Having your cake and eating it

Everyone knows that mainframes are the best computers you can have. You can run them locally, you can hide them in the cloud, and you can link them together into a massive processing network. But we also know that there are smaller platforms out there that work differently. Wouldn’t it be brilliant if you could run them all from one place?

Last summer we were excited by the announcement from IBM of its new zBC12 mainframe computer. The zBC12 followed the previous year’s announcement of the zEC12 (Enterprise Class), and 2011 saw the z114, with 2010 giving us the z196. So what’s special about those mainframes?

Well, in addition to IFL, zIIP, and zAAP specialty processors, and massive amounts of processing power, they came with the IBM zEnterprise BladeCenter Extension (zBX), which lets users combine workloads designed for mainframes with those for POWER7 and x86 chips, like Microsoft Windows Server. So let’s unpick this a little.

One issue many companies have after years of mergers and acquisitions is a mixed bag of operating systems and platforms. They could well have server rooms across the world and not really know what was running on those servers.

IBM’s first solution to this problem was Linux on System z. Basically, a company could take hundreds of Linux servers and consolidate them onto a single mainframe. They would save on energy to drive the servers and cool them, and they would get control of their IT back.

IBM’s second solution, as we’ve just described, was to incorporate the hardware for Linux and Windows servers in its mainframe boxes. You’d just plug in the blades that you needed and you had full control over your Windows servers again (plus all the benefits of having a mainframe).

But what about if you could actually run Windows on your mainframe? That was the dream of some of the people at Mantissa Corporation. They did a technology demo at SHARE in August 2009. According to Mantissa’s Jim Porell, Network World read their abstract and incorrectly assumed that they were announcing a product – which they weren’t. The code is still in beta. But think about what it could mean: running all your Windows servers on a mainframe. That is quite a concept.

Again, according to Jim, they can now have real operating systems running under their z86VM, although, so far, they are the free versions of Linux. Their next step will be to test it with industrial strength Linux distros such as Red Hat and Suse. And then, they will need to get Windows running. And then they’ll have a product.

Speaking frankly, Jim said that currently they have a bug in their Windows ‘BIOS’ processing in the area of plug-and-play hardware support. Their thinking is that it’s a mistake in their interpretation of the hardware commands and, naturally, they’re working to resolve it.

The truth is that it’s still early days for the project, and while running Linux is pretty good, we can already do that on a mainframe (although you might quibble at the price tag for doing so). But once the Mantissa technical people have cracked the problems with Windows, it will be a product well worth taking a look at, I’m sure. But they’re not there yet, and they’re keen to manage expectations appropriately.

Jim goes on to say that once the problems are solved it’s going to be about performance. Performance is measured in a couple of ways: benchmarks of primitives, benchmarks of large-scale for capacity planning, end user experiences, and what they are willing to tolerate. So it seems that they have business objectives around performance where they could be successful if they supported only 10 PCs, then more successful with a 100 PCs, and even have even greater success if they can support a 1000 PCs.

Jim Porell describes z86VM as really just an enabler to solving a wide range of ‘customer problems’ by enabling direct access between the traditional mainframe and the PC operating systems that are co-located with it.

I think that anything that gets more operating systems running on mainframe hardware has got to be good. And I’m prepared to wait for however long it takes Mantissa to get Windows supported on a mainframe. I’ll definitely feel then that I’m having my cake and eating it!

IBM's big baby

IBM has traditionally announced powerful mainframes in alternative years to more affordable models. Last year saw the z114, so this year we were expecting IBM to deliver something big – and that’s what it’s done with the zEnterprise EC (Enterprise Class) 12.

The zEC12 has a hexa-core design that runs at an eye-watering 5.5GHz and is pretty unique in supporting transactional execution. That means it treats system resources in much the same way as a transactional database system and so eliminates the overhead of software locking systems. You can expect to hear more about this Transaction Execution Facility. The z12 chip is implemented in a 32 nanometer high-K metal gate process – its predecessor (z11) used a 45 nanometer process.

The top end machines can take 120 cores, and the processors include an Enhanced DAT2 facility, which allows languages to exploit 2GB page frames. The processors also have a decimal-floating-point zoned conversion facility that will be supported by the next release of PL/I.

The zEC12 also supports Flash Express memory, which is used by systems that anticipate bursts of activity. And IBM claims this is one of the most secure enterprise systems ever with its Common Criteria Evaluation Assurance Level 5+ security classification. It even has a tamper-resistant cryptographic co-processor, which is called the Crypto Express 4s.

IBM said the new mainframe was the product of over $1bn in research and development investment and work was carried out in 18 IBM labs around the world and with IBM’s clients.

And as well as being the mainframe for the cloud computing generation, it also provides hybrid computing support. So not only can it run z/OS, users can consolidate their Linux workloads with Linux on System z (using SUSE Enterprise Server or Red Hat Enterprise Linux), and run other z/ operating systems – z/VM, z/VSE, and z/TPF. And the IBM zEnterprise BladeCenter Extension (zBX) lets users combine workloads designed for mainframes with those for POWER7 and x86 chips, Microsoft Windows Server.

The zEC12 increases the performance of analytic workloads by 30 percent compared to its predecessor. It seems that support for the DB2 Analytics Accelerator enables clients to run complex business analytics and operational analytics on the same platform. There are also IT systems analytics capabilities. It analyses internal system messages to provide a near real-time view of the system’s health, including any potential problems. Called IBM zAware, the technology learns from the messages to recognize patterns and quickly pinpoint any deviations, using the information to identify unusual system behaviour and minimize its impact.

Unusually, sites will be able to run the zEC12 without a raised data centre floor using new overhead power and cabling support. And for ‘green’ sites, zEC12 can deliver 25 percent more performance and 50 percent more capacity with the same energy footprint as its predecessor.

In a statement of direction, IBM said the zEC12 will be the last high-end System z server to offer support for zAAP specialty engine processors. Although IBM will continue to support running zAAP workloads on zIIP processors (zAAP on zIIP). This is intended to help simplify capacity planning and performance management, while still supporting all the currently eligible workloads.

The new processors should be available from about the end of September. It looks like a very powerful piece of kit with a number of interesting features that should keep mainframe users happy for a year or two.