Dave is talking about Winning with Cloud Computing Step-by-Step. The presentation is up on slideshare. [link added 4.1.2009] Dave’s presentation is based on work from a forthcoming book.
The basic idea is you can extend your SOA to the cloud, utilizing external resources, either business or informational services, or infrastructure resources.
Cloud & SOA lets us mix an enterprise architecture cocktail. [works for me]
You can’t replace enterprise architecture with cloud computing. You can’t replace SOA with cloud computing. You always need an architectural strategy. Adding cloud computing allows you to cash-in on SOA. Saving money because you are using other people’s work.
If you want to boil cloud computing down, it’s switching the use of resources that are currently available in your enterprise, out in the cloud, but in a virtualized, reliable, secure manner via a subscription model.
Cloud computing is neither completely evil or completely good. Lots of hype at this conference. Lots of skeptism in IT. Cloud computing has tendency to be cost-effective, but need to consider “all-in” cost. Still need to maintain your systems (applications, databases etc), do integration, deal with security on the cloud.
On-premise, you’ll always be under utilizing your hardware and software investments. Many on-premise systems are only at 3% capacity. Cloud delivered, you pay as you go. So, your entire investment is productive – running business transactions.
Warnings:
- Not all computing resources should exist in the cloud
- Not always cost-effective – all-in cost
- Don’t move crap to the cloud (not what he said today, but I heard him say it last week)
When it’s a fit:
- processes, applications and data are largely independent
- integration points well defined
- lower level of security is ok
- core enterprise architecture is healthy
- web is desired platform
- cost is an issue
- applications are new
When it’s not a fit: essentially the opposite for 1-6, plus: application is legacy, application requires native interface
Start with the architecture: understand your business drivers, information under management, services under management and core business processes. Use whatever EA framework works for your organization.
Getting ready:
- accept that it’s ok to leverage external services in your SOA
- create a strategy for the consumption and management of cloud services
- create a proof of concept now. get a login. try stuff.
Dave has his 17 point stepping into the clouds list up now. I’ve blogged on this before. Items include, assessing culture, business, value, data, services, processes, cloud resources. Then, identify candidate data, services, processes. Create governance strategy. And some more.
Dave’s 17 points could be called, “stepping over the hype to architect your way to the cloud”.
Man! Dave just crashed a lightening bolt through my fluffy clouds. He shares, we’ll never, ever, have a magic console that allows us to drag and drop icons and suddenly have cloud applications and cloud deployments. He implies, we’ll actually have to do some work. It’s almost as though he’s been through a paradigm shift or two.
Other thoughts from Dave:
- external cloud services should function like any other enterprise resource – should work the same
- evaluate cloud providers using similar validation patterns as you would with business and data center solutions
- consider private clouds; right now, that’s really a virtualization story, recouping that unused capacity
Q&A
Discussion on Vertical clouds versus private. A hedge fund cloud is not private, because there would be multi-parties involved. Private cloud is for an organization and their data.
Discussion on SLAs, compensation and risk. If you get (expect) compensation when the public cloud doesn’t meet the service levels, you’ve increased the cost, and therefore the value of the public cloud. [This point is interesting, I’ll be exploring it further. Certainly, there needs to be some penalty for not meeting an SLA, otherwise, why have one].
Comments