Revision history [back]

Suppose you have to support many streaming: one server will not be enough. Beyond that, to provide a service resilient to failure you would need multiple instances of the service. Before spending time in experiments on the scalability and how to achieve it, I would like to know if there is available any guideline. Is it enough to have multiple instances behind a load balancer? Somehow I doubt it, because even if you may be using pipelines that are stateless, the problem is that Kurento relies on the concept of session. Supposing than the kurento instance to which I am connected fails, and the load balancer redirect me to another instance, will it work? Or given that the session is not shared the user will experience a failure? Is there any way to share sessions among multiple instances?