Wednesday, June 02, 2010

A life less cloudy

For 18 months I ran the cloud strategy at Canonical. It was a real pleasure to work with some very amazing people and play a small part in Canonical's journey. Today, Ubuntu has become the dominant cloud operating system and leads the field in provision of hybrid clouds combining both public use on Amazon EC2 with private cloud technology.

However, I've been in the cloud space for five years, I've got my long service medal and it's time to scratch another itch but I thought I'd wrap up with some general thoughts.

Definition of Cloud

Given that 200 years after the industrial revolution we still can't agree on a good enough definition of that, it was never likely the industry was going to agree on cloud. The problem with cloud is it's not a thing but a transition of certain IT activities from a product to a more service based economy provided through large scale utilities. This change was predicted by Douglas Parkhill in his 1966 in his book the "Challenge of compute utilities". He made the comparison to the electricity industry and talked of a future of public, private, community and government utilities providing computer resource online, on demand, with elastic supply and charged on a utility basis. These characteristics permeate cloud today but just don't try and get everyone to agree.

Why now?

A number of factors were needed before this change could occur. Firstly, you needed the concept which Parkhill kindly provided us with. Secondly, you needed the technology to achieve this service provision but we've had that for the best part of a decade. Thirdly, you needed those IT activities to be well defined and ubiquitous enough to be suitable for the volume operations needed to support a utility business. This has also happened over the last decade.

Lastly, you needed a change in attitude and a willingness of business to adopt these new models of provision. This change in business attitude started with Strassman in the 1990s when he pointed out that spending on IT amounted to little more than an arms race with dubious links to any value created. Nick Carr developed these concepts further and this wisdom has now peculated throughout the business world.

Is everything suitable for the cloud?

The answer to this is "depends". To understand why we must first appreciate that IT consists of many ubiquitous and well defined activities that are suitable for cloud provision. However, IT also contains many new activities that are not ubiquitous nor well defined and hence aren't suitable for provision as cloud services. This of course doesn't mean that such novel activities can't be built on cloud services.

There is a difference between that which can be provided as a cloud service and that which can be built upon cloud services. However, don't get comfortable yet because activities have a lifecycle. Business activities themselves undergo a constant transition from innovation to commodity. In short that which can be built upon cloud services will eventually become that which is provided as cloud services.

Benefits & Risks.

The benefits and risks of cloud have been listed countless times before so I'll avoid going through these again and just point out some highlights.

First, cloud is fundamentally about enabling and increasing rates of user innovation (a combination of creative destruction and componentisation). Don't get bogged down into the cost saving arguments because whilst cloud will increase efficiency, a number of factors will explode consumption.

Second, you don't have choice over cloud. Your company is competing in an ecosystem with others who will adopt cloud computing. The benefits of cloud will create pressure for you to adopt in order to remain competitive (this is the "Red Queen Hypothesis")

Third, you do have a choice over how to implement cloud. The hybrid model of public & private service is a standard supply chain trick of balancing benefits vs risks. Expect lots of this and ignore anyone who tells you that private cloud isn't cloud computing.

Fourth, it's not about virtual data centres. This one is a tough nut to crack since lots of people have spent vast sums on virtual data centres and they want them to be infrastructure clouds. The problem is simply one of economics. VDCs are based upon concepts of resilience which have their origin in physical product world and come with a wide range of overheads. Infrastructure clouds are based upon the economics of volume operations with vast numbers of cheap & reliable enough VMs and resilience created through design for failure at the management layer.

Fifth, providers ain't your buddies. In a commodity world of utility services then the market, the exchanges, brokers, standardisation and assurance bodies (providing price vs QoS comparisons) are your buddies. Whilst some providers get this, others will try and convince you of their innovative cloud features. These are the enemies of portability & easy switching and the fastest way to end up paying more than you need.

Six, security, auditing and governance have to change in the cloud world and it's not going to be the same as the perimiterised, defense in depth approach of old.

There's a whole host of topics to talk about from the future formation of exchanges, the creation of rating agencies, the introduction of insurance and derivative instruments, the componentisation effects of the platform space, the growth in importance of orchestration, the …. but this has been done to death for the last five years.

I'll mention one last thing. The secret to Ubuntu's success in the cloud is execution.

Rather than fighting battles which couldn't be won, Canonical choose to adopt the dominant approach in the market, provide users with choice and real technology and then combine this with a clear story. Hence Canonical launched a hybrid cloud strategy making it easy for users to build with Ubuntu in the public cloud computing space combined with simple and easily installed technology for creating private clouds that matched the same APIs.

By combining these approaches, management tools could be used across both public and private clouds with machine images that could run in both environments. Users had their first real opportunity to exploit a hybrid cloud environment, choosing for themselves the right balance of public and private with an open source technology that limited lock-in and took advantage of both the Ubuntu and the Amazon EC2 ecosystems. Since it's launch in April '09. this approach has gone from strength to strength and has created a beach-head for Canonical to extend into the orchestration, platform and marketplace fields.

However, the strategy wasn't new, what was critical was the way in which Canonical's excellent engineers achieved this.

Despite the various prognostications on the disruptive nature of cloud, it's old hat. The strategies around cloud were formulated many many years ago. The concepts and ideas around cloud, orchestration, platforms, the economics, the barriers to adoption and how to play in this game are all well known.

If you're looking for that great strategy to bring your company into the cloud, be aware that whatever you think of is almost certainly being done by others and has been contemplated to the nth degree by many more. Today, success in cloud is all about partnerships, execution, tactical moves and timing in a great game of shogi that is already well advanced.

The time for strategists in the cloud space is over and it's all about how the game is played.

Don't get me wrong, the game is still an exciting place. There are all manner of tactical plays, twist and turns and efforts to help standardise and form this future industry. However, strategy is my passion, it's where my heart lives.

Which is why I have joined CSC's Leading Edge Forum to pursue a new challenge at the very centre of change.