How to build better clouds

Ten or 15 years ago it was common to nominate each consecutive year The year of the Linux desktop. This doesn’t happen anymore. Do you know why? I feel that desktops have become less relevant than they used to be. And singular servers are less relevant, too. Everything is seemingly moving to the “cloud”.

Yes, I’m telling an old story. If you expect me to raise privacy concerns, that’s correct. But this time I’ll approach the subject from a different angle. Nobody’s going to argue that if you don’t control your server, from data centre premises to software, you don’t control your data. A landlord may seize your hardware if you don’t own the building. A cloud provider can do anything with your bytes and bits, because it can. If the data centre is yours, but is in a different country so it’s closer to your customers, you don’t own it even if you do on paper.

The “everything yourself” vs “a shared thing” dilemma isn’t new. I bet you have a boiler in your house somewhere. I don’t because we rely on central heating. This means I have no direct control over the temperature in my bedroom, but also no liability to fix the pipes when they break. I pay my utility bills, and these things “just work” for me. In a nutshell, this is all about the costs of running it myself and the risks of not having the service when I need it.

Something similar happens in clouds. In many cases, having something as a service is cheaper and the risks are justified. Sometimes this isn’t the case and you really need it as your own. The point is that cloud computing concept isn’t about to go away anytime soon. So let’s focus on how to make them safe. Pervasive cryptography (where you rather than the provider own keys) and replication all feel like steps in the right direction.