Original URL: https://www.theregister.com/2010/10/06/hp_integration/

Close integration cannot stop at the HW/SW boundary

Even if openness suffers

By Chris Mellor

Posted in Storage, 6th October 2010 11:59 GMT

Why doesn't HP's Converged Infrastructure (CI) include system software? HP says it does: according to the IT giant you can have any hypervisor you want, any operating system you want, any database or any other middleware you want. So what am I talking about?

HP's CI could stand for Close Integration if we are talking about hardware. The servers, the storage and the networking hardware is all HP technology: ProLiant servers, ProCurve networking and StorageWorks storage - BladeSystem Matrix is a wonderfully well-integrated piece of hardware. Why integrate it so closely? Because that way HP can ensure the parts work very well indeed together. Data and control bits travel from storage to server to network as efficiently as possible because HP owns all the IP and can knit it all together to make it work as fast and as smoothly as possible.

So far so good, so very good, but as soon as we cross the hardware/software boundary, the situation changes. Open house rules, literally, because HP prides itself on being open, on being agnostic to flavours of hypervisor for example. Steve Dietch, HP's cloud infrastructure marketing VP, says that while Oracle's Exadata and Exalogic machines are integrated stacks of hardware and software, they are not open, so you have to use Oracle software.

But this argument applies in reverse at the hardware level, with BladeSystem Matrix you have to use HP hardware, to get the advantages of integration, just as with Oracle's Exadata and Exalogic systems you have to use Oracle hardware. If openness is such a virtue in the software layer then it should be an equal virtue in the hardware layers.

Equally the benefits of close integration apply to software as well as to hardware and, I might add, particularly to the integration of hardware and software.

What will HP's stance be if the Oracle integration of hardware and software pays off in appreciably better application performance, and, say, the Oracle database runs 50 per cent faster on Exadata than it does on equivalent BladeSystem Matrix hardware?

Dietch did not want to speculate about what HP might do in that case. He admitted that the scenario contained a valid point and went so far as to say: "The innovation is not standing still; we know what we have to do."

We might note here that HP has just recruited the ex-CEO of SAP, Leo Apotheker, to be its CEO, and Ray Lane, an ex-president of Oracle when it was purely a software company, to be its chairman. It's a safe bet that HP is cognisant of the strong competition coming from Oracle, and we can read the "we know what we have to do" statement as indicating that HP has to integrate its BladeSystem Matrix hardware stack with system software and middleware.

For example, SAP will most probably run more effectively on BladeSystem Matrix hardware if it is tuned to do so and if the underlying operating system and hypervisor layers are also tuned. The whole point of an IT stack is to run an application as fast and as effectively and as reliably as possible. The application is at the tip of the IT pyramid and, without the application, there is no need for the IT pyramid at all.

It seems to me that, at some stage, HP will have to set about integrating BladeSystem Matrix with specific hypervisor-operating system-middleware software combinations. There is the existing development relationship with Microsoft and that might indicate that we'll see Windows and Hyper-V, plus SAP say, integrated before, say Windows and VMware or a flavour of Unix and VMware plus SAP. But market forces will probably influence HP's choices in the matter.

If the convergence and integration of the IT infrastructure is as strong a driving force as HP believes - HP executives at a briefing in Barcelona today used he terms "paradigm shift" and "secular shift" - then that means integration of the IT stack cannot and will not stop at some arbitrary level, such as the hardware/software boundary.

HP, I'm convinced, knows and believes this, but it has its existing product set to market and support, so the "openness" tactic will be well-used as long as it is expedient. ®