Clouding over the issues.

Forget Grids, and the "network is the computer", and Distributed Computing and all those ideas from yesterday. We now have Cloud computing, which will solve all those tricky problems the last N iterations of networked server computing could not. The zeitgeist is upon us, with posts on my reader from far and wide, and has even made Techmeme via Businessweek:

The term "cloud computing" encompasses many areas of tech, including software as a service, a software distribution method pioneered by Salesforce.com about a decade ago. It also includes newer avenues such as hardware as a service, a way to order storage and server capacity on demand from Amazon and others. What all these cloud computing services have in common, though, is that they're all delivered over the Internet, on demand, from massive data centers.

A Sea Change in Computing

Some analysts say cloud computing represents a sea change in the way computing is done in corporations. Merrill Lynch (MER) estimates that within the next five years, the annual global market for cloud computing will surge to $95 billion. In a May 2008 report, Merrill Lynch estimated that 12% of the worldwide software market would go to the cloud in that period.

A fortnight ago I was on a panel at the Wealth of Networks Conference talking about the evolution of Service Infrastructures (Disclosure - we strategise, design, build and fix these infrastructures for a living), of which Cloud Computing is the current topic du Jour (Service Oriented Architectures being so 2007). I think I was put on as the token sceptic . And I must admit to being a little nonplussed - about once a year we tend to pick up a client assignment predicting the future of the overall service infrastructure market, and what is required for it to progress - and whether it's Top Meme is Clouds, or Grids, or SOA, or PaaS or whatever, its all comprised of the same basic submarkets, comprising of the Digital Logistics - the "trucks n' sheds" of the internet. The datacentres, backbones, OSS/BSS, middleware, and client side software subsectors to make all this happen.

And if I were just a bit skeptical (surely not) I would argue that this is merely an attempt to rebrand this overall area yet again, as it hasn't really taken off the last times. Nothing has really changed conceptually - we have datacentres, we have the internet, we have PC's (and increasingly mobiles etc) on the receiving end. They are all getting cheaper, bigger, faster etc but that is all pretty predictable (in fact, the $100bn market 5 years from now is pretty much a constant part of the story as well - it comes from aggregating all these submarkets up and hitting it with 5 years projected GDP growth globally).

So what has changed this time - what is the real difference between "The Cloud" and what has come before? There are some subtleties, as Wikipedia notes for example:

Cloud computing is often confused with grid computing (a form of distributed computing whereby a "super and virtual computer" is composed of a cluster of networked, loosely-coupled computers, acting in concert to perform very large tasks), utility computing (the packaging of computing resources, such as computation and storage, as a metered service similar to a traditional public utility such as electricity) and autonomic computing (computer systems capable of self-management). Indeed many cloud computing deployments are today powered by grids, have autonomic characteristics and are billed like utilities, but cloud computing is rather a natural next step from the grid-utility model. Some successful cloud architectures have little or no centralised infrastructure or billing systems whatsoever including Peer to peer networks like BitTorrent and Skype and Volunteer computing like SETI.

I await with interest to see how "monetisation" will happen without a billing system (except flogging yourself to eBay )

But technically this is still not much more than the inevitable progress of innovation learning curves, and economically not much more than the inevitable progress of Moore's and Reed's Laws. It is hard to see why this is a step change from what has come before.

Strategically its far more interesting, because what we are seeing is The Convergence, and corporates of various stripes scrabbling for the high ground (and in Dell's case, the Right to Brand). Consider the stripes - from the Web 2.0 corner comes Google, from the Web 1.0 corner comes Amazon, from the Web 0.0 corner comes Microsoft, and from the 4th corner come the resurgent Telco 2.0's who have networks and think they may as well be computers too.

In the centre sit a host of "traditional" ICT incumbents - the hardware and software companies, who are finding these large tanks increasingly parking on their (and each others) lawns.

The prize they all believe is worth winning is that moment of magical Enterprise 2.0 Enlightenment, when large corporates and SME's outsource all that infrastructure to them. Just in time, for today's outsourcers, to avoid the crippling race to the bottom of pricing in that market that has occurred in he last 5 years or so.

