Monday, 28 April 2008

What is BPEL and should I distinguish between the music and the dancers?

Well, let’s answer the first part of the question posed in the title and see what BPEL is. The letters stand for Business Process Execution Language, and it is a language for specifying the behaviour of business processes based on Web services. Importantly, processes use only Web services to import and export functionality.

That all sounds fairly sensible – it’s a language to specify how business processes should execute, and it uses Web services, which are software systems supporting interoperable machine-to-machine interactions over a network. And, of course, WSDL (Web Services Description Language) is the machine-readable description of the service. And we’re all familiar with SOAP and WSDL etc these days.

But BPEL gets more complicated. These business processes come in two varieties – executable and abstract. So, definition time… An executable business process models the actual behaviour of a participant in a business interaction. An abstract business process may hide some of the concrete operational details – making it more descriptive. This is important for a real-life business model.

IBM and others developed BPEL4WS 1.1. The current standard (with the current naming convention) is WS-BPEL 2.0. In June last year, IBM and others published the BPEL4People and WS-HumanTask specifications, which describe how human interaction in BPEL processes can be implemented.

So what’s all this staff about "the music and the dancers"? Well, BPEL is an orchestration language, not a choreography language. It seems that an orchestration language specifies an execution process involving messages being exchanged with other systems – specifically, the message exchange sequence is controlled by the orchestration designer. A choreography specifies a protocol for peer-to-peer interactions.

Let me put this another way. Orchestration defines the flow of a process that involves distinct services and choreography defines the interaction between independently-defined processes. If you are writing a choreography, then there is WS-CDL – Choreography Description Language.
In truth, people are using orchestrations, but no-one is yet using choreography.

How did they get these names and what analogies were their creators thinking of? Well, if you think about an orchestra, although there are numerous players, they all follow the conductor. So an orchestration is more about centralized control. Now think of a stage full of dancers – they all have their parts of the performance but they do not have a conductor, they are in many ways on their own.

So, as a died-in-the-wool mainframer, why do you care about BPEL and dancers and orchestras? The answer is really all to do with SOA. If making your mainframe part of a distributed system is the way forward, and you don’t want a mixed bag of point solutions that work with a bit of your CICS system or your DB2 database, then you probably are looking for a simple single strategy. You’re also probably looking to future-proof your SOA development by adopting a recognized industry standard.

BPEL might just be able to help you achieve that.

Lastly, if you’re an IMS user, then please complete our IMS survey. You can find it at

Monday, 21 April 2008

2 become 1

Am I talking about the Spice Girls 1996 Christmas mega-hit (the USA had to wait until July 1997, when it reached number 4 in the US charts)? No, obviously, I’m talking about IBM’s recent announcement that System i and System p have officially converged.

System i is still often thought of by its old old name of AS/400, and System p is still commonly called RS/6000. However, in Nashville at its COMMON user group meeting, IBM announced two standard configuration systems. These are the IBM POWER 520 Express and the IBM POWER 550 Express. So what’s the big deal? Well they can run AIX, Linux, or IBM i (formerly-known as i5/OS or OS/400).

Clearly this is another step in IBM’s not-very-secret-at-all Project ECLipz, in which their three different processor lines would eventually merge. This isn’t the first sign, you may recall the Power5 processor could run both System i and System p. The logic behind Project ECLipz is that it makes no sense financially for a single company to develop and maintain three separate CMOS and chipset architectures. In addition, there are the associated three separate hardware and support infrastructures. Why not have just one?

So, apart from coming out with a new name for OS/400 or i/OS (that’s the IBM i tag), what was actually announced? The 520 can ship with up to two 4.2GHz Power6 processors, and customers can activate one, two, or four cores on the systems. The 550 ships with two, four, six, or eight cores, meaning it can have up to four Power6 chips running at 3.5GHz or 4.2GHz. These are both low-end models.

Also announced was an i Edition Express for the BladeCenter S. Customers can slot the JS12 Express blade into the BladeCenter S chassis, allowing Power and x86 systems to be run in the same case.

Forgetting the new name for the System i software part, the real significance of the announcement is that we see another small step on the road towards convergence of IBM’s different systems. Cynics might say it will take a long time before System z, with all its major differences, can merge with the other two. However, it looks like those 2 (Systems i and p) will definitely become 1.

Monday, 14 April 2008

IMS is dead!?!?

