# US 2023/0418505 A1 · Systems and methods of forecasting temperatures of storage objects using machine learning

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2023/0418505 A1
- Document ID: US20230418505A1
- Status: Published application
- Kind: A1
- Application number: 17/847,449
- Filed: 2022-06-23
- Published: 2023-12-28
- Earliest claimed: 2022-06-23
- Assignees: Dell Products L.P.
- Applicants: Dell Products L.P.
- Inventors: Shaul Dar; Ramakanth Kanagovi; Vamsi K. Vankamamidi; Guhesh Swaminathan; Swati Smita Sitha
- Web page: https://patentgrep.com/patent/US20230418505A1

## CPC

- G06F 3/0653: {Monitoring storage devices or systems}
- G06F 3/0685: {Hybrid storage combining heterogeneous device types, e.g. hierarchical storage, hybrid arrays}
- G06F 3/0604: {Improving or facilitating administration, e.g. storage management}
- G06F 3/0608: {Saving storage space on storage systems}
- G06F 3/061: {Improving I/O performance}
- G06F 3/0649: {Lifecycle management}
- G06F 3/0679: {Non-volatile semiconductor memory device, e.g. flash memory, one time programmable memory [OTP]}
- G06F 11/3034: {where the computing system component is a storage system, e.g. DASD based or network based (digital input from or digital output to record carriers G06F3/06; digital recording or reproducing G11B20/18; for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS], H04L67/1097)}
- G06F 11/3058: {Monitoring arrangements for monitoring environmental properties or parameters of the computing system or of the computing system component, e.g. monitoring of power, currents, temperature, humidity, position, vibrations (thermal management in cooling arrangements of a computing system G06F1/206)}
- G06F 11/3409: {for performance assessment}
- G06N 20/00: Machine learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Techniques for forecasting temperatures of storage objects in a storage system using machine learning (ML). The techniques can include forecasting at least one temperature of a storage object using at least one ML model, modifying storage of the storage object based on the at least one temperature of the storage object, and, having modified storage of the storage object, obtaining at least one performance metric associated with the storage object. The techniques can further include, based on the performance metric(s), varying a frequency of forecasting the at least one temperature of the storage object, retraining the at least one ML model used in forecasting the at least one temperature, and/or adjusting at least one operational parameter of the system. The techniques provide increased accuracy over known statistical approaches to forecasting temperatures of storage objects, leading to increased performance gains in terms of IO latency, IO operations per second, and bandwidth.

## Claims

### Claim 1

A method comprising:
forecasting at least one temperature of a storage object using at least one machine learning (ML) model;
modifying storage of the storage object based on the at least one forecasted temperature of the storage object;
having modified storage of the storage object, obtaining at least one performance metric associated with the storage object; and
based on the at least one performance metric, performing one or more of varying a frequency of forecasting the at least one temperature of the storage object, retraining the at least one ML model used in forecasting the at least one temperature, and adjusting at least one operational parameter of the storage system.

### Claim 2 (depends on 1)

The method of claim 1 wherein forecasting at least one temperature of a storage object includes forecasting a plurality of temperatures of the storage object using a plurality of ML models, respectively.

### Claim 3 (depends on 2)

The method of claim 2 further comprising:
producing a plurality of target variables by the plurality of ML models, respectively, the plurality of target variables corresponding to the plurality of temperatures of the storage object, respectively,
wherein the plurality of target variables further correspond to a plurality of temperature definitions, respectively, each temperature definition defining a different manner of expressing a temperature of a storage object.

### Claim 4 (depends on 2)

The method of claim 2 wherein one or more input/output (TO) operations are performed on the storage object over a plurality of first time intervals, and wherein the method further comprises:
for each IO operation performed on the storage device during each first time interval, updating one or more in-memory counters.

### Claim 5 (depends on 4)

The method of claim 4 further comprising:
at an end of each first time interval, generating a plurality of features relevant to the at least one temperature of the storage object by aggregating the updated in-memory counters,
wherein the plurality of features include at least a number of IO operations per second (IOPS) performed on the storage object, a number of read IO operations performed on the storage object, and a number of write IO operations performed on the storage object.

### Claim 6 (depends on 5)

The method of claim 5 wherein the plurality of features generated at the end of each first time interval forms an input dataset for the plurality of ML models, and wherein the method further comprises:
assigning each input dataset to one of a plurality of data buckets,
wherein each data bucket corresponds to a different range of temperatures of a storage object.

### Claim 7 (depends on 6)

The method of claim 6 further comprising:
for each data bucket, training the plurality of ML models using at least some input datasets assigned to the data bucket.

### Claim 8 (depends on 6)

The method of claim 6 wherein forecasting a plurality of temperatures of the storage object using a plurality of ML models, respectively, includes, for each data bucket at an end of each of a plurality of second time intervals, (i) applying input datasets assigned to the data bucket to the plurality of ML models, and (ii) producing a plurality of target variables by the plurality of ML models, respectively, the plurality of target variables corresponding to the plurality of temperatures of the storage object, respectively,
wherein the plurality of target variables further correspond to a plurality of temperature definitions, each temperature definition defining a different manner of expressing a temperature of a storage object.

### Claim 9 (depends on 8)

The method of claim 8 wherein modifying storage of the storage object includes applying one or more storage optimization policies based on the forecasted plurality of temperatures of the storage object.

