A Switching Compression Technology for the Fast Boot of Embedded Systems

• Main Authors: Meng-Hsiu Kuo, 郭孟修
• Other Authors: Chiu-Ching Tuan
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/v9kjj7

碩士 === 國立臺北科技大學 === 電腦與通訊研究所 === 102 === The demand for fast boot is mainly motivated by embedded system, especially for consumer devices, such as mobile phone, automotive vehicles, entertainment products, etc. Therefore, how to effectively improve the boot time so that the users can have a better feeling is still a major challenge to be conquered so far. For this, many fast boot approaches have been proposed. Among all the proposed approaches, most of which are based on hibernation technologies, which is an effective way to improve boot time. However the use of hibernation technology can result in bad block in flash memory due to the erase and write procedure once the hibernation mode is triggered. Aside from improving the boot time, how to reduce the amount of data erased and written to the Flash is also very important. Based on this reason, Linux system applies the LZO compression algorithm to compress hibernation image, and then write the compressed image to flash memory. Although the amount of data written to the flash can be reduced, the improvement on the boot time is still not satisfactory. Therefore, we proposed in this paper a more efficient algorithm for the compression and fast decompression of the hibernation image during software resuming process. A very good compression ratio on the hibernation image and better resuming time than that of LZO can be obtained with the proposed algorithm. In order to verify the proposed algorithm, a lot of experiments are performed on embedded system platforms. Simulation results show that an average improvement of 10.83% on the decompression time can be obtained when compared with that of LZO. Besides, when applied to software suspend, an average improvement around 8.66% can be obtained than that with LZO. We also compare the proposed algorithm with that without using any compression algorithm. By using a variable page pool size, an average improvement around 13% to 24% on the resuming time can be obtained by using the proposed algorithm when compared with that without using any compression algorithm. In our experiments, an average compression ration around 3.0 for the hibernation image can be achieved. Therefore less flash erasion and write operation is performed, and hence can extend the life time of flash around 2/3. Experimental results show that not only the boot time but also the life time can effectively improved by using the proposed approach which demonstrates the usefulness of the proposed algorithm on embedded systems.