"When will my server run out of resources?" This is a question on the minds of many experienced database administrators. This is particularly relevant in Cassandra given that, as data grows, compactions get triggered to keep the database healthy. Compactions cause a temporary spike in disk usage and disk I/O while old and new SSTables coexist, and organizations can get caught off guard with a saturated cluster that becomes tricky to expand and unable to respond effectively to user requests.
In this talk we describe how we use Prophet, Facebook's open source project, at Pythian to perform historical trend analysis on relevant Cassandra metrics, and predict the future trends of Cassandra resources, considering seasonality and holiday effects. This allows us to grow clusters in a proactive way, helping companies reduce costs and improve availability.
