Virtualization, The Recession and The Mainframe
What do virtualization, the recession and the mainframe have in common?
- Nothing at all
- Quite a lot
- Everything
Your answer to this question is going to depend very much on what you know about the modern zSeries mainframe. So I’ll assume you don’t know very much and lead you through the forest.
The modern mainframe has its origin in the IBM System/370, a range of machines that superseded the early IBM 360 mainframes and were themselves replaced by the IBM System/390 before it was transformed into the zSeries. There is, naturally, a huge difference between the IBM System/370 as it was and the modern zSeries, but there are also some important similarities. The similarities that matter are:
- The System/370 introduced pretty much all the technical tricks required to allow the architecture to scale over decades, from virtual memory to an increase in address space (initially from 24 to 31 bit) and the ability to use multiple address spaces.
- The operating system was built in highly efficient assembler language.
- The operating system was built to optimize the use of all resources under its control in a highly efficient way.
- A virtual machine capability was introduced on the System/370
We can get very blasé about virtual machines, after all, everyone and their grandma uses virtual machines nowadays. But the point to understand is that a VM running on an IBM mainframe is a different VM from the one running on a Unix server – whether that Unix server comes from IBM, HP, Sun or Johnny-come-lately. That mavericky old IBM mainframe is utterly miserly in the way in manages resources – and not only that, decades of evolution have provided it with a management ecosystem that Unix can only dream about.
IBM has demonstrated over 40,000 VMs with Linux loaded happily running on a mainframe, but that’s the stuff of demos and it’s no more astonishing than seeing a dog walk on its hind legs, because while it’s possible, it’s not useful. The real point is that zSeries machines can achieve resource utilization in the “90% and above” area. As far as that is concerned, all you can say is “eat your heart out VMware.” Under Unix or Windows (with VMware) you’ll never get close to that, not even going down hill with the wind behind you.
Of course, this doesn’t mean that everyone ought to buy zSeries mainframes for running Linux. It’s a little more complicated than that. First, if you don’t already have an existing mainframe and mainframe expertise then that’s a serious barrier to mainframe adoption. But if you do, using the mainframe as a virtualization platform makes a great deal of sense for a whole series of workloads.
What has tended to happen with both Linux and Windows is that companies have deployed single purpose servers (or blades) that run a single application. If the cpu usage of that single application is light, then it is a natural candidate for virtualization – particularly for mainframe virtualization. It is also a natural candidate for Unix-based virtualization on a big Unix box.
However, ROI models I’ve recently examined show a huge cost advantage for the mainframe in most circumstances. It starts to become apparent when you add everything up; the power costs, the floor space, the air-conditioning, the virtualization software, the virtualization management software, the system management software and so on. If there’s server consolidation to be done then the zSeries mainframe can be the most economic option – and in some situations the cost differences are striking.
This type of consolidation does need more publicity, especially the “resource utilization in the “90% and above””. After all, we’re talking about real money (let alone the green side of thing).
One of the challeneges is the initial setup cost. A hosting company, for example, can start off with several large intel servers and start scaling out / up. Starting off with a Z Series machine (or even a P5xx) and the associated infrastructure seems a bit frightening in comparison.
Actually Robin, while it’s true that the S/360 operating systems were written in Assembler, and much of the 370 operating systems, PL/X was already in use for some of the large and complex components.
It is also widely know that virtualization, as you know it on the mainframe today, was first introduced on the S/360 model-67. This was a “bastard child” of the S/360 processors that had virtual memory extensions. At that point, the precursor to VM/370 used on the S/360-67 was CP-67.
I think you’ll also find that IBM never demonstrated 40,000 Linux virtual machines on a single VM system, it was David Boyes of Sine Nomine, who also recently ported Open Solaris to VM. http://www.theregister.co.uk/2008/10/17/solaris_on_mainframe/
Also, there’s no such thing as pSeries Unix in the marketing nomenclature any more, it’s now Power Systems, whose virtualization now supports AIX, System i or IBM i to use the the modern vernacular.
Wikipedia is a pretty decent source for information on mainframe virtualization, right up until VM/XA where there are some things that need correcting, I just have not had the time yet.
Oh yeah, by the way. While 2TB of memory on a mainframe gives pretty impressive virtualization capabilities, my favorite anecdote, and it’s true because I did it, was back in 1983, to virtualize a complete,production, high availability, online credit card authorization system, by adding just 4Mb of memory boosting the total system memory to a whopping 12Mb of memory! Try running any Intel hypervisor or operating system on just 12Mb of memory, a great example of how efficient the mainframe virtualization is!
Many thanks for the corrections. Looks like I should have done a little more fact checking. Never mind, the thrust is right and the details have been corrected….