Scaling social networks — beyond Moore's law limits

Even though the Moore's law surprisingly still continues to apply, it yields only that — doubling of transistors in consumer-level CPUs every two years. It yields, however, smaller increase in CPUs computational capabilities and we used to demand more, more and more from our machines.

As the size of the Web increases, number and size of social networks follow this trend. And so does the Azouk network of expert and professional content. To handle it, to process it, to serve it — we need a system that can scale faster and cheaper than allows the Moore's law.

Azouk  is distrbuted from the ground up. This, in fact, is a no-brainer. It's a must-have for an expert social site like this. Sure, it may be not perfect from the very beginning, but we aim to avoid the costs of redesigning and rewriting all website's backend as our network grows.

We believe in free and open software, so the website's backend is based on Linux running inside OpenVZ containers, website is written in Django/Python, our gateway is nginx and we use GlusterFS distributed filesystem. We even contributed to the Open Source community, providing the heart of our distributed system architecture, Multiplexer. The latter provides distributed, fault-tolerant environment of loosely coupled services that can be run in a cloud.

Ready to see, how it works?