Hit ratio of caching disk buffers

Divided Disk Cache and SSD FTL for Improving Performance in Storage

“…However, since metadata is accessed more frequently, the number of metadata hits increases significantly, compensating for the reduced data hits [2,5,6]. On the other hand, metadata is relatively small, and metadata cache space must be allocated carefully.…”

“…Hong evaluated the performance impact from using micro electromechanical systems (MEMS) as metadata storage and disk cache [6]. MEMS have seek-times 10-20 times faster than hard drives, storage density 10 times higher, and also lower power consumption.…”

“…Hospodor provides the total access time when the required data is not present in the disk cache and the physical disk must be accessed [6]. For simplicity, we consider cache transfer time to be equal that for a magnetic disk, although the latter is much higher as it involves rotating the physical disk to read from the sectors [7].…”

“…Every disk cache hit results in substantially reduced (1-4 ms) I/O than would be required to access the disk itself (10-100 ms) [16]. We evaluate the performance of a disk cache divided into data and metadata regions, where metadata represents only a small portion of the memory, but is accessed very frequently [2,5,6].…”

A Space-efficient On-disk Prefetching Algorithm

“…In similar reason, modern hard disk drives can also do caching and prefetching with their own CPU and memory unit. When an operating system want to read a block from the hard drive and the data of that block is found in the on-disk memory, the read request can be served without any mechanical movement [1][2][3]. This makes hard disk drives have faster response time overall, especially in sequential workloads [4,5].…”