Adaptable online breakpoint detection over I/O trace time series via deep neural network autoencoders re-parameterization

Patent metadata

Field	Value
Publication number	US-11550686-B2
Application number	US-201916402110-A
Country	US
Kind code	B2
Filing date	May 2, 2019
Priority date	May 2, 2019
Publication date	Jan 10, 2023
Grant date	Jan 10, 2023

Abstract

One example method includes accessing I/O traces, generating parameters based on the I/O traces, and defining an autoencoder deep neural network, training the autoencoder deep neural network using the parameters, collecting and storing new I/O traces, computing an encoded features difference series using the new I/O traces, detecting breakpoints in the encoded features difference series, evaluating a utility of the breakpoints, and performing an action based on the breakpoint utility evaluation.

First claim

What is claimed is:

1. A method, comprising:
obtaining a multi-channel time series comprising I/O traces from a computational appliance;
defining parameters based on the I/O traces, and generating an autoencoder deep neural network;
training the autoencoder deep neural network using the parameters;
collecting and storing new I/O traces samples, wherein the I/O traces and/or the new I/O traces comprise multi-channel time series data;
obtaining encoded features resulting from the autoencoder deep neural network using samples of the new I/O traces as input;
computing an encoded features difference series using the encoded features of the new I/O traces samples;
detecting breakpoints in the encoded features difference series;
evaluating a utility of the breakpoints, and the evaluating of the utility of the breakpoints results in a weighted composition of a true-positive factor and a false positive factor, where:
the true-positive factor=a sum of distances from each ground truth point to a closest discovered breakpoint, divided by a length of the multi-channel time series; and
the false-positive factor=a sum of distances from each discovered breakpoint to a closest ground truth point, divided by the length of the multi-channel time series; and

performing an operational action involving the computational appliance based on the weighted composition obtained from the evaluating of the utility of the breakpoints.

2. The method as recited in claim 1 , wherein the computational appliance comprises a storage array or hyperconverged infrastructure appliance, and wherein obtaining the I/O traces comprises collecting and storing data from one or more logging systems and/or monitoring systems.

3. The method as recited in claim 1 , wherein when an unexpected breakpoint is detected that leads to an undesired operational state of an underlying system, the operational action comprises repairing the system.

4. The method as recited in claim 1 , wherein detecting breakpoints in the encoded features difference series comprises one-dimensional local maxima detection.

5. The method as recited in claim 1 , further comprising using the multi-channel time series data to characterize an operational state of an underlying system.

6. The method as recited in claim 2 , further comprising using the identified breakpoints as a basis for implementing cache policy optimization.

7. The method as recited in claim 1 , wherein the I/O traces and new I/O traces are in the same domain as a set of target data.

8. The method as recited in claim 1 , wherein defining parameters is performed again after being triggered by evaluation of the utility of the breakpoints.

9. The method as recited in claim 8 , further comprising retraining the autoencoder deep neural network.

10. The method as recited in claim 2 , wherein the operational actions performed based on the breakpoint utility evaluation comprise any one or more of: automatic tuning of a storage appliance to optimally satisfy application requirements; generating a data prefetching configuration; performing an adaptation of a cache policy; and performing automated anomaly detection.

11. A non-transitory storage medium having stored therein computer-executable instructions which are executable by one or more hardware processors, to perform operations comprising:
obtaining a multi-channel time series comprising I/O traces from a computational appliance;
defining parameters based on the I/O traces, and generating an autoencoder deep neural network;
training the autoencoder deep neural network using the parameters;
collecting and storing new I/O traces samples, wherein the I/O traces and/or the new I/O traces comprise multi-channel time series data;
obtaining encoded features resulting from the autoencoder deep neural network using samples of the new I/O traces as input;
computing an encoded features difference series using the encoded features of the new I/O traces samples;
detecting breakpoints in the encoded features difference series;
evaluating a utility of the breakpoints, and the evaluating of the utility of the breakpoints results in a weighted composition of a true-positive factor and a false positive factor, where:
the true-positive factor=a sum of distances from each ground truth point to a closest discovered breakpoint, divided by a length of the multi-channel time series; and
the false-positive factor=a sum of distances from each discovered breakpoint to a closest ground truth point, divided by the length of the multi-channel time series; and

performing an operational action involving the computational appliance based on the weighted composition obtained from the evaluating of the utility of the breakpoints.

12. The non-transitory storage medium as recited in claim 11 , wherein the computational appliance comprises a storage array or hyperconverged infrastructure appliance, and wherein obtaining the I/O traces comprises collecting and storing data from one or more logging systems and/or monitoring systems.

13. The non-transitory storage medium as recited in claim 11 , wherein when an unexpected breakpoint is detected that leads to an undesired operational state of an underlying system, the operational action comprises repairing the system.

14. The non-transitory storage medium as recited in claim 11 , wherein detecting breakpoints in the encoded features difference series comprises one-dimensional local maxima detection.

15. The non-transitory storage medium as recited in claim 11 , wherein the operations further comprise using the multi-channel time series data to characterize an operational state of an underlying system.

16. The non-transitory storage medium as recited in claim 11 , further comprising using the identified breakpoints as a basis for implementing cache policy optimization.

17. The non-transitory storage medium as recited in claim 11 , wherein the I/O traces and new I/O traces are in the same domain as a set of target data.

18. The non-transitory storage medium as recited in claim 11 , wherein defining parameters is performed again after being triggered by evaluation of the utility of the breakpoints.

19. The non-transitory storage medium as recited in claim 18 , wherein the operations further comprise retraining the autoencoder deep neural network.

20. The non-transitory storage medium as recited in claim 11 , wherein the operational actions performed based on the breakpoint utility evaluation comprise any one or more of: automatic tuning of a storage appliance to optimally satisfy application requirements; generating a data prefetching configuration; performing an adaptation of a cache policy; and performing automated anomaly detection.

Assignees

• Emc Ip Holding Co Llc

Inventors

• Sousa Eduardo Vera
• Gottin Vinicius Michel
• Salas Percy Rivera

Classifications

• G06F11/3409

  for performance assessment · CPC title

• G06N3/084 Primary

  Backpropagation, e.g. using gradient descent · CPC title

• G06F11/3055

  Monitoring arrangements for monitoring the status of the computing system or of the computing system component, e.g. monitoring if the computing system is on, off, available, not available (error or fault processing without redundancy G06F11/0703; error detection or correction by redundancy in data representation G06F11/08; error detection or correction of the data by redundancy in operations G06F11/14; error detection or correction by redundancy in hardware G06F11/16) · CPC title

• G06F11/3485

  for I/O devices · CPC title

• G06N3/08

  Learning methods · CPC title