And of course they are all hoping that they can get the benefit of network effects - aka the rich get richer - in which early leaders find they get a positive feedback loop going and pull ahead of les autres. Similarly, the risk of not getting in early is being consigned to the long tail of Also-Ran's.

But in this noble rush for the New Outsourcing Eldorado, the cloud enthusiasts are prone to being a bit dewy eyed over its benefits, and try and pull the mist over the eyes of customers as to why - by and large - these efforts have failed in the past. The issue with Cloud computing is that its a new tech fix, but core reasons for failure in the past have had little to do with the technology, and everything to do with business risk:

(i) Risk of service failure - you need high service availability. Students of probability theory will know that to get 99.999% reliability in a one chain link system (from your computer to your server) is "interesting". To get it in a 2 chain link system requires each chain link to to be 99.9995% reliable, which is most kindly described as "challenging". Clouds usually have more links in the chain, especially those envisioned by enthusiasts where the functionality is distributed all over the cloud. The transaction messaging required for this distributed scenario is still in its early days.

(ii) Risk of Service Centralisation - because of the above issue, the most rational architecture is not a cloud, but to centralise all the services somewhere. But then you run into the problem of "who owns the service". It works like this:

- Hand service over to supplier
- Sign supplier's Service Level Agreement
- Hand your data, workflow, etc to the supplier (after considerable cost & effort to systemise it for a 3rd party to be able to run)
- Find the SLA is more for the supplier's, not your benefit, and their response to your needs as a company is not as flexible as you may desire. Get used to filling change requests!
- Find, in disputing the SLA, that the supplier has a tight grip on your corporate testicles - switching you off sends you rapidly down Swanee, sans paddle

If you have been around the block more than once, its possibly the opposite case - you negotiate such a good SLA with the supplier that they can't actually afford to maintain it, and things slowly fall apart.

(Or, you find that its a startup and there is no SLA in effect)

(iii) Risk of security compromise - when there are multiple parties in the transaction, even of good faith, then breaching security becomes far more of a possibility. It just does.

(iv) Risk of Handing over Control - the reason the PC was so popular vs the mainframe (the "first cloud" architecture) was that the user had control of the operating environment.

There is a more subtle version of this last one - ie Who Do We Trust with our data (as in who won't peek / datamine / resell) . This is a battleground that will be interesting to watch, as in general the lower cost the service, the more its costs will potentially be offset in this way (see this post on FreeConomics - why your data is free but everywhere in chains)

The issue quite simply is this - in my business, I am committed. The infrastructure partner is only involved.

Now, I hear you argue, Outsourcing has done very well - and it has - but it is instructive to look at what is usually outsourced and by whom. By and large its non core, non customer facing applications (we have in fact done a number of pieces of work re-insourcing customer facing services). So Email, HR, Sales Analysis and Order Management etc are outsourced, but core workflow far less so. Also (whisper this who dars) outsourcing is often a piece of financial, not technical, engineering. Smaller companies do tend to outsource more critical services, but its driven by their economic necessity (lower resource levels, lower cost plus trying to get a jump ahead of large players) rather than choice in many cases.

The other point about existing outsourcing and webservicing is that it has - in our view - still scooped up the 80/20 economic benefits from the low hanging fruits (I'm looking at a McKinsey study from 2006/7 as I write this, I can't see where the next "step change" comes from via Cloud Computing. In other words, the existing "good enoughs" are probably good enough to prevent most Cloud business cases from passing corporate muster in the near term, and the real next level benefits are still in the "hard to do" category.

So what to do?

For larger companies, our view is that this a technology to try out in some small, non core activity - maybe an area that does need a certain bit of sizzle. It feels to us like a "prepare to be a fast follower" game at the moment. For SOHOs and SE's its more of a gambler's advice gambit - make sure that you can afford to lose what you stake (and even better, back it up to the nines).

Update - Gapingvoid has posted an article on this that I liked - ( it comes to similar conclusions ) and Confused of Calcutta has an interesting argument, that Open Source will be the antidote to the "Rich Get Richer" dynamic. I don't think so - Open Source is a tool here, we have built various pieces of service infrastructure using Open Source in this space - the key issue is who is willing to take on the SLA - ie under-write the Risks