In this paper, we study the special requirements of a special storage system: streaming media server, and propose a
solution to improve I/O performance of RAID storage system. The solution is suitable for streaming media applications.
A streaming media storage subsystem includes the I/O interfaces, RAID arrays, I/O scheduling and device drivers. The
solution is implemented on the top of the storage subsystem I/O Interface. Storage subsystem is the performance
bottlenecks of a streaming media system, and I/O interface directly affect the performance of the storage subsystem.
According to theoretical analysis, 64 KB block-size is most appropriate for streaming media applications. We carry out
experiment in detail, and verified that the proper block-size really is 64KB. It is in accordance with our analysis. The
experiment results also show that by using DMA controller, efficient memory management technology and mailbox
interface design mechanism, streaming media storage system achieves a high-speed data throughput.
This paper introduces a storage system which can adjust its storage strategy automatically according to the
current running environment, evolutionary storage system (ESS). In the evolutionary processes, ESS can
automatically choose one system organizing strategy which fits for the current running environment best, to keep the
dynamic balance of the whole system. ESS can always adapt to the request of current application, no matter the
storage device is out of date or not. So the most remarkable character ESS shown is that the system will be better
and better along with the running. The performance of ESS will increase gradually with the evolution. Three
evolutionary ways, system physical structure evolution, system logical structure evolution and data distribution
evolution were described and demonstrate separately.