People underestimate how much cloud providers present a challenge to IT as it's practiced today in most organisations. A couple of news stories brought home this point. They illustrate the existential threat that cloud computing and its practices present to corporate IT groups.
The first item that came across my desk was a blog entry about AOL and how it has built a completely unmanned data centre. The writer noted that he came to AOL relatively recently and was tasked with streamlining its data centres. He says, "It was during this process that I came to realise that our particular legacy challenge, while at 'internet' scale, was more closely related to the challenges of most corporate or government environments than the biggest internet players."
The writer continues to say that, despite its early "King of the Internet" position, AOL's collection of hundreds of bespoke, inconsistently configured and implemented systems made it very similar to most enterprise environments: manually administered, high cost and difficult to change.
The piece then goes on to describe how AOL restructured its entire infrastructure, and more importantly, its supporting tools and processes, to create a lights-out, completely automated deployment environment. No one needs to enter the data centre to provision machines, configure networks, run cables, etc.
The writer describes how a key piece of the automation systems is a CMDB (configuration management database), which allows quick, dynamic provisioning of resources. How quick and dynamic? Virtual machines in eight seconds. Injection of application and middleware software packages in another eight seconds. Reconfiguration to move into production? Sixty seconds.
AOL's story, stripped to its basics, is that of an organisation applying today's cloud-inspired IT management best practices to achieve highly scalable, extremely cost-efficient computing environments.
The next day, I came across another blog post written by an enterprise IT professional. In it, he pooh-poohed the advantage and serviceability of automated IT tools. He posited that, for most organisations, a solution of human computing (meaning smart, experienced, people) is the right way to achieve a consistent, efficient infrastructure.
I often come across people who denigrate cloud computing. They point out systems that require careful hand tuning to achieve acceptable performance.
Database configuration in production ERP environments is commonly cited as an example of why cloud computing automation isn't sufficient. The suggested conclusion is that IT organisations should continue existing practices because they ensure that finicky applications requiring trial-and-error optimisation by experienced professionals can be managed properly.
A third event illustrating the existential threat that cloud computing poses to IT came about as a result of Apple's release of the new iPhone 4S and the accompanying iCloud. Apple includes 5GB of free storage to let customers manage photos, videos, etc. Seeing an opportunity to piggyback on this high profile event, box.net announced that it would make available to any iPhone or iPad user 50GB of free storage.
What struck me about box.net's announcement is the contrast between this distribution of free resources with the typical stringent rationing of computing resources within most companies (which, of course, is driven by the restricted budgets doled out to IT organisations).
Obviously, box.net has made a financial calculation and concluded that, based on its business opportunities and cost structure, it can provide a large amount of IT resource at no cost to users. And, even more obviously, this has now raised the bar for other cloud providers. How long will it be before another provider ups the ante to 100GB?
Using low costs driven by purchasing economies of scale, along with the kind of low operating costs provided by automation techniques such as those described in the piece on AOL, cloud providers can vastly undercut internal IT service offerings and costs.
Of course, one can point out that offerings such as box.net suffer from shortcomings that enterprise IT offerings don't: They fail to provide compliance certainty, in-person support, integration with internal applications, etc. There's no doubt that this is true.
What both the enterprise IT professional and the people who point out shortcomings in popular cloud services fail to comprehend, however, is that the number of applications that require that stringent certainty and that hand-optimisation is a small percentage of overall enterprise applications. Not to mention the fact that there is an enormous universe of applications that, today, are not even considered because they can't possibly justify themselves in the resource-constrained, high-cost, rationed world of enterprise IT.
The problem with arguing that current practices must be continued to address necessary application optimisations or compliance requirements is that it imposes the inefficiency and high cost structure of current practices on all the applications that make up the corporate portfolio-even though most of them don't require those practices. It's like insisting that one's family use a Ferrari to make the weekly grocery run because occasionally it's necessary to turn a 56.12sec lap at Lime Rock Raceway.
The biggest threat to IT organisations faced with these new cloud developments is that they set the new bar of expectations. If AOL can get a server up in eight seconds, and the IT organisation takes two weeks, it's going to be seen as hopelessly ineffective-and ripe for a complete teardown.
Far better would be to evaluate the application portfolio and identify those applications for which the skills of experienced personnel are required to ensure performance, compliance or other characteristics uncongenial to cloud deployment. A case should be made for those applications to be deployed, administered and monitored by the IT organisation. The rest of the application portfolio should be aggressively migrated to cloud providers who can apply their efficiency and cost advantages to them.
The biggest mistake most IT organisations make is failing to recognise the urgency of this application sorting. When an industry adopts a new technology that provides a significant cost advantage, those who tarry in adoption threaten the survivability of the entire company. Viewing cloud computing with skepticism, representing it as another turn in the centralisation versus decentralisation cycle fails to comprehend how big a shift it represents in computing. Cloud computing is no timesharing-with-a-gloss, as many people think of it. Believing that one has plenty of time to maintain current practices fails to draw lessons from previous cost revolutions, and poses the threat that one's company will become the Studebaker of today.