Roughly described, a database accelerator is installed transparently in a network between database client and server systems. It maintains a pool of connections to the database server(s) for re-use as needed. Connection requests from client systems are received and completed by the accelerator, and queries received on such client-side connections are forwarded through pool connections toward the database servers. If no appropriate connections are available when needed for a particular query, the accelerator forms one by emulating a client system requesting a connection to the desired database server. If the maximum number of connections has already been reached, then the query is queued until a connection becomes available in the pool. By installing the accelerator transparently in the network, no modifications are required to the software in either the client systems or the database servers to add connection pooling functionality to an existing client/server database environment.