Optimizing flash device write operations
There is disclosed a technique for use in optimizing write operations for flash devices. A system having a plurality of flash based solid state drives receives a write request to overwrite existing data stored on the solid state drives with new data. The write request data is formatted using a write granularity having a first size and the solid state drives are configured with a write granularity having a second size. Corresponding existing data is retrieved. The new data and the existing data are subdivided into multiple corresponding subunits where each subunit has a size equal to the second size. Each new data subunit is compared with each corresponding existing data subunit to identify which new data subunits include modified data. The new data subunits identified as having modified data are written to corresponding locations on the solid state drives.
1. A method for use in managing data storage, the method comprising:
receiving, at a system having a plurality of NAND flash memory based solid state drives, a write request to overwrite existing data stored on the solid state drives with new data, wherein the write request is formatted using a first write granularity size corresponding to a hard disk drive write granularity and the solid state drives are configured with a write granularity having a second write granularity size, wherein the first write granularity size is larger than the second write granularity size;
avoiding overwriting the existing data with the new data using the first write granularity size by:
reading existing parity data from cache;
comparing parity data of the new data with the existing parity data from cache to identify which new data subunits in the write request include modified data; and
writing the new data subunits identified as having modified data to corresponding locations on the solid state drives.
2. The method as claimed in claim 1 , wherein a plurality of the solid state drives are configured and arranged as a redundant array of inexpensive disks (RAID) group and data writes are stripped across two or more drives.
3. The method as claimed in claim 2 , wherein reading the existing data includes retrieving the existing data and parity data associated with the existing data.
4. The method as claimed in claim 2 , wherein reading the existing data includes retrieving the existing data and parity data associated with the existing data by reading the existing data and parity data from SSDs storing the data thereon.
5. The method as claimed in claim 4 , further including calculating new parity data associated with the partial block write.
6. The method as claimed in claim 5 , wherein calculating the new parity data includes removing the existing partial block write data from the existing parity data and recalculating the parity data using the new data.
7. The method as claimed in claim 6 , wherein reading the existing data includes retrieving the existing data and parity data associated with the existing data by reading the existing data and parity data from cache subsequent to recalculating the parity data.
8. The method as claimed in claim 2 , wherein the new data is a partial block write.
9. The method as claimed in claim 1 , further including:
determining the write data size is less than the second write granularity size;
retrieving existing data;
creating a new write with a write size equal to the second write granularity size, wherein the new write data includes the new data and the remaining write data; and
writing the new write to SSD.
10. A system for use in managing data storage, the system configured to:
receive, at a system having a plurality of NAND flash memory based solid state drives, a write request to overwrite existing data stored on the solid state drives with new data, wherein the write request is formatted using a first write granularity size corresponding to a hard disk drive write granularity and the solid state drives are configured with a write granularity having a second write granularity size, wherein the first write granularity size is larger than the second write granularity size;
avoid overwriting the existing data with the new data using the first write granularity size by:
reading existing parity data from cache;
comparing parity data of the new data with the existing parity data from cache to identify which new data subunits in the write request include modified data; and
writing the new data subunits identified as having modified data to corresponding locations on the solid state drives.
11. The system as claimed in claim 10 , wherein a plurality of the solid state drives are configured and arranged as a redundant array of inexpensive disks (RAID) group and data writes are stripped across two or more drives.
12. The system as claimed in claim 11 , wherein reading the existing data includes retrieving the existing data and parity data associated with the existing data.
13. The system as claimed in claim 11 , wherein reading the existing data includes retrieve the existing data and parity data associated with the existing data by reading the existing data and parity data from SSDs storing the data thereon.
14. The system as claimed in claim 11 , wherein the new data is a partial block write.
15. The system as claimed in claim 14 , further configured to calculate new parity data associated with the partial block write.
16. The system as claimed in claim 15 , wherein calculating the new parity data includes removing the existing partial block write data from the existing parity data and recalculating the parity data using the new data.
17. The system as claimed in claim 16 , wherein reading the existing data includes retrieve the existing data and parity data associated with the existing data by reading the existing data and parity data from cache subsequent to recalculating the parity data.
18. The system as claimed in claim 10 , further configured to:
determine the write data size is less than the second write granularity size;
retrieve existing data;
create a new write with a write size equal to the second size, wherein the new write data includes the new data and the remaining write data; and
write the new write to SSD.
19. A computer-program product including a non-transitory computer-readable storage medium encoded with computer-program code that, when executed on a processor of a computer, causes the computer to manage data storage in a data storage system, the computer-program product further comprising:
receiving, at a system having a plurality of NAND flash memory based solid state drives, a write request to overwrite existing data stored on the solid state drives with new data, wherein the write request is formatted using a first write granularity size corresponding to a hard disk drive write granularity and the solid state drives are configured with a write granularity having a second write granularity size, wherein the first write granularity size is larger than the second write granularity size;
avoiding overwriting the existing data with the new data using the first write granularity size by:
reading existing parity data from cache;
comparing parity data of the new data with the existing parity data from cache to identify which new data subunits in the write request include modified data; and
writing the new data subunits identified as having modified data to corresponding locations on the solid state drives.