Nicholas Carr's book The Big Switch: Rewiring the World, from Edison to Google predicts a transition to the delivery of IT as a utility service that parallels the revolutionary effect on industry of the availability of electricity as a source of power nearly a century ago. He works on the presumption that service providers will fill their data centres full of converged IT systems: integrated stacks of virtualised servers, storage and network hardware layers. The question posed by that prediction is, Who will deliver those stacks?
Presumably, they will be bought as single line items, SKUs in the trade, but who will own the manufacturing capacity that brings them to market? Will it be the system software supplier, the server supplier, the storage or the network supplier?
This question was prompted by news that the VMware, Cisco, EMC (VCE) coalition now has 62 customer accounts signed up for its Vblocks. Meanwhile, NetApp has launched its FlexPods in partnership with Cisco and VMware -- yes, the same partners to EMC in the VCE. And Microsoft has announced a Vblock-like Hyper-V Cloud Fast Track with support from Hitachi Data Systems. The idea is to combine Hitachi compute and storage hardware with network infrastructure and Windows Server 2008 R2 Hyper-V and System Center software.
Four distinct tiers of these integrated stacks are appearing. The first tier is the complete in-house stack, with everything supplied and sold by one vendor. HP's BladeSystem Matrix and Oracle's Exadata and Exalogic are the main examples of this tier.
Next is the single product produced by a group of suppliers and implemented as a single entity. The VCE Vblock is the leading example of this kind of integrated stack product.
Next we have NetApp's FlexPod. It is seen as NetApp's FlexPod even though it is a partnership product that uses VMware with Cisco's UCS servers and Nexus Ethernet switches and will be sold by channel partners and installed by them. There isn't a single entity that delivers FlexPods to customers though; the partners order the bits and assemble them themselves.
The fourth tier is a long way from where we started as it is just a set of templates that Microsoft channel partners can use to assemble a valid configuration to run Hyper-V. Other suppliers supporting Microsoft in this effort are Dell, Fujitsu, HP, IBM and NEC. On the face of it Microsoft owns this stack, but Microsoft doesn't sell servers, networking and storage; its partners do.
In the early days of the automobile industry you could buy a chassis and engine from a manufacturer and then ship it to a coachbuilder to have the body put on it. That almost never happens today. You buy the complete, integrated automobile from BMW, Ford or whoever, and that surely is what will happen with converged IT systems for data centres.
The stack will be owned by the supplier who integrates the bits and sells them as a single, orderable entity. HP and Oracle are closest to this model today.
At first, the stack will be owned by the business that puts the pieces together, meaning channel partners in the Microsoft and NetApp FlexPod cases. But, the integration point will be compelled to move back up the supply chain to gain economies of scale. It would be ludicrous to have 1,000 NetApp, Cisco and VMware partners assembling FlexPods from components delivered to them by Cisco, NetApp and VMware. Far more sensible would be a scheme of assembly plants per region, with rationally sequenced production systems. This is Manufacturing 101.
It would also be nonsensical for a software supplier such as VMware or Microsoft to own these stacks and stack assembly plants. Software is just a component; the stack plants will need to be owned by a hardware provider.
As things stand, Oracle and HP will be able to make their converged IT systems more effectively, meaning faster and at less cost, than NetApp FlexPods can be made. So it is surely inevitable that for FlexPods to survive, a FlexPod manufacturing site will need to be established. Logically, as hardware suppliers in the FlexPod ménage a trois, either Cisco or NetApp has to own the plant.
But, as Cisco is also partner to VMware and EMC in the VCE threesome and has invested in it, there is no practical way Cisco can own the FlexPod manufacturing plant. Either Cisco and NetApp invest in a third party to do this, as VCE created Acadia, or NetApp has to own it.
If NetApp doesn't own it and a third party doesn't own it, then FlexPods will die. Just as it's pointless to bet against Ethernet economics, it's pointless to fight the laws of manufacturing economics when thousands of integrated stack boxes need to be built. If NetApp is going to bet against such certainties, all it is doing is handing the keys of the integrated stack kingdom to HP, Oracle and VCE.
As with car and aeroplane plants, there has to be one controlling organisation with responsibility for all major aspects of the product. Given this view, VCE will have to get more control over the design and development of the Vblock components. Unless it does this or unless either EMC or Cisco does this, they are handing the keys of the integrated IT stack kingdom to HP and Oracle.
We are looking at a time scale of years here: five or 10 or more. But the strategists at all these companies need to realise the inexorable logic of what's going on. Compared with Oracle and HP, the others are a long way behind. The trailing pack is led by VCE, with NetApp a way behind that and Microsoft miles behind that. Server system suppliers supporting Microsoft's Hyper-V Fast Track need to get real; this is not the way forward. Their way forward is to own their own integrated stacks, brand them and be capable of building them in large numbers. Hyper-V and ESX are just software icing on the hardware cake -- as essential as the icing, but no way are they the cake.
It's the server system suppliers who should own the integrated IT stacks, and they need to step up before they get swept aside.
The big auto firms that implemented mass-production methods to build cars made every other type of automobile manufacture irrelevant. They bought the ones with any value and let the others fail.
How do you think General Motors came into being? Standalone suppliers of major integrated components for one style of integrated IT stack cannot survive on their own. They either become suppliers to multiple integrated stack vendors or join with one of them. Read up on the history of the Fisher Brothers coachbuilding operation in the US; it ended up becoming a GM operating division. QED.
About the author
Chris Mellor is storage editor with The Register.
This was first published in December 2010