, and 160 GB of disk capacity to run. This virtual HMC has a license charge of $3,000, which included one year of Software Maintenance (SWMA), and SWMA cost another $1,200 for three years beyond that. The physical HMCs cost at least twice that.
This was a step in the right direction, of course. But there was still that pesky X86 server in the mix. Now, customers are going to have some Power iron choices for the HMC. Starting September 15, customers will be able to run the HMC stack in a logical partition running Linux on their IBM i systems, or they will be able to but a Power server running Linux to host the HMC stack. No more X86 iron, virtual or physical, in the mix.
For those who like to have a distinct management console and a single point of management for multiple machines, in announcement letter 117-033, IBM is offering a Power-based HMC, which is a server that comes in a 1U rack mounted chassis that has a single six-core Power8 processor, 32 GB of main memory, two 2 TB SATA disks, and four 10 Gb/sec Ethernet ports for fanning out to systems under management. This is sold under the product number 7063-CR1, and it requires a variant of Linux that supports little endian data formats and has to run HMC code level 8.7.0 or higher. (FYI: All Power8 chips support little endian data formats; no prior Power chips do.) Customers won’t even know they are not on an X86 variant of the HMC, we suspect. This puppy costs $6,875, but we know from the pricing of the virtual appliance that nearly half of that is coming from the software license fee.
If you want to go virtual and ditch the vHMC on X86 iron at the same time, finally, you can do so as you can see in announcement letter 217-200. This vHMC runs atop the PowerVM hypervisor that also supports IBM i, AIX, and Linux, and that stands to reason since the vHMC software was ported to Linux back in 2015. The Power LPAR version of the vHMC appliance is based on Version 8.8.7 of the HMC code, and it can manage machines using Power6 through Power8 processors.
Customers can mix and match Power HMC and Power vHMC controllers within their datacenter, but we presume only one console can control one machine and its partitions and other functions at a time. Pricing for the Power vHMC was not announced, but presumably this is no more expensive than the X86 version of the vHMC.
So what does it all mean?
There are a bunch of different schools of thought about this. Some of the techies that we talked to says that about a third of the IBM i boxes out there in the world don’t use any virtualization at all, and that Operations Console is good enough for them. As far as they can tell, the HMC is only good if you want to add more than four LPARs on a system and manage their resources; otherwise customers should just use Integrated Virtualization Manager and fire up a few partitions. That said, others pointed out that the HMC does make linking to storage area networks easier and it also interfaces nicely with VIOS and does a good job taming it a bit.
An interesting possible scenario is evolving now, however. With the HMC now native on Power iron in logical partitions, the task of creating a redundant HMC – something that all big shops tend to do – could be a whole lot easier, particularly for those customers using IBM’s PowerHA or some other replication techniques such as MIMIX from Vision Solutions. Customers could fire up a vHMC on an LPAR on their Power system and have a mirrored partition on their backup box, and if they lost their primary machines, they would have it replicated remotely as well as the HMC that runs it. Right now, cross-coupling two physical HMCs is a bit kludgey. Yes, the HMC has built-in replication, but cross-connecting them through ports is a bit tricky, and they have to be individually linked to each Power System server under management. A replicated vHMC would be, in essence, a perfect copy, and would link to each machine under management automagically.
While we are thinking about it, it probably makes sense to have PowerVC cloud infrastructure management (IBM’s implementation of OpenStack) and PowerSC security management tools in the same partition as the vHMC, although we do not know if IBM supports such a configuration. But maybe all of the tools should be integrated and in partitions.
The i in IBM i does stand for integration, after all.