Archive for February, 2012

Why OpenStack?

Saturday, February 25th, 2012

As you probably saw, Ericsson has recently joined the OpenStack project. One may ask: why OpenStack, what’s the point of it?

The answer is not connected to Ericsson (and, of course, represents my personal view only), but it’s simple: it is really about openness. Openness is not about open source, alone but it’s about freedom of choice: freedom to pick and mix your hardware according to you needs; freedom to design your network as you want; freedom to use the virtualization technology of your choice. The ITC industry was extremely successful at creating a bewildering choice of compute, storage and network hardware, virtualization platforms, virtual network models and so on – to be fair, each with its benefits (and more often than not, drawbacks). Trying to create the perfect, one fits all “standard” HW and virtualization platform is doomed to fail and would just result in yet another set of offerings. Better leave that area alone: if you are building a mission critical cloud, you are better off with e.g. a telecom blade system; if you are just after a vanilla IaaS platform, many vanilla IT blade vendors will be happy to serve you: the choice is – and should be – yours.

This is where OpenStack got it right: it aims at providing a software abstraction layer above the HW and virtualization layer that can hide and enable management of this bewildering diversity. It’s a tacit recognition of the fact that real value comes from the rest of your stack: how you manage automation, elasticity, scalability, resiliency; how you integrate with BSS systems and so on. Let the guys fight out the race to the bottom, wrap them into OpenStack and create real value.

When we first started using OpenStack, it was soon enough strikingly clear how powerful their model is: we were able to add ground-breaking features such as WAN elasticity, distributed cloud support, SIM based authentication and many more (about which I will talk at the World Telecommunication Congress) – while keeping the abstraction layer untouched. In addition, being an open source project, it was like a wish come true process: whenever we identified the need for some feature, soon enough turned out that someone actually was working on it.

However, in order to live up to its promise, OpenStack has to keep following the same principles: support the broadest possible set of HW and support the broadest set of hypervisors, with equal quality. If the community starts relaxing this holistic approach, it will reduce the appeal of OpenStack and might eventually have the same fate as many other open source projects: loose momentum, loose corporate backing and eventually fold.

So, let the era of cloud freedom roll on and I’m sure some cool stuff will be coming soon from the OpenStack-based community.