Adaptive super block wear leveling
A system can include a memory device a memory device comprising multiple dies, and a processing device, operatively coupled with the memory device, to perform various operations including identifying multiple management units to be programmed, where one management unit contains memory cells from a die having one endurance metric and another management unit contains memory cells from a die having another endurance metric, and determining a value of a media endurance metric for each management unit. The operations further include determining, for each management unit, a respective endurance exhaustion parameter defined by a relationship media endurance metrics, and distributing operations to each management unit based on the endurance exhaustion parameter.
1 . A system comprising:
a memory device comprising a plurality of dies; and
a processing device, operatively coupled with the memory device, to perform operations comprising:
identifying a plurality of management units of the memory device to be programmed, wherein a first management unit comprises a first plurality of memory cells residing on a first die exhibiting a first value of a media endurance metric and a second management unit comprises a second plurality of memory cells residing on a second die exhibiting a second value of the media endurance metric, wherein a respective value of the media endurance metric is associated with a respective data retention capability of a corresponding die, wherein the second value of the media endurance metric is different from the first value of the media endurance metric due to variation during manufacturing;
determining, for each management unit, a respective endurance exhaustion parameter that depends on a ratio between a value of a media endurance metric of a die and a predefined media endurance rating assigned to the management unit during manufacturing, wherein the value of the media endurance metric corresponds to the die on which the plurality of memory cells of the management unit reside, and wherein the predefined media endurance rating depends on an aggregated value of the media endurance metric of two or more dies in the plurality of dies, and wherein the endurance exhaustion parameter indicates remaining lifetime of the management unit; and
distributing memory access operations among the plurality of management units based on the respective endurance exhaustion parameters, wherein measurement of error bits of memory cells of the memory device is not relied upon.
2 . The system of claim 1 , wherein the respective endurance exhaustion parameter for a respective management unit is a normalized value of the media endurance metric of the die on which the plurality of memory cells of the management unit reside.
3 . The system of claim 1 , wherein the respective endurance exhaustion parameter for a respective management unit is a normalized count of program-erase operation cycles.
4 . The system of claim 1 , wherein distributing the memory access operations further comprises:
maintaining a difference between a count of program-erase operation cycles performed on the first management unit and the count of program-erase operation cycles performed on the second management unit, such that the difference is below a maximum difference.
5 . The system of claim 4 , wherein the maximum difference is defined in terms of a percentage of the first value of the media endurance metric or in terms of a percentage of the second value of the media endurance metric.
6 . The system of claim 4 , wherein the maximum difference is defined as a difference in the counts of program-erase operation cycles performed.
7 . The system of claim 4 , wherein each count of program-erase operation cycles is a normalized count and wherein the maximum difference is defined as a difference of the normalized counts of program-erase operation cycles performed.
8 . The system of claim 4 , wherein distributing the memory access operations further comprises:
mapping a first logical block address range to the first management unit and mapping a second logical block address range to the second management unit, wherein data in the first logical block address range is modified more frequently than data in the second logical block address range.
9 . A method comprising:
identifying a plurality of management units to be programmed on a memory device comprising a plurality of dies, wherein a first management unit comprises a first plurality of memory cells residing on a first die exhibiting a first value of a media endurance metric and a second management unit comprises a second plurality of memory cells residing on a second die exhibiting a second value of the media endurance metric, wherein a respective value of the media endurance metric is associated with a respective data retention capability of a corresponding die, wherein the second value of the media endurance metric is different from the first value of the media endurance metric due to variation during manufacturing;
determining, for each management unit, a respective endurance exhaustion parameter that depends on a ratio between a value of a media endurance metric of a die and a predefined media endurance rating assigned to the management unit during manufacturing, wherein the value of the media endurance metric corresponds to the die on which the plurality of memory cells of the management unit reside, and wherein the predefined media endurance rating depends on an aggregated value of the media endurance metric of two or more dies in the plurality of dies, wherein the endurance exhaustion parameter indicates remaining lifetime of the management unit; and
distributing memory access operations among the plurality of management units based on the respective endurance exhaustion parameters, wherein measurement of error bits of memory cells of the memory device is not relied upon.
10 . The method of claim 9 , wherein the respective endurance exhaustion parameter for a respective management unit is a normalized value of the media endurance metric of the die on which the plurality of memory cells of the management unit reside.
11 . The method of claim 9 , wherein the respective endurance exhaustion parameter for a respective management unit is a normalized count of program-erase operation cycles.
12 . The method of claim 9 , wherein distributing the memory access operations further comprises:
maintaining a difference between a count of program-erase operation cycles performed on the first management unit and the count of program-erase operation cycles performed on the second management unit, such that the difference is below a maximum difference.
13 . The method of claim 12 , wherein the maximum difference is defined in terms of a percentage of the first value of the media endurance metric or in terms of a percentage of the second value of the media endurance metric.
14 . The method of claim 12 , wherein the maximum difference is defined as a difference in the counts of program-erase operation cycles performed.
15 . The method of claim 12 , wherein each count of program-erase operation cycles is a normalized count and wherein the maximum difference is defined as a difference of the normalized counts of program-erase operation cycles performed.
16 . A non-transitory computer-readable storage medium comprising instructions that, when executed by a processing device, cause the processing device to perform operations comprising:
identifying a plurality of management units of a memory device to be programmed, wherein a first management unit comprises a first plurality of memory cells residing on a first die exhibiting a first value of a media endurance metric and a second management unit comprises a second plurality of memory cells residing on a second die exhibiting a second value of the media endurance metric, wherein the media endurance metric reflects a quantity of memory access operations, and wherein a respective value of the media endurance metric is associated with a respective data retention capability of a corresponding die, wherein the second value of the media endurance metric is different from the first value of the media endurance metric due to variation during manufacturing;
determining a respective count of memory access operations for each management unit;
determining, for each management unit, a respective endurance exhaustion parameter that depends on a ratio between the count of memory access operations of the management unit and a predefined media endurance rating assigned to the management unit during manufacturing, wherein the predefined media endurance rating depends on an aggregated value of the media endurance metric of two or more dies on which the plurality of memory cells of the management unit reside, wherein the endurance exhaustion parameter indicates remaining lifetime of the management unit; and
distributing the memory access operations among the plurality of management units based on the respective endurance exhaustion parameters, wherein measurement of error bits of memory cells of the memory device is not relied upon.
17 . The non-transitory computer-readable storage medium of claim 16 , wherein the respective endurance exhaustion parameter for a respective management unit is a normalized count of memory access operations.
18 . The non-transitory computer-readable storage medium of claim 16 , wherein distributing the memory access operations further comprises:
maintaining a difference between a count of program-erase operation cycles performed on the first management unit and the count of program-erase operation cycles performed on the second management unit, such that the difference is below a maximum difference.
19 . The non-transitory computer-readable storage medium of claim 18 , wherein the maximum difference is defined in terms of a percentage of the first value of the media endurance metric or in terms of a percentage of the second value of the media endurance metric.
20 . The non-transitory computer-readable storage medium of claim 18 , wherein each count of program-erase operation cycles is a normalized count and wherein the maximum difference is defined as a difference of the normalized counts of program-erase operation cycles performed.