RCAT : A Scalable Architecture for Massively Multiuser Online Environments

We describe RCAT, an architecture designed to scale real-time online environments up to thousands of interacting users in an applicationindependent manner. RCAT takes inspiration from the principles that allow Web applications to scale horizontally, while being aware of the challenges brought up by real-time multi-user interactions. We present a performace study that shows where the bottlenecks of RCAT are, and the thresholds at which additional cores are needed. Our laboratory experiments with a reference application that produces 2.5 updates/second show that 229 interacting users can be appropriately served with 16 cores. More generally, our experiments show that to support twice the number of users, the server side needs four times more cores, independent of the update frequency of the application. This finding is important in order to engineer the next generation of cloud-based massive multiuser environ-