Abstract: A distributed application has a client software component on a client computer which communicates with server software on a server computer by way of a remote procedure call (RPC) system. The server has a persistent memory manager which keeps data available in high-speed memory and periodically backs up the data to disk or other lower-speed memory. Keeping transactions short and synchronous permits the entire database to be locked during each transaction. The system can provide responsiveness and usability approaching that of a stand-alone application.