Retention-aware data tiering algorithm for hybrid storage arrays
A memory array controller includes memory media scanning logic to sample a bit error rate of memory blocks of a first memory device. A data management logic may then move data from the first memory device to a second memory device if the bit error rate matches a threshold level. The threshold level is derived from a configurable data retention time parameter for the first memory device. The configurable data retention time parameter may be received from a user or determined utilizing various known machine learning techniques.
1. A hybrid storage system comprising:
a first memory device;
a second memory device;
memory media scanning logic configured to periodically sample a bit error rate of memory blocks of the first memory device;
data management logic configured to:
apply an evaluation algorithm to determine an initial data temperature selected from hot data and cold data;
write hot data blocks to the first memory device; and
write cold data blocks to the second memory device; and
a wear leveling logic in the first memory device configured to:
determine a configurable data retention time parameter for the first memory device, wherein the configurable data retention time parameter corresponds to a defined time period for moving data blocks from the first memory device to the second memory device;
map the configurable data retention time parameter to a bit error rate threshold level; and
redirect, responsive to the memory media scanning logic generating an indication that the bit error rate matches the bit error rate threshold level, a write of a hot data block from the first memory device to the second memory device.
2. The hybrid storage system of claim 1 , wherein:
the first memory device is a solid state drive; and
the second memory device is a hard disk drive.
3. The hybrid storage system of claim 1 , wherein:
the first memory device is a first solid state drive; and
the second memory device is a second solid state drive having a higher latency and higher capacity than the first solid state drive.
4. The hybrid storage system of claim 1 , wherein the memory media scanning logic comprises static wear leveling logic.
5. The hybrid storage system of claim 1 , wherein:
the memory media scanning logic is implemented on a local controller of the first memory device; and
the data management logic is implemented on a central controller for the first memory device and the second memory device.
6. The hybrid storage system of claim 1 , wherein:
the first memory device is a solid state drive; and
the data management logic, the wear leveling logic, and the memory media scanning logic are implemented by a controller of the solid state drive.
7. The hybrid storage system of claim 1 , wherein the configurable data retention time parameter is correlated to the bit error rate threshold level in a memory lookup table.
8. The hybrid storage system of claim 7 , further comprising a host device interface configured to receive the configurable data retention time parameter from a host device.
9. The hybrid storage system of claim 1 , wherein the configurable data retention time parameter is generated by machine learning logic.
10. The hybrid storage system of claim 1 , wherein:
the memory media scanning logic comprises the wear leveling logic; and
the wear leveling logic is configured to:
apply the configurable data retention time parameter and the bit error rate to overrule the determination by the evaluation algorithm that the initial data temperature of the hot data block is hot data; and
initiate a memory migration process by the data management logic to redirect the write of the hot data block to the second memory device.
11. A hybrid storage system comprising:
a first memory device;
a second memory device having different data storage characteristics than the first memory device;
means in the first memory device to periodically sample a bit error rate of the first memory device;
means to apply an evaluation algorithm to determine an initial data temperature selected from hot data and cold data;
means to write hot data blocks to the first memory device;
means to write cold data blocks to the second memory device;
means to determine a configurable data retention time parameter for the first memory device, wherein the configurable data retention time parameter corresponds to a defined time period for moving data blocks from the first memory device to the second memory device;
means to map the configurable data retention time parameter to a bit error rate threshold level; and
means to move, responsive to the bit error rate meeting or exceeding the bit error rate threshold level, hot data blocks from the first memory device to the second memory device.
12. The hybrid storage system of claim 11 , further comprising means to apply the bit error rate and the configurable data retention time parameter to overrule the determination of the initial data temperature as hot data.
13. A computer-implemented method comprising:
applying an evaluation algorithm to determine an initial data temperature selected from hot data and cold data;
writing hot data blocks to a first memory device;
writing cold data blocks to a second memory device;
determining a configurable data retention time parameter for the first memory device, wherein the configurable data retention time parameter is a defined time period for moving data blocks from the first memory device to the second memory device;
mapping the configurable data retention time parameter to a bit error rate threshold level;
periodically sampling, by the first memory device, a bit error rate of memory blocks of the first memory device;
comparing, for a hot data block, the sampled bit error rate to the bit error rate threshold level; and
redirecting, responsive to the sampled bit error rate exceeding the bit error rate threshold level, a write of the hot data block from the first memory device to the second memory device.
14. The computer-implemented method of claim 13 , further comprising:
receiving the configurable data retention time parameter from a user input, wherein the configurable data retention time parameter is a user-defined time period for moving data blocks from the first memory device to the second memory device.
15. The computer-implemented method of claim 13 , further comprising:
receiving the configurable data retention time parameter from a machine learning component.
16. The computer-implemented method of claim 13 , wherein mapping the configurable data retention time parameter to the bit error rate threshold level uses a lookup table that stores bit error rate threshold levels as a function of the configurable data retention time parameter.
17. The computer-implemented method of claim 13 , further comprising:
marking, responsive to the sampled bit error rate exceeding the bit error rate threshold level, the hot data block to be moved to the second memory device; and
moving, the hot data block to the second memory device.
18. The computer-implemented method of claim 17 , further comprising:
checking, responsive to marking the hot data block to be moved, cache data for the hot data block; and
flushing the cache data for the hot data block to the second memory device.
19. The computer-implemented method of claim 17 , wherein comparing the sampled bit error rate to the bit error rate threshold level overrules the configurable data retention time parameter for marking the hot data block to be moved to the second memory device.
20. The computer-implemented method of claim 13 , further comprising:
applying the configurable data retention time parameter and the sampled bit error rate to overrule the determination by the evaluation algorithm that the initial data temperature of the hot data block is hot data; and
initiating a memory migration process to redirect the write of the hot data block to the second memory device.