Cache Optimization And Real-time Transaction Scheduling Of Embedded Database

With the development of embedded systems used in all kinds of environments, data management in embedded devices has been a critical problem. Researchers have developed database management system suited in embedded devices to store data and provide efficient data management. Generally embedded database has the feature of small foot-print, high speed, and real-time processing capability. Its data mainly resides in memory to remove the uncertainty brought by accessing the peripheral storage device.As the speed discrepancy between processors and memory becomes more and more evident, embedded architecture today adopts cache as a fast storage to save the mostly likely used instructions and data. It greatly enhances the performance of the system because it works at the same speed of the processor. But cache is usually managed by hardware and its behavior is unpredictable. It reduces the level of predictability for real-time transactions;therefore, currently used embedded database system has to consider this kind of uncertainty.To solve this problem embedded processors developed these days has the ability to lock instruction and data cache to prevent the contents in them from being swapped out. Using this feature we have designed an embedded database which is based on cache memory. It employs the technique of locking instruction cache and partitioning data cache and uses cache information to schedule real-time transactions to provide better predictability. It lays a good foundation for achieving high performance and predictability goal in processing data in embedded environments.