Hello for us Redis was a great solution for this problem: we run a kamailio based routing server that needs to handle millions of entries. mtree module worked great but it had a problem: after every change in the DB you need to perform a reload in order for the data to be in memory. For several million entires the reload took more than 3 minutes. That means that we needed to batch many changes and do a reload every few minutes (it made no sense to do a full reload for a single change). We moved to redis an soon as the ndb_redis module was released. With redis we have these advantages: -No reload needed, any change is instanly available (no batch either) -We can access the same redis instance from several servers -Using redis pipelining and unix sockets (Vicente provided a patch for using them in ndb_redis module), the performance is really astonishing. The disadvantage is that you no longer have a mt_match funcion, you have to build it yourself. The same logic can be applied to other modules that rely on a db backend but need to be loaded into memory for speed. For small size data the reload is very fast, but when it starts growing it may take a while. Regards Javi