Whether done with the support of an outsourcing partner or done in house with your own staff, most IT in most enterprises is still done locally.
Cloud vendors (the -as-a-service types) follow the logic of Nicholas Carr (The Big Switch) in saying that, just as enterprises had Chief Electrical Officers and small power plants a century ago but moved to the grid to power their buildings, so, too, IT will move to standardised offerings in the cloud.
Will we use some of that? Absolutely.
Will we be fanatical about it? Probably not.
There are a few good reasons why it won’t all be standardised in the cloud.
First, differentiation. For most businesses, some information-based differentiation is a part of their competitive structure.
That may run on a cloud-based platform, but it’ll be custom designed. That’s the nature of differentiation: it’s breaking ground, isn’t standard.
It’s not at all the same as running a standard package and simply buying the service.
Second, getting from there to here. Taken a look recently at the number of enterprises still running vintage 1990s client-server applications? (There are still over 6,000 installations running vintage 1970s mainframe code.)
In all of these cases, this stuff persists because it works, and because it’s highly customised. Packages were deeply modified to “fit us precisely” (which is why there’s no business case to rip those mods out, or to upgrade: the costs outweigh the benefits). Or the package does a unique task (which can be a commodity offering, but not one common enough for a package market to have developed for it, the reason so much mainframe work continues).
You can’t just throw that stuff overboard and run to the cloud, much though it may be nice to dream about it. There’s a real commitment to follow a change track — time, money, other things that won’t be done — to get to the point where a migration could occur.
In tougher economic straits, that’s unlikely to happen.
Third, integration. Cloud vendors offer specific services. Salesforce.com does a fine job of sales force automation and customer relationship management; it doesn’t even try to do continuous flow materials management, or payroll.
Despite the many attempts to be a Swiss Army knife that can handle any and every need, there are very few companies able to run on just one piece of software.
So integration issues — keeping master data in synch, passing results from service A to service B, handling business intelligence needs, etc. — have to be worked out.
Could they be, in an all-cloud world? Certainly. Where a technology runs makes no difference.
But moving data through networks is not free. Moving data around a data centre is (once the infrastructure is installed). Until the charges for transferring data fall below what you can do it for yourself, workloads will remain mixed.
What this means is that the challenge is neither “get to the cloud” nor “hold it off”. It’s “optimise the financial profile”.