From a performance perspective, the master database is a critical component of the system. All synchronized transactions created in the system are eventually committed in the master database. Similarly, publication data is refreshed from the master database. From the system infrastructure point of view, this means two things:
▪ The capacity of the master database server must be sufficient to manage the CPU and disk-I/O load that is caused by replica database transactions and refreshes. Some additional disk I/O is caused by the store and forward messaging of the synchronization architecture. If the additional disk I/O that is caused by the store and forward messaging becomes a problem, consider using the synchronous refresh method described in Synchronous replication: Messageless REFRESH.
▪ The fault tolerance of the master database server must be at a sufficient level. Since the replica databases communicate with each other only through the master database, the master database server is the single point of failure. If the master database server goes down, synchronization between replica databases stop. To avoid problems caused by master database server failure, consider using a solidDB high availability component, see High availability.