Clustering – load distribution on generation

On the hand side, FirstSpirit supports “vertical scaling”, i.e. it is possible to increase the system performance by adding resources, such as more CPUs or by increasing the main memory, as processes such as multi-threading and caching are used to a large extent.

On the other hand side, FirstSpirit supports “horizontal scalability”, too.

One aspect of this “horizontal scalability” is load distribution to the members of the cluster during generation of the FirstSpirit content. The generation is segmented at schedule level. The generation actions (within one or several schedules) can be distributed to the cluster nodes. A generation action is completely dealt with on a cluster node. Other, parallel pending generation actions can be distributed to other cluster nodes. Scaling of the actions via the cluster nodes takes place automatically. If a valid licence is available for the function the required settings can be defined via the FirstSpirit ServerManager.

Apart from generation of the preview, the generation of content is one of the most time-critical operations in a FirstSpirit environment. Here high computing performance requirements and the wish for the shortest possible waiting time have been addressed together. The aim of the FirstSpirit “Clustering” function is to increase performance in multi-user environments. To this end the computing-intensive generation of the Master server is moved to one (or several) other servers (generation servers). 

The following pages describe the use of the function in the generation of FirstSpirit content.