Efficient Invocation of Transaction Sequences Triggered by Data Streams

Streaming media such as GPS, wireless sensor nodes, RFID and electric commerce are spreading over our lives. To process data streams from such streaming media, DSMS has been developed and widely used. While DBMSs are rarely used to process data streams directly, there is a growing need to integrate DSMS and DBMS, which means to use DBMS as a source of information related to data streams and to use DBMS to archive data streams. In such integration of DSMS and DBMS, transactions are triggered by data streams. Since arrival rate of data streams can be extremely high, it is necessary to invoke those transaction sequences efficiently to keep performance of DSMS. In this paper, we model invocation of transaction sequences and propose three schemes named Synchronous Transaction Sequence Invocation (STSI), Asynchronous Transaction Sequence Invocation (ATSI) and Order-Preserving Asynchronous Transaction Sequence Invocation (OPATSI). We built a prototype system which implements those schemes and conducted experiments to evaluate their performance. Our primary experiments showed that we experienced an order of magnitude higher throughput in transaction processing by using ATSI compared with STSI while breaking the order of transaction results. The results of experiments also showed that throughput of OPATSI is comparable to that of the ATSI, though it preserves the order of transaction results.