The technosaurs continue their battles to become the kings of the cloud, by innovation and aquisition. Oracle just put $1.5bn on the barrel head for RightNow and it doesn't look like being the last software as a service (SaaS) company that they will snap up. The purpose of this spree is fairly clear: if you are going to fill an aquarium you need fish. The Oracle environment will be schools of servers for each customer: not a true multi-tenancy system where developers can dreamily forget about "size" in their deployment - leaving it for human and machine application monitors to expand and contract on demand.
As a reality check let's note that service definition files in Azure dictate size and Amazon machine images are very much sized. SalesForce champions the sizeless but gives clear limit guidelines. It really is all too easy to lose sight of cost of usage even if it is much better than any internal I'll-buy-my-own-silicon-thanks-very-much solution, so all of these system let you size in one way or another because of business reality.
Let's return to the point of development and business making. Part of the cloud vision is that I shouldn't have to worry about silicon or sub-system: buying them, upgrading them or even, within limits, understanding them. In an ideal world where Sir Richard Friend, say, invents a whole new optical computing thingamejig, a single instance of which can run a cloud and additional resource is grown rather than expanded by networked: we cloudizens should be blissfully unaware of anything except lowering hosting costs and increased capacity. It's a fanciful example but the point is that abstraction really does matter.
Oracle's system will not be very abstract by latest reports. So what are the benefits of them not going true multi-tenancy or is it just that they are late to the game and need to cobble Fusion into a fit-for-cloud* technology. There is a big one - the real-walls-a-must argument.
It goes like this: for legal/political/trust reasons a client insists on a physical air-gap between their information or application space and any other clients, there is no room for play and they don't accept that your virtual partitions are inviolable. Oracle gives you a separate pod and polaroids of the arrangement, your compliance officer takes them off to whichever watchdog is currently chewing the remains of his ankle. I also should point out that Oracle provides for the multiple tenancy of pods for smaller customers, a model which should be familiar from days of old server farm hosts. Is the truth that many of these customers will find it easier to keep their solution on premises rather than move to any cloud solution? Quite likely. A lot of these customers won't be shopping in the SaaS/PaaS/IaaS market anyhow.
It's possible that the latest Azure federations may offer an oblique answer to this by forcibly locating a single client/tenants information by a SaaS customer on a machine but it's not a real wall/air gap. Ultimately SalesForce and Azure will/can both provide very convincing answers to real-walls-a-must, an answer that needs to be check boxed and formatted to satisfy non-technical legislature.
Buying 'customers' can only really be a short term solution: designing for those bought companies may turn out to be short-sighted. If you really want the right kind of fish in your aquarium then lower the gate to the ocean and, if your fish tank is that good, they might just swim in by themselves.
*I'd like to patent that phrase, especially in the sense: "before I started architecting for the cloud I could run a marathon, then I got absorbed into the things I could do and before you know it I was fit-for-cloud, my software was incredible but my trousers didn't fit."