Storage device feature extraction optimization
Systems and methods for compacting and anonymizing telemetry data in a storage system. A controller of a storage device may generate telemetry data based on collected features indicative of the performance of the storage device. The controller may store the telemetry data in the telemetry memory of the storage device. The controller may then transform the telemetry data into transformed telemetry data based on a dimension reduction algorithm, and transmit the transformed telemetry data to the host device.
1. A system, comprising:
a memory device in a storage device configured to store and retrieve data in response to a host device;
a telemetry memory in the storage device configured to store telemetry data; and
a controller in the storage device configured to:
generate the telemetry data based on original collected features indicative of performance of the storage device;
store the telemetry data in the telemetry memory of the storage device;
compact the telemetry data into transformed telemetry data based on a dimension reduction algorithm, wherein the transformed telemetry data comprises a smaller number of transformed collected features than a number of the original collected features; and
transmit the transformed telemetry data to the host device.
2. The system of claim 1 , wherein the dimension reduction algorithm is a linear algorithm.
3. The system of claim 1 , wherein the dimension reduction algorithm is one of a Principal Component Analysis (PCA) algorithm, a Linear Discriminant Analysis (LDA) algorithm, a Generalized Discriminant Analysis (GDA) algorithm, and a kernel PCA algorithm.
4. The system of claim 1 , wherein the dimension reduction algorithm is a non-linear algorithm.
5. The system of claim 1 , wherein the controller is further configured to:
receive a selection of the original collected features from the host device; and
transform selected original collected features of the telemetry data corresponding to the selection of the original collected features into the transformed telemetry data based on the dimension reduction algorithm to compact the telemetry data.
6. The system of claim 5 , wherein the selection of the original collected features comprises the number of transformed collected features in the transformed telemetry data.
7. The system of claim 1 , wherein the controller is further configured to:
aggregate, after compacting the telemetry data, to anonymize an identity of the storage device to form the transformed telemetry data transmitted to the host device.
8. The system of claim 1 , wherein the controller is further configured to:
receive a telemetry request from the host device; and
transmit, in response to receiving the telemetry request, the transformed telemetry data to the host device.
9. The system of claim 1 , wherein the controller is further configured to:
receive operational parameters for the storage device, wherein the operational parameters are based on the transformed telemetry data further processed by the host device.
10. A computer-implemented method, comprising:
generating, by a controller of a storage device, telemetry data based on original collected features indicative of performance of the storage device;
storing the telemetry data in a telemetry memory of the storage device;
compacting, by the controller of the storage device, the telemetry data into transformed telemetry data based on a dimension reduction algorithm, wherein the transformed telemetry data comprises a smaller number of transformed collected features than a number of the original collected features; and
transmitting, by the controller of the storage device, the transformed telemetry data to a host device.
11. The computer-implemented method of claim 10 , wherein the dimension reduction algorithm is a linear algorithm.
12. The computer-implemented method of claim 10 , wherein the dimension reduction algorithm is one of a Principal Component Analysis (PCA) algorithm, a Linear Discriminant Analysis (LDA) algorithm, a Generalized Discriminant Analysis (GDA) algorithm, and a kernel PCA algorithm.
13. The computer-implemented method of claim 10 , wherein the dimension reduction algorithm is a non-linear algorithm.
14. The computer-implemented method of claim 10 , further comprising:
receiving a selection of the original collected features from the host device; and
transforming, by the controller of the storage device, selected original collected features of the telemetry data corresponding to the selection of the original collected features into the transformed telemetry data based on the dimension reduction algorithm to compact the telemetry data.
15. The computer-implemented method of claim 14 , wherein the selection of the original collected features comprises the number of transformed collected features in the transformed telemetry data.
16. The computer-implemented method of claim 10 , further comprising:
aggregating, after compacting the telemetry data, to anonymize an identity of the storage device to form the transformed telemetry data.
17. The computer-implemented method of claim 10 , wherein transmitting the transformed telemetry data to the host device comprises:
receiving, at the controller of the storage device, a telemetry request from the host device; and
transmitting, by the controller of the storage device in response to receiving the telemetry request, the transformed telemetry data to the host device.
18. The computer-implemented method of claim 10 , further comprising:
receiving, at the controller of the storage device, operational parameters for the storage device, wherein the operational parameters are based on the transformed telemetry data further processed by the host device.
19. The computer-implemented method of claim 18 , wherein the operational parameters include at least one of:
partition sizes of mixed level memory devices in the storage device;
memory device wear balancing parameters; and
programming rate parameters of memory devices in the storage device.
20. A system, comprising:
a memory device in a storage device configured to store and retrieve data in response to a host device;
a telemetry memory in the storage device configured to store telemetry data;
means, in the storage device, for generating the telemetry data based on original collected features indicative of performance of the storage device;
means for storing the telemetry data in the telemetry memory of the storage device;
means, in the storage device, for compacting the telemetry data into transformed telemetry data based on a dimension reduction algorithm, wherein the transformed telemetry data comprises a smaller number of transformed collected features than a number of the original collected features; and
means, in the storage device, for transmitting the transformed telemetry data to the host device.