Monday, December 07, 2009

Private or Public clouds?

There is ample evidence to suggest that many common and well defined activities in I.T. are shifting from a product to a service based economy. Naturally this change creates a broad range of risks including :-

  • the risk of doing nothing as competitors gain advantage from economies of scale through volume operations, utility charging, ability to focus on core activities and a faster speed to market through componentisation .
  • transitional risks including confusion, security of supply, trust in new providers, transparency and governance.
  • outsourcing risks including suitability, vendor lock-in, pricing competition, second sourcing options and loss of strategic control.

For any organisation, it is a case of balancing the risk of not using the cloud against the risks of using it. There appears to be two general schools of thought on this subject.

The first school states that whilst the outsourcing risks will be solved by the formation of competitive utility computing markets (with easy switching between providers) these markets do not exist today. Hence, whilst the movement to public clouds is considered inevitable (bar for the largest companies and governments), we're still in a time of transition. Private clouds can help solve many of these transitional risks whilst preparing for a future movement towards public cloud services.

The first school accepts that private clouds have a role, it emphasises the importance of standards, of reducing barriers to education and promotes a hybrid model of both public and private clouds. It encourages a compromise between economies of scale and transitional risks during this time of change.

I've been an advocate of this first school of thought for many years (since before 2006).

The second school of thought states that private clouds aren't cloud computing and advocates adoption of public clouds. It dismisses the transitional phase and talks of continuous innovation in the provision of what is fundamentally a commodity (commonplace, well defined and hence suitable for service provision). It is almost purist by nature, sometimes describing public clouds as true cloud computing and finding little distinction between private clouds and virtualisation platforms.

I don't subscribe to this second school of thought. Basic economic sense and risk management would suggest that in this time of transition, organisations will attempt to gain some of the benefits whilst mitigating the dangers of this phase. Hence, for the next few years I'd expect the cloud industry to be dominated by hybrid models.

After which I'd expect it to become more slanted towards public provision (as competitive markets form) but nevertheless hybrid models will continue to have a role.


Anonymous said...

Simon, great post and you raise a good point. We're one of the companies that really wants to accelerate the adoption of public clouds, mainly because we believe that public clouds are the way enterprises can really benefit from the cloud. Our main beef is with virtualized data centers being called private clouds. The nomenclature makes people believe that their work is done once they move to a "private cloud" but as you rightly point out it's a transition phase.

By the way, heard from folks on the CSC LEF Study tour that you gave a killer presentation on cloud computing. Do you have that in your blog or on YouTube?

swardley said...

Thank you.

Regarding increasing adoption of public clouds, the real issue here is overcoming the concerns such as lack of competitive marketplace, easy switching between providers and second sourcing options.

This is why back in 2006, we planned to open source Zimki (one of the first PaaS before it was called PaaS) in order to encourage formation of a marketplace of providers.

Back in 2003/4, the company that I ran heavily used virtualisation throughout our own data centre (i.e. we virtualised the data centre). However, we didn't produce a "private" cloud environment with more self-service capabilities (our own early equivalent to EC2) until around 2004/5.

I'd suggest that there is more to "private" cloud than just data centre virtualisation. However, even with this "private" clouds are more of a hybrid transition strategy and not necessarily an end goal in themselves.

Regarding my talk, thanks but alas there is no video. I'll have to put something together, however there are many videos of me talking about cloud (nee utility computing) and commoditisation over the last 3 years.

A more recent, light hearted one is my talk at OSCON '09.

It's a variation of the talk I gave at OSCON'07.

Unknown said...

Hi Simon,

I agree with your position.

The short answer: Yes, a private cloud is a viable alternative to a public cloud environment.

Long answer:

A private cloud is a deployment model. It has the same characteristics, by definition, that you'd find with a public cloud. The primary difference is that it is run for the sole use of a given organization and, most often, deployed within the four walls of an organization's data center. 

Yes, there has been a lot of debate about this. People like Werner Vogels of Amazon loathe the idea. And, he has a point. That is, even if you create a private cloud within the enterprise, you still own the hardware and software that make it go. You may even deliver apps or servers on-demand and bill users internally for services consumed, but you're not eliminating Capex. 

Okay, I get that argument. But, that still doesn't mean that you can't deploy a private cloud. For many companies - whether enterprise or not - public clouds are not (yet) ready for prime-time. So, a private cloud deployment makes a lot of sense. And, there are many vendors - from Canonical to VMware - that can help companies leverage this deployment model. Heck, some have even argued that Amazon's own VPC gave more legitimacy to the concept. 

At the end of the day if a private cloud opens the door to more organizations - especially enterprise customers - then it's more than a viable model.