I took part in an enterprise survey last week. The first question was to name companies that I thought were in the enterprise IT space. Of course, I trotted out names like IBM, CA, and BMC. When prompted for more I said DataDirect, NEON, Mackinney, William Data Systems, SDS, and Oracle. I would have moved on to high-end hardware suppliers, but I stopped – to see whether I had said enough. The second part of the survey was about Hewlett-Packard. Now if HP thinks it’s in the enterprise space, where does that put IBM and the other mainframe suppliers. Where does that put things like IMS?

Hewlett-Packard has been around for a long time, but I always think of them as the company that sells cheap printers and expensive ink cartridges. In truth, I’m actually writing this blog on an HP Pavilion laptop. In terms of positioning in the market place, I tend to put HP and Sun as mid-range players. I can remember when HP took over DEC, who at one time had been king of the mid-range. In fact, PDP11s were like the Volkswagon Beetle or Morris Minor of their day – they were still in use at many companies long after you would have expected them to have been replaced. I even worked on a DEC VAX machine (in the days when VAX was nothing to do with vacuuming the carpet!). I still remember the excitement in the press when Carly Fiorina became CEO – and when she left. Although I knew about their workstations and their OpenView software, I still thought of them as mid-range.

So if everyone else in the world sees HP as an enterprise player, where does that put IBM? Does this mean that IBM is now really little more than a niche player in the enterprise space? And if that is the case, where does it put specialized packages like IMS? Here’s the logic: only a part of the enterprise space is IBM’s, and only some of those IBM sites use IMS. Therefore the IMS part must be very small and probably shrinking to nothing over the next few years.

If you’re an IMS user, how does that make you feel? Well, I’m hoping that I’m completely wrong with everything I’ve just said about IMS. But I think it’s important that IMS users have a chance to have their say. I think it would be very useful to know what real IMS users are doing with their IMS. So, if you are an IMS users, why don’t you fill in the IMS survey at When the results are in, I’ll let you know whether IMS users think IMS is dead or not.

IMS users who haven’t already signed up may be interested in the Virtual IMS Connection user group at

Monday, 7 April 2008

IBM and extinction-level events

Last week I was suggesting that IBM mainframes, particularly the new z10, were, for all sorts of reasons, the best computing solution in town, and yet many people were making other choices. I was suggesting that if IBM didn’t do something they would find themselves as extinct as the dinosaurs they are so often compared to. And with dinosaurs ruling the Earth for 186 million years, this isn’t necessarily a bad thing!

So, despite IBM having the best product, many people are choosing an alternative option. What can IBM do about it? What are their choices to avoid becoming extinct?

IBM could decide to do nothing. This solution solidifies them in their current position. They can continue producing new hardware at regular intervals. They can continue upgrading their software. And, of course they can continue selling at very high prices to very big and very wealthy corporations.

IBM can continue with their SOA (Service-Oriented Architecture) developments. But while this makes bits of CICS programs available to Web browsers, it does nothing to generate more mainframe sales. It really just satisfies current mainframe users, who can argue against people wanting a change to Unix or Microsoft in order to have this kind of computing available.

The future looks bleak for IBM because they will only be selling to a diminishing number of companies. They don’t currently seem to have a strategy to suck in middle size VMware users from server farms.

What’s needed is a physically small box, perhaps running z/VM (as you know I’m a big VM fan, I’ve even written two books on the subject) that needs almost no tweaking once it’s first set up. This box should then run any operating system you like – Linux, Solaris, and, dare I say it, Windows. And it should run an incredibly large number of them. And it should cost almost nothing (to begin with).

This is a bad strategy because IBM will immediately lose all its small and some middle-sized mainframe users. Even large sites will be working out the cost-benefits.

It’s a good strategy because it will retain all its smaller customers who would have bought other products. It’s good because it will encourage sites using almost any other server to change to IBM. Having killed all the mainframe emulator companies, it will be able to do this.

IBM then has to bite the bullet of small revenues while other competing companies go out of business. Then, in perhaps 10 years time, this new-look IBM starts to raise its costs, or the costs of add-on speciality engines, and gradually goes back into mega-profit.

This mutation of IBM will allow it to continue to thrive long after the asteroid has hit the Earth and lumbering giants have turned into fossils.

Meanwhile in the real world, IMS professionals will be pleased to know that Tuesday 8 April at 10:30 CDT is when the next Virtual IMS Connection ( user group meeting takes place. The meeting is free to members (and signing up to become a member is also free) and contains part 2 of the talk by Bill Keene on IMS disaster recovery preparation.