IBM and the "system of systems" zSeries: a new battleship for the data center wars ?

Nothing "official" was said - and we do not analyse rumors - but some remarks collected here and there give the indication that IBM is making a move towards a consolidation of platforms under the zSeries roof.

What some analysts call "z11" or "zNext" was described as a "system of systems" by Rod Adkins IBM STG Manager.

What has IBM done so far ?

Two types of integration already exist and are noteworthy in the zSeries : the "specialty engines" and Linux. They correspond to two different specific needs for zSeries customers :

- Be able to accomodate workloads such as Java or Internet oriented technology within the "z box" but out of control of the zOS system (at least out of its billing scope). In that way, customers continue to use the platform and avoid the temptation to look at the competition (HP or Sun). The value proposition with these ZaaP and ZiiP "specialty engines" is that you are under the same cover for your development lifecycle and production methods but you do not pay the zOS premium. The customer can still benefit from a "marginal cost" for the additional workload.

- Accept some Linux based activity without being forced to buy a new box. According to the slogan "a new processor is better than a new box" a lot of customers decided to add IBM engines (called IFL for Linux) to run Linux and associated applications. The fact that zVM can manage thousands of Linux machines is also an interesting value add.

In these two situations, IBM has strong control over two things : the engine (understand : processor and microcode, the IBM way) and the internal connection within the zBox (keep Cisco out).

If IBM decides to go one step forward, they will most probably keep an eye on these aspects : processor and connection.

What might IBM do next ?

If this box has been called "system of systems" by an IBMer, that means that it brings under the zBox cover other IBM systems. Potential candidates are Power (for the i and pSeries) and x-86 or x-64 platforms. Let's analyse these two options :

- Power under zSeries : from a technical point of view, IBM has control of the processor and the connection. This change is not very difficult for IBM. The Power processor is different from the zProcessor and even if there are some elements in common, the coupling can only be limited ... as is the advantage for customers. For IBM users having many pSeries servers, IBM can -and does- offer a better consolidation under the pSeries itself. The same apply for i-Series and mixture of i and pSeries.

- x-86 or x-64 (or xSeries in IBM jargon) is a different story. The processor in this case is not "home made" and the consolidation may be tedious. Nevertheless, IBM recently announced a new chipset around Intel (or AMD) microprocessors. This new set could have been structured in such a way that it facilitates the interconnection within a zBox. The blade form-factor may be appropriate to accomodate this physical hosting in a slightly revised zBox that would accept chassis frames internally.

It should be noted here that there exist other options that we do not consider as realistic : a pure x-64 architecture emulation on zProcessor or a "specialty processor for x-64", or even a Windows emulation on IBM circuitry... but these solutions are not easy to do and would raise issues of copyrights, patents, licences, etc...not to mention maintenance and support.

The more realistic view is a x-64 integration the way the AS400 offered it for x-86, some decades ago.

What can be the IBM advantage ?

Integrate to simply put things together has no interest. An important aspect, that we want to point out at this stage, is the key role of the "glue" in such an integration. There is a need here for additional functionalities proposed by IBM that would convince the customers that the migration to that box is worth it. We can list :

- better efficiency in power consomption,
- better management of resources (memory, cache, storage),
- improved utilisation ratio,
- more efficient communications inside the box (I/O in memory, networking,...)
- better RAS (reliability, availability, serviceability)
- better elasticity for...cloud computing

This could be managed by a new species of hypervisor or of "control hyperprocessor"... that is traditionally a domain of excellence for IBM .

What is the real target ?

IBM market segmentation was based up to now on a clear separation :

- the zSeries was characterized by a high price, high margin, low quantity strategy,

- the x-Series was low price, low margin, high volume, even if IBM tried to justify a premium on some "nice to have" features.

-the pSeries was in between, depending of the application and performance range.

Why would IBM be interested in opening the golden "high margin" box ?

- Replace the existing xSeries or pSeries by their "equivalents" in the z environment ? This is not really convincing.

- Eliminate the Wintel servers of the competition ? Maybe, but it is a high volume market were IBM is not the best, and the premium of the zSeries world is a handicap.

- Capture large chunks of the x-64 installed base and let it enter a new management scheme ? Sounds better...

We really think that this move makes sense only if the new zBox is aimed at the total data center. The ultimate goal is to shrink the data center into the box. But for the users, might it be a voyage ... of no return ?