PDC-NH: Popular data concentration on NAND flash and hard disk drive

In this paper, we present a network server architecture which concentrates popular data on both HDD and NAND flash. Popular Data Concentration (PDC) is a well known technique to reduce the power consumption of disk arrays. We extended PDC by adding NAND flash based Solid State Drive (SSD) and achieved better performance and energy savings. We developed a novel file placement policy which places popular files on either one of flash disk or hard disk based on their size, popularity and access characteristics. Our placement policy places large and sequential read files on HDD and small and random read files on SSD. This placement takes advantage of the benefits of each medium and hides their shortcomings. To boost the HDD´s performance, we further clustered correlated files and placed them on the HDD in a sequential manner. This reduces seek counts hence leading to a performance improvement. And we reserved some space of HDD to use it for write caching. By this, all the requests including reads and writes are concentrated on a designated hard disk and flash disk letting other disks to stay longer in low-power mode. We evaluated our architecture with trace driven simulator and the results show that our architecture is more energy efficient and performance effective compared with existing energy efficient architectures.