This Wednesday I heard Simon Wardley give a fascinating overview of cloud computing to the Records Management Update 2010 Conference organised by UNICOM. Simon is responsible for Cloud Computing Strategy for Canonical.
Simon told us that cloud computing is difficult to define because it is not a thing, it is a transition.
The transition is from:
- computing-as-a-product (the enterprise buys in hardware, operating systems and applications as products which they install, configure, deploy and maintain themselves)
- computing as a service (the enterprise subscribes to applications which are hosted in the cloud, deployed on operating systems that are hosted in the cloud, sitting on hardware that exists somewhere on the cloud)
The transition period will be problematic and disruptive for companies. They won’t know who to trust, or what standards they should be insisting on, or which suppliers will be viable over time, or how they should govern their information in the cloud. There will exist many different cloud models during the transition period. Enterprises won’t just flick a switch and move all their IT infrastructure, all their development and all their applications straight over to the cloud.
Cloud computing is not a new phenomenon. John McCarthy and Douglas Parkhill both (separately) came up with the concept of utility computing in the 1960s. The necessary technology (virtualisation, which allows a provider to run several different virtual servers on one peice of hardware) was described by Jerry Popek et al in the 1970s. But the crucial attitude change came more recently, with the writings of Paul Strassman and Nicholas Carr who argue that the majority of IT activities in an enterprise do not give the organisation any strategic advantage. In other words IT has become a commodity, well enough defined that no enterprise gets a strategic advantage from doing it better or differently than their competitors.
Simon told us that some organisations will lose out in the transition. One of the key risks is supplier lock in. To avoid supplier lock in you need to be able to get your data out of your existing supplier. You also need alternative suppliers to be both willing and able to accept your data in the form that you extracted it from the previous supplier. The cloud has not yet reached that level of maturity, it is still in a transition period and there is no agreed de facto standard for APIs (the application programming interfaces through which organisations access, amend and add to the data held by their cloud provider). Most cloud providers have set up their own APIs to their services. Simon showed us a long list of different cloud APIs. If you choose a supplier whose API is not compatible with any other vendor then changing supplier would be hampered by the cost of adapting their data so that it can be deposited through the different API.
Simon showed us the complexity of the cloud computing stack. with its three levels of infrastructure, platform and applications.
- Software-as-a-service vendors provide applications – tools such as Google Apps, SharePoint Online and Salesforce that end users can utilise
- Platform-as-a-service vendors provide programming language support and development tools to enable an enterprise to develop new applications and to deploy those applications [see this information week article for a survey of the platform-as-a-service market]
- Infrastructure-as-a-service vendors provide the’ bare metal’ hardware, plus an operating system and a virtualisation tool
Simopn anticipates that there will be cloud computing disasters and failures.
Cloud computing creates dependencies up and down this stack. For example you may trust your software-as-a-service provider to provide you with a stable, reliable application. They in turn may trust their platform-as-a-service provider. But if the infrastructure-as-a-Service provider at the bottom of the stack fails, this would have a knock-on effect up the stack causing both the platform-as-a-service and the software-as-a-service to fail. Simon drew parallels with the the catastrophic banking collapse of late 2008, which resulted from bankers being shielded from the true risks of their investment through the existence of different layers separating their investments from the properties on which those investments were secured.
However Simon does not anticipate that the risks, disruption and losses involved in the transition to cloud computing will prevent that transition happening. Organisations that continue to host their own IT may find themselves at the wrong end of a competitive gap between them and organisations that use cheaper, more efficient cloud computing services. This may mean that many organisations will have eventually no choice but to move to a cloud computing model, in some shape or form, for most non-strategic computing provision.