Simulation Environment And Application For Flash Memory

Hard disk is the most popular data storage device so far, because of its fast I/O speed, high capacity data storage characteristics. But, with the develop of electron technology recent years, hard disk is not sufficient for some new data storage requirements, such as: in aerospace field, people often require storage media with good shock-resistent, and can work in high/lower temperature; in handheld embedded devices, people need storage media with a small volume. However, due to the mechanical structure, hard disk cannot fit these requirements, moreover, these mechanical structures also affected the increse of I/O speed. The problem of speed dismatch between hard disk, CPU and high-speed data bus has been becoming more and more serious. In this situation, the flash memory, as a new data storage medium, comes out.Flash memory appeared in 1980s, with high speed, non-volatile, power economic, shock-resistant, small size, etc. these fine characteristics make it to be a preference to hard disk. With the lower price and the rapid growth in capacity of flash memory chips, flash memory has become the most important secondary storage mediums, and have been used in enterprise-level computing environments (Sina 2008).Recently, researchers have proposed many data management algorithms and techniques on flash memories. However, researchers realize that it's difficult to evaluate their new algorithms which are based on flash memories on current DBMS. Researchers always use software simulation for their evaluation experiments, but these make much hard and redundant work. Here, the redundant work refers to that people usually need to adopt some similar algorithms and architectures in the simulating tools. And different simulation environments make researchers cannot compare their results between each others. So if we build a reusable and flexible framework for the algorithms evaluation on flash memories, it will be much better so that different researchers can reuse the framework and easily perform the evaluation work of their algorithms.There're two kinds of flash-based researches: One is out-of-SSD algorithms (for example: flash-based DBMS system, flash file system et al.). Another one is in-SSD algorithms (for example: flash translation layer, wear leveling, garbage collection et al.). Researchers use software simulator or SSD device to evaluate out-of-SSD algorithms all the time, but they can only use simulator for in-SSD algorithms evaluation instead due to the encapsulation of SSD device. So, a good flash simulator should be able to evaluate those two kinds of flash-based algorithms. Meanwhile, the simulator should be easy to configure and use. Researchers can adjust simulator to make it suitable for their experiment with little work. Now we know, a good flash simulator should be with these features: feasible, easy to use and configure, be suitable for all flash-based researches.The contributions of the dissertation are listed as follows:(1) A new, flexible and efficient flash simulation platform is introduced in this paper. We use object-oriented and modular technology to design and develop this platform, which make it can be used in almost flash-based evaluation experiments.(2) The flash simulation will be used to evaluate two kinds of algorithms. First, buffer management evaluation, second, flash translation layer evaluation. These two experiments show that our simulator is suitable for all kinds of algorithm evaluations.