Open source: is it ready to run your datacentre?

Usually, updating a datacentre involves adding more hardware. More servers, more switches, more storage - just more stuff. But what if that's precisely not what you need?

It's an old saw but worth re-stating: unless you're getting the most out of what you've got, adding more hardware is often like rolling a cheese down a hill in order to find the last one you rolled down. It's good money after bad, in other words. Maybe you just need to manage it better. And maybe open source is the way to do it.

Such thoughts were prompted following a recent meeting with the newly-named Cloud.com (it used to be VMOps), an open source, datacentre management software developer.

Cloud.com has produced CloudStack, which aims to be, according to the press release, "an integrated software solution for delivering virtual data centres". Essentially, the company is repackaging software that's already in use in datacentres and making it open source. The heritage of the company includes execs from the likes of Sun and Zimbra, so they have previous - or as marketing exec Peder Ulander said: "All the execs and investors have been in and out of the open source market."

The way it works, claimed Ulander, is to sit above the hardware but below application management systems from the likes of IBM, CA, BMC and HP. providing a consistent user interface with the aim of speeding the introduction of cloud-based services -- spun as infrastructure-as-a-service. This then allows service providers and enterprises to deliver services while managing, provisioning and charging for those services. The key is that security, service management, metering and charge-back are integral elements of everything the system does.

This is territory currently partially occupied by Google App Engine and Microsoft Azure, but unlike those development tools, CloudStack delivers a set of completed features. It also shares attributes with cloud services offerings from such as Amazon, RackSpace and GoGrid.

While there is a commercial version of the product, you have to ask whether the OSS approach can work at this level. Looking at existing models, it makes sense for commercial system software developers to focus development efforts on core functionality, which helps lock users into your product, rather than interoperability, which doesn't. So what commercial systems tend to do is checkbox interoperability with other vendors' systems but, in practice, users often find that interoperability isn't quite as smooth as it was sold.

On the other hand, what OSS has going for it is an inherent connectedness, a willingness -- indeed eagerness -- to play well with other systems from other developers. If you're inserting a shim of software into a datacentre that purports to manage everything, interoperability isn't a nice to have, it's everything. And OSS does that well.

I haven't tried CloudStack so I can't comment on its goodness or badness. I do think however that it's an intriguing development for this area -- and could be a straw in the wind for the future.