Analysis of memory sub-systems based on threshold distributions
Disclosed is a system comprising a memory component having a plurality of memory cells capable of being in a plurality of states, each state of the plurality of states corresponding to a value stored by the memory cell, and a processing device, operatively coupled with the memory component, to perform operations comprising: obtaining, for the plurality of memory cells, a plurality of distributions of threshold voltages, wherein each of the plurality of distributions corresponds to one of the plurality of states, classifying each of the plurality of distributions among one of a plurality of classes, generating a vector comprising a plurality of components, wherein each of the plurality of components represents the class of a respective one of the plurality of distributions, and processing, using a classifier, the generated vector to determine a likelihood that the memory component will fail within a target period of time.
1. A system comprising:
a memory component comprising a plurality of memory cells, wherein each memory cell of the plurality of memory cells is capable of being in a plurality of states, each state of the plurality of states corresponding to a value stored by the memory cell; and
a processing device, operatively coupled with the memory component, to perform operations comprising:
obtaining, for the plurality of memory cells, a plurality of distributions of threshold voltages, wherein each of the plurality of distributions corresponds to one of the plurality of states;
classifying each of the plurality of distributions among one of a plurality of classes;
generating a vector comprising a plurality of components, wherein each of the plurality of components represents the class of a respective one of the plurality of distributions; and
processing, using a classifier, the generated vector to determine a likelihood that the memory component will fail within a target period of time.
2. The system of claim 1 , wherein classifying a first distribution of the plurality of distributions of threshold voltages comprises identifying a first tail in a quantile-quantile representation of the first distribution relative to a control distribution.
3. The system of claim 2 , wherein the control distribution is a normal distribution.
4. The system of claim 2 , wherein classifying the first distribution of the plurality of distributions of threshold voltages further comprises:
determining that the first tail corresponds to events for which a corresponding probability of occurrence is below a first cut-off value; and
treating the first tail as non-existent.
5. The system of claim 2 , wherein classifying the first distribution of the plurality of distributions of threshold voltages further comprises:
determining that the first tail corresponds to a deviation from the control distribution, wherein the deviation is below a second cut-off value; and
treating the first tail as non-existent.
6. The system of claim 2 , wherein classifying the first distribution of the plurality of distributions of threshold voltages comprises identifying the first tail as one of a top tail or a bottom tail, wherein the top tail corresponds to threshold voltages that are above a mean value for the first distribution, and wherein the bottom tail corresponds to threshold voltages that are below the mean value for the first distribution.
7. The system of claim 2 , wherein classifying a second distribution of the plurality of distributions of threshold voltages comprises identifying a shift of the second distribution relative to the control distribution.
8. The system of claim 1 , wherein the classifier comprises a machine learning model.
9. The system of claim 8 , wherein the machine learning model is trained based on a training input, a target output, and an association between the training input and the target output, wherein the training input comprises a training plurality of distributions of threshold voltages, wherein the training plurality of distributions comprises one of a plurality of distributions of threshold voltages for a failed memory component, a plurality of distributions of threshold voltages for a good memory component, or a plurality of distributions of threshold voltages for a good memory component, modified to simulate a failed memory component.
10. The system of claim 9 , wherein the target output comprises a likelihood that a memory device having the training plurality of distributions of threshold voltages will fail within the target period of time.
11. The system of claim 1 , wherein obtaining the plurality of distributions of threshold voltages comprises:
collecting unstructured threshold voltage data; and
converting the unstructured threshold voltage data into the plurality of distributions of threshold voltages corresponding to each one of the plurality of states.
12. The system of claim 1 , wherein obtaining the plurality of distributions of threshold voltages is responsive to an error-correction module detecting that the memory component failed to correctly store data following a store operation.
13. A method comprising:
obtaining, by a processing device operatively coupled with a memory component comprising a plurality of memory cells, a plurality of distributions of threshold voltages, wherein each of the plurality of distributions corresponds to one of a plurality of states of a memory cell of the plurality of memory cells, wherein each state of the plurality of states corresponds to a value stored by the memory cell;
classifying each of the plurality of distributions among one of a plurality of classes;
generating a vector comprising a plurality of components, wherein each of the plurality of components represents the class of a respective one of the plurality of distributions; and
processing, using a classifier, the generated vector to determine a likelihood that the memory component will fail within a target period of time.
14. The method of claim 13 , wherein classifying a first distribution of the plurality of distributions of threshold voltages comprises identifying a first tail in a quantile-quantile representation of the first distribution relative to a control distribution.
15. The method of claim 14 , wherein the control distribution is a normal distribution.
16. The method of claim 14 , wherein classifying the first distribution of the plurality of distributions of threshold voltages further comprises:
determining that the first tail corresponds to events for which a corresponding probability of occurrence is below a first cut-off value; and
treating the first tail as non-existent.
17. The method of claim 14 , wherein classifying the first distribution of the plurality of distributions of threshold voltages further comprises:
determining that the first tail corresponds to a deviation from the control distribution, wherein the deviation is below a second cut-off value; and
treating the first tail as non-existent.
18. The method of claim 14 , wherein classifying the first distribution of the plurality of distributions of threshold voltages comprises identifying the first tail as one of a top tail or a bottom tail, wherein the top tail corresponds to threshold voltages that are above a mean value for the first distribution, and wherein the bottom tail corresponds to threshold voltages that are below the mean value for the first distribution.
19. The method of claim 14 , wherein classifying a second distribution of the plurality of distributions of threshold voltages comprises identifying a shift of the second distribution relative to the control distribution.
20. The method of claim 13 , wherein obtaining the plurality of distributions of threshold voltages is responsive to an error-correction module detecting that the memory component has failed to correctly store data following a store operation.
21. A method comprising:
obtaining, by a processing device, a training input comprising a plurality of distributions of threshold voltages of a memory component, wherein each of the plurality of threshold voltages corresponds to one of a plurality of states, each state of the plurality of states corresponding to a value stored by memory cells of the memory component;
obtaining, by the processing device, a target output comprising a target likelihood of failure of the memory component within a target time interval;
processing the training input through a neural network model comprising a plurality of neurons to obtain, by the processing device, a training output, wherein the training output comprises a predicted likelihood of failure of the memory component within the target time interval;
determine, by the processing device, a difference between the predicted likelihood of failure and the target likelihood of failure; and
modify, by the processing device, parameters of the neural network model based on the determined difference.
22. The method of claim 21 , wherein the neural network model comprises a first subnetwork of neurons and a second subnetwork of neurons, wherein the first subnetwork is to process the training input and to output, a class for the plurality of distributions of threshold voltages, and wherein the second subnetwork is to determine, based on the output class, the predicted likelihood of failure.