Marco Seiriö's CEP Blog

First there was the service bureau, then the application service providers (ASP), after that came software as a service (SaaS) and now its time for Cloud Computing. Basically we have one single concept which is as old as IBM Mainframes. Every generation seems to rediscover it, and create their own version of it.

The goal this time seems to be to have a huge number of servers called a Computing Cloud, it sounds really good if you don’t understand anything about software development. Just throw your software into this huge friendly cloud and it will automatically execute on an “Elastic” could of computing power. Just buy capacity where it happens to be cheap today and when you need more power just click in a slick web gui and add more processing power to your application.

Can we see a problem here?

The problem is that only a tiny fraction of todays software can be written to run in this manner. Even smaller fraction is actually written in this manner. That’s why every generation fails with taking over the world with ASP/SaaS/Cloud computing. It’s just not for everyone.

But for a small subset of applications this is just pure heaven. If you build applications which contains a lots of small things which all can be done concurrently you could in theory build an application which can take advantage of a service bureue, eh sorry, I mean an elastic computing cloud of course.

To get there, you need two things. First your application must naturally fit into this concept (most don’t) and you need to have some clever coders to implement something which actually runs smoothly on a large number of servers (most can’t)

If, and I say if, the event driven way ever becomes the natural and mainstream way of building system I think we have a good chance to actually use these processing clouds for something. If not, there will soon be lots of silent data centers out there.

But there’s hope. Event processing and event driven architectures are gaining some traction currently. I think the analysts and tool vendors will need something new to analyze and build tools for. SOA is nothing new anymore and hard to profit from. Hopefully event driven architectures (EDA) will be this new thing that can (once again) save the world and promise new wonderful types of software solutions.

When the cloud hype starts to fade I think we will have left some companies which can provide us with some really good execution environments for event processing software. To get there, we need at least ten years of evolution of the current state of event processing tools and systems. But I think we as a community are on the right track. The revolution just takes a bit longer than most of us would like.