When monitoring multiple clusters of servers, we need a data collection method that's both scalable and configurable. The collectd daemon is a scalable statistics-gathering service, perfect for large clusters as it operates on a client-server model. A common collectd cluster may look like this, with collectd running on every server:

We can direct the statistics of several PostgreSQL servers to a central aggregate server. This server may process the data directly or forward it to a graph system for easy visualization. To gain this type of functionality, we need to spend some time installing and configuring collectd.