Abstract
An increase in the mapping table unit size without corresponding increases in the size of the host logical block address unit size results in a size mismatch that can cause a data write operation to be unaligned between the LBA and a solid-state drive. The read-modify-write operations to deal with unaligned write requests can take a relatively long time, resulting in a large performance drop. This disclosure describes techniques to improve the performance of unaligned data writes to SSD storage by employing a smaller Indirect Unit (IU) table on top of the original IU table in conjunction with a Bloom filter. Whenever data is to be written to or read from the SSD, the host LBA can be routed to the appropriate mapping table by looking it up with the Bloom filter. An unaligned write operation is routed to the smaller IU table as long as it is not full, with the host LBA added to the Bloom filter. In case the smaller IU table is full, the unaligned write can fall back to the traditional approach. Implementation of the techniques can help avoid the I/O performance hit because of unaligned write operations to an SSD.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Tao, Jun and Gao, Wuzi, "Mapping Table with Bloom Filter to Improve SSD Unaligned Write Performance", Technical Disclosure Commons, (February 11, 2025)
https://www.tdcommons.org/dpubs_series/7828