### Claim 10 (depends on 9)

The method of claim 9 wherein obtaining at least one performance metric associated with the storage object includes evaluating the at least one performance metric resulting from application of the one or more storage optimization policies.

### Claim 11

A system comprising:
a memory; and
processing circuitry configured to execute program instructions out of the memory to:
forecast at least one temperature of a storage object using at least one machine learning (ML) model;
modify storage of the storage object based on the at least one forecasted temperature of the storage object;
having modified storage of the storage object, obtain at least one performance metric associated with the storage object; and
based on the at least one performance metric, perform one or more of varying a frequency of forecasting the at least one temperature of the storage object, retraining the at least one ML model used in forecasting the at least one temperature, and adjusting at least one operational parameter of the storage system.

### Claim 12 (depends on 11)

The system of claim 11 wherein the processing circuitry is further configured to execute the program instructions out of the memory to forecast a plurality of temperatures of the storage object using a plurality of ML models, respectively.

### Claim 13 (depends on 12)

The system of claim 12 wherein the memory includes a plurality of counters, wherein one or more input/output (IO) operations are performed on the storage object over a plurality of first time intervals, and wherein the processing circuitry is further configured to execute the program instructions out of the memory, for each IO operation performed on the storage device during each first time interval, to update at least some of the plurality of counters.

### Claim 14 (depends on 13)

The system of claim 13 wherein the processing circuitry is further configured to execute the program instructions out of the memory, at an end of each first time interval, to generate a plurality of features relevant to the at least one temperature of the storage object by aggregating the updated plurality of counters, wherein the plurality of features include at least a number of IO operations per second (IOPS) performed on the storage object, a number of read IO operations performed on the storage object, and a number of write IO operations performed on the storage object.

### Claim 15 (depends on 14)

The system of claim 14 wherein the plurality of features further include at least some of:
an average time between successive arrivals (“average interarrival time”) of the IO operations performed on the storage object;
an average interarrival time of the read IO operations performed on the storage object;
an average interarrival time of the write IO operations performed on the storage object;
an average difference between start logical block addresses (“average delta LBA”) of successive IO operations performed on the storage object;
an average delta LBA of successive read IO operations performed on the storage object;
an average delta LBA of successive write IO operations performed on the storage object;
a percentage of successive pairs of IO operations performed on the storage object that include two read IO operations;
a percentage of successive pairs of IO operations performed on the storage object that include a read IO operation followed by a write IO operation;
a percentage of successive pairs of IO operations performed on the storage object that include a write IO operation followed by a read IO operation;
a percentage of successive pairs of IO operations performed on the storage object that include two write IO operations;
a percentage of successive pairs of read IO operations performed on the storage object in which a start LBA of a second read IO operation immediately follows an end LBA of a first read IO operation; and
a percentage of successive pairs of write IO operations performed on the storage object in which a start LBA of a second write IO operation immediately follows an end LBA of a first write IO operation.

### Claim 16 (depends on 14)

The system of claim 14 wherein the plurality of features generated at the end of each first time interval forms an input dataset for the plurality of ML models, and wherein the processing circuitry is further configured to execute the program instructions out of the memory to assign each input dataset to one of a plurality of data buckets,
wherein each data bucket corresponds to a different range of temperatures of a storage object.

### Claim 17 (depends on 16)

The system of claim 16 wherein the processing circuitry is further configured to execute the program instructions out of the memory, for each data bucket at an end of each of a plurality of second time intervals, to apply input datasets assigned to the data bucket to the plurality of ML models, and to produce a plurality of target variables by the plurality of ML models, respectively, the plurality of target variables corresponding to the plurality of temperatures of the storage object, respectively,
wherein the plurality of target variables further correspond to a plurality of temperature definitions, each temperature definition defining a different manner of expressing a temperature of a storage object.

### Claim 18 (depends on 17)

The system of claim 17 wherein the processing circuitry is further configured to execute the program instructions out of the memory to apply one or more storage optimization policies based on the forecasted plurality of temperatures of the storage object, and to evaluate the at least one performance metric resulting from application of the one or more storage optimization policies.

### Claim 19

A computer program product including a set of non-transitory, computer-readable media having instructions that, when executed by processing circuitry, cause the processing circuitry to perform a method comprising:
forecasting at least one temperature of a storage object using at least one machine learning (ML) model;
modifying storage of the storage object based on the at least one forecasted temperature of the storage object;
having modified storage of the storage object, obtaining at least one performance metric associated with the storage object; and
based on the at least one performance metric, performing one or more of varying a frequency of forecasting the at least one temperature of the storage object, retraining the at least one ML model used in forecasting the at least one temperature, and adjusting at least one operational parameter of the storage system.

### Claim 20 (depends on 19)

The computer program product of claim 19 wherein forecasting at least one temperature of a storage object includes forecasting a plurality of temperatures of the storage object using a plurality of ML models, respectively, and wherein the method further comprises:
producing a plurality of target variables by the plurality of ML models, respectively, the plurality of target variables corresponding to the plurality of temperatures of the storage object, respectively,
wherein the plurality of target variables further correspond to a plurality of temperature definitions, respectively, each temperature definition defining a different manner of expressing a temperature of a storage object.
