Macss: A metadata-aware combo storage system

To improve the I/O performance of servers, replacing Hard Disk Drives (HDD) with Solid State Drives (SSD) may be the best way, for SSDs have higher throughput than HDDs in terms of write and read, especially for random access pattern. But the cost per gigabytes (GB) of SSDs is ten or more times higher than that of HDDs. So constructing storage system with HDDs and SSDs is a practical way to improve the performance while at a lower cost. In this paper, we presented Macss, a combo storage prototype with block mapping in generic block layer based on Linux Ext3/Ext2. In Macss, SSD is used as metadata keeper, read-data buffer, and synchronous write data pool. We also evaluated Macss with the benchmark and the real workload. The experimental results show that even with a small proportion of SSD, more than 35% I/O requests can be served by the SSD for real workload. As a result, the average I/O response time is reduced.