Cloud Computing – A Real World Guide

First session of the Enterprise 2.0 conference 2009 and how could I resist this one… “Cloud Computing, A Real World Guide”.

Alistair Croll, Co-author, “Complete Web Monitoring” and Principal Analyst, Bitcurrent

“Cloud computing is like modern art – I can’t describe it but I know it when I see it”. Apparently Amazon Web Services has over 60000 customers of which the majority are (somewhat surprisingly) enterprise level. An interesting analogy that cloud computing is either like soup or stew – either very configurable or easy to digest but “one size fits all”. Google app engine is soup – you can code in any language so long as it’s Python or Java – but if you fit to their rules – it’s easy, quick and simple. Amazon as an example of stew – more complex but more flexible.

Why to switch to clouds;

• Better economics (but not always – and unhelpful to only seel cloud computing based on cost savings)
• Developer empowerment

Croll gave the example of when The Washington Post got access to Hilary Clinton’s diaries during the election and compared the time and cost of making those diaries searchable on a local processing base versus in the clouds. They used cloud computing to convert more than 17,000 pages into a searchable database within twenty-four hours.

Croll talked around the case of Coghead – the PaaS provider that is now in the deadpool. He pointed out that developing for an agnostic PaaS provider makes no sense – leveraging a user base (a la force.com with salesforce.com, the Intuit Partner Platform with QuickBooks) makes much more sense. He also reminded people that thinking about mobility is important for startups – building an application that is inherently tied to any one particular platform is risky in that it very much limits the possible exit strategies that can be followed.

He told the story of one massive financial services company that is considering moving all of its IT infrastructure over to Google – he wasn’t prepared to divulge details (not surprisingly) but indicated that this enterprise was adamant that Google could do it better than they themselves could with internal infrastructure.

Pick the right battle – the best place for an enterprise to use cloud computing Croll suggests are;

• Places where traditional IT wasn’t used before because of cost, time or process limitations (witness the Clinton diary example above)
• Where lower concerns exist around security, configuration and control

Croll pointed out a CapGemini study that found that “over 80 percent of the information a worker needs to complete their job is held outside of the organisation” and as such we should embrace being more porous with the outside rather than siloing data. He contrasted it to a similar report done 20 years ago where the figure was 20%.

A cautionary tale: the Ada language that NASA used in the eighties to replace some Fortran… While the results that Ada could bring where very attractive, it never achieved critical mass – the naysayers blocked adoption of something new while the evangelists promised too much, too soon.

An amusing metaphor – Croll pointed out that in the early days of electricity those who worked with its generation and distribution where superheroes – now they’ve been relegated to being smelly dirty workers as electricity moves to being a core utility. He contends that the same will occur with those he calls “server huggers” – as the cost of computing trends towards zero, the infrastructure behind that will become a complete commodity utility.

Croll then went on to try and refute Nick Carr’s assertion about cloud computing being a utility item like electricity – he pointed out lack of standards, low levels of interoperability, compliance requirements and the sheer number of services that a business may be able to use as being reasons that cloud computing should be thought of as different. To be honest I think his points were somewhat flawed – much of the issues he brought up are symptoms of an immature market rather than failings with cloud computing per se. In essence he was contrasting the entire cloud stack with only one level of electricity as a utility. A fairer contrast would be contrasting generation only with cloud processing.

Croll spent some time discussing the concerns and unknowns around the use of meta-data. He used the example of the image view data that Flickr has with individual images – possible a better example (given the audience) would have been the use of aggregate data for benchmarking as we’ve discussed recently with regards to mint.com.

And lastly the kitchen-aid analogy – it used to be that motors were expensive and attachments were cheap – hence buying a kitchen-aid and a bunch of attachments that go with it. A similar thing is happening with computing – once processing was expensive but attachments were cheap. Now everything has its own processing as the cost of processing plummets. Now what is important is the synchronisation, it’s an evolution towards ubiquitous computing where everything should be available everywhere – the ultimate move from a device centric world to a data-centric one.