Design And Implementation Of SSD Flash Cache In PostgreSQL

With the advent of the era of big data, the size and growth rate of data is growing rapidly. Massive data processing raises a great challenge for computer. The speed of disk has become a bottleneck for storage system performance. SSD, which based on flash technology, provide an efficient way to solve this problem. Flash memory is a semiconductor medium, without mechanical seek latency. So it can read and write very fast. Besides, SSD expend less power than disk. But compared with HDD, SSD are more expensive. Hybrid storage technology is the balance between performance and cost-effective.In this project, we start with the characteristic of flash and the access of SSD. Only have a knowledge of SSD can help to develop SSD cache. As this project based on PostgreSQL, we analyzed its source code. Because it’s context of SSD cache.Finally we developed SSD cache module based on ext3 and UBIFS. Then we did TPC-C test. SSD generally consists of multiple NAND flash memory chips and FTL. Common file management system cannot directly run on NAND. The FTL encapsulates the NAND flash memory chips, making the ordinary file management system access SSD and HDD the same way. But the mode of common file management system + FTL do not fully take advantage of the performance of the NAND flash memory.UBIFS is a kind of NAND flash memory file management system. It is better than ext3 + FTL this model. Using UBIFS can access the data on the SSD more quickly. Using UBIFS can bypass the FTL layer, which can take advantage of NAND flash memory.