A Concurrent Tuple Set Architecture for Call Level Interfaces

Abstract

Call Level Interfaces (CLI) are low level API aimed at providing services to connect two main components in database applications: client applications and relational databases. Among their functionalities, the ability to manage data retrieved from databases is emphasized. The retrieved data is kept in local memory structures that may be permanently connected to the host database. Client applications, beyond the ability to read their contents, may also execute Insert, Update and Delete actions over the local memory structures, following specific protocols. These protocols are row (tuple) oriented and, while being executed, cannot be preempted to start another protocol. This restriction leads to several difficulties when applications need to deal with several tuples at a time. The most paradigmatic case is the impossibility to cope with concurrent environments where several threads need to access to the same local memory structure instance, each one pointing to a different tuple and executing its particular protocol. To overcome the aforementioned fragility, a Concurrent Tuple Set Architecture (CTSA) is proposed to manage local memory structures. A performance assessment of a Java component based on JDBC (CLI) is also carried out and compared with a common approach. The main outcome of this research is the evidence that in concurrent environments, components relying on the CTSA may significantly improve the overall performance when compared with solutions based on standard JDBC API.