Imagine that you are a process owner, or some other senior person in the business, wanting to deliver a new system. Lets say that you have come up with a great product that takes advantage of the current market uncertainties and has the chance of becoming big, viral, social and web 2.0 buzzword compliant. You've just walked out of the unscheduled 'budget adjustment and cost cutting' committee meeting and have been asked to deliver the same thing for half the money. You pour over your budget and notice that the infrastructure costs make up more than half of the budget and, wondering if there is some way you can shave off some costs, you call up the infrastructure architect:
<ring> <ring>
You : This is the VP of Fantastical Product Developments speaking. I want to go over the hardware requirements quickly.
IA: Sure, go ahead
You: We don't expect this system to be that big initially and requested only one server. Can we get away with one?
IA: Yes, but you need at least two for load balancing.
You: Ok, so we need two then?
IA: Yes, but for each server we need a mirrored server for fail-over
You: Hmmm, ok. Four servers?
IA: Yes, but the corporate Enterprise Architecture standard says that your testing and UAT environment needs to be configured exactly the same as production.
You: You mean eight servers?
IA: Yes, and one for integration testing, one for the build server, one for the development server and two for the developers' VM's
You: You are pulling my chain! How can one server turn into thirteen servers?
IA: That is what the infrastructure design policy document specifies.
You: And why, if I take the total cost and divide by thirteen I get a gazillion dollars per server?
IA: Well that cost includes the OS licences, the database licences, the backup licences, the management interface licences and a few other bits and bobs
You: So that is my total capex?
IA: Let me look. Hmmm, it seems that you cost schedule excludes disk space.
You: What the f! At that price the servers come without disks and frikkin' "lasers"?
IA: Yes they have disks, but just for the OS, we need to buy disk space on the SAN from the Data Storage Centre department. I'll send you a price on that.
You: I bet you there is more!
IA: Well actually you are right there! There is an initial setup cost by the data centre to get your servers installed - that will cost 32.5% per server. And there is an annual cost of 21.3% per server to keep it going.
You: What do you mean 'keep it going'?
IA: Um... backups, monitoring, reporting, fire drills - that sort of thing. I'll sent you prices.
You: Never mind, I'm cancelling this frikkin project!!!
<slam phone down>
The looming cloud is destined to change the above scenario. The dark could of recession is going to slash IT budgets (already evident in slow server sales) and business may find it tough to deliver with limited funds. But apparently clouds have a silver lining and in the case of Microsoft it is Azure rather than silver. Microsoft's Azure is a platform for delivering applications in the cloud - it includes some software bits and great big data centres owned and operated by Microsoft themselves.
'The Cloud' in this context refers to a bunch of computing resources out there on the Internet that you won't know anything about physically and neither do you care. If you use Gmail you are already using a cloud-based application. Gmail runs on nobody-knows-how-many servers and the data is stored somewhere, but we don't know where. In the case of Google, or YouTube or SkyDrive and others - most users don't really care. All that you care about is that it is available when you need it.
Cloud-based platforms that developers can use are not new, I took a look earlier this year at the Google App Engine, but tired of the need to wade through the platform language and framework (Python and Django). Finally Azure provides a platform for delivery of cloud based applications that use a lot of the existing skills that we have - lots of .Net and a web server that is familiar ASP.net running in IIS7. So with little effort .net developers can scale out their skills onto the cloud which will push a lot of the demand and I imagine that there will also be a lot of pull from business.
In the hypothetical scenario above, the process owner can forgo the hardware costs and huge capex, concentrating on optimising their budget for development or creative work. Hardware costs will come later but will not be as high because someone else will have economies of scale to bring down the cost of the data centre and hardware is only required when needed. The testing and UAT environment can be 'purchased' for the few days that it is needed and then given back.
I think that the global recession will be a trigger for different ways of delivering IT systems in terms of development processes, maintenance and infrastructure. Cloud computing is a good candidate for managing hardware costs during lean times and the shortage of cash for hardware purchases will definitely put it on the CIO's radar - us developers just need to make sure that we are ready for it.
Disclaimer
Maybe you should also talk to the hardware people to get their opinion. I am, of course, biased and being in software development I (apparently) know very little about these things. Luckily for me the time is coming when the only required knowledge is how to press a button in Visual Studio labelled 'Publish to Cloud'. I press the button and the application goes forth onto the cloud and I need to know as much about it as a tea leaf needs to know the history if the East India company (apologies to Douglas Adams).
Simon Munro