# US 2024/0004716 A1 · Method and Apparatus for Dynamic CPU Allocation Based on Feature Usage Profiles of Shared Global Memory

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

- Number: US 2024/0004716 A1
- Document ID: US20240004716A1
- Status: Published application
- Kind: A1
- Application number: 17/856,960
- Filed: 2022-07-02
- Published: 2024-01-04
- Earliest claimed: 2022-07-02
- Assignees: Dell Products L.P.
- Applicants: Dell Products, L.P.
- Inventors: John Creed; Owen Martin
- Web page: https://patentgrep.com/patent/US20240004716A1

## CPC

- G06F 9/5038: {considering the execution order of a plurality of tasks, e.g. taking priority or time dependency constraints into consideration (scheduling strategies G06F9/4881 and subgroups)}
- G06F 9/3836: {Instruction issuing, e.g. dynamic instruction scheduling or out of order instruction execution}
- G06F 9/3851: {from multiple instruction streams, e.g. multistreaming}
- G06F 9/3877: {using a secondary processor, e.g. coprocessor (peripheral processor G06F13/12)}
- G06F 9/505: {considering the load}
- G06F 9/5061: {Partitioning or combining of resources}
- G06F 9/544: {Buffers; Shared memory; Pipes}
- G06F 2209/508: Monitor
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Feature usage profiles of shared global memory are monitored and the profiles are used to allocate computational resources (CPU cores) between thread groups. A global memory monitor is used to monitor global memory metadata to determine the amount of space currently being consumed by a particular feature or service, and dynamically adjusts the amount of computational resources allocated to the thread group responsible for processing IO operations related to the feature or service. By exploiting an inversely proportional relationship between allocated computational resources and usage of shared global memory, it is possible to monitor feature usage profiles of shared global memory to dynamically allocate computational resources to thread groups. By looking at anomalies in usage of slots of shared global memory, and in a targeted way speeding up the processes working on such slots, it is possible to improve overall utilization of shared global memory.

## Claims

### Claim 1

A non-transitory tangible computer readable storage medium having stored thereon a computer program for dynamic CPU allocation based on feature usage profiles of shared global memory, the computer program including a set of instructions which, when executed by a computer, cause the computer to perform a method comprising the steps of:
waiting for expiration of a monitoring interval;
upon expiration of the monitoring interval, determining by a global memory monitor a feature usage profile of shared global memory for the monitoring interval, the feature usage profile of shared global memory being based on a number of slots of shared global memory containing data for processing by each of a plurality of thread groups at the end of the monitoring interval, each thread group being responsible for processing a respective type of memory access operation;
allocating computational resources from a dynamically allocated portion of computational resources to the plurality of thread groups based on the feature usage profile of shared global memory; and
dynamically allocating the dynamically allocated portion of computational resources over time according to evolving feature usage profiles of shared global memory by iterating the steps of waiting for expiration of the monitoring interval, determining the feature usage profile of shred global memory, and allocating computational resources after each subsequent monitoring interval.

### Claim 2 (depends on 1)

The non-transitory tangible computer readable storage medium of claim 1, wherein:
a first of the plurality of thread groups is responsible for processing read hit memory access operations;
a second of the plurality of thread groups is responsible for processing read miss memory access operations;
a third of the plurality of thread groups is responsible for processing write destage memory access operations; and
a fourth of the plurality of thread groups is responsible for processing local replication (LREP) write memory access operations.

### Claim 3 (depends on 1)

The non-transitory tangible computer readable storage medium of claim 1, further comprising providing a fixed minimum allocation of the computational resources to each of the thread groups that are not part of the dynamically allocated portion of the computational resources.

### Claim 4 (depends on 1)

The non-transitory tangible computer readable storage medium of claim 1, wherein the step of allocating computational resources from a dynamically allocated portion of computational resources to the plurality of thread groups comprises comparing the number of slots of shared global memory containing data for processing by each of a plurality of thread groups with a plurality of thresholds.

### Claim 5 (depends on 4)

The non-transitory tangible computer readable storage medium of claim 4, wherein the plurality of thresholds comprise respective promotion and demotion thresholds for each of the respective thread groups.

### Claim 6 (depends on 5)

The non-transitory tangible computer readable storage medium of claim 5, wherein if a number of slots of shared global memory containing data for processing by a first of the thread groups is above a first promotion threshold for the first thread group, the step of allocating computational resources comprises increasing a first portion of the dynamically allocated computational resources that are assigned to the first thread group during an upcoming monitoring interval.

### Claim 7 (depends on 5)

The non-transitory tangible computer readable storage medium of claim 5, wherein if a number of slots of shared global memory containing data for processing by a second of the thread groups is below a first demotion threshold for the second thread group, the step of allocating computational resources comprises decreasing a portion of the dynamically allocated computational resources that are assigned to the second thread group during an upcoming monitoring interval.

### Claim 8 (depends on 1)

The non-transitory tangible computer readable storage medium of claim 1, wherein the step of allocating computational resources from a dynamically allocated portion of computational resources to the plurality of thread groups comprises comparing the number of slots of shared global memory containing data for processing by each of a plurality of thread groups a number of slots of shared global memory containing data for processing by each of the plurality of thread groups during a prior monitoring interval.

### Claim 9 (depends on 8)

The non-transitory tangible computer readable storage medium of claim 8, wherein if a number of slots of shared global memory containing data for processing by a first of the thread groups has increased since the prior monitoring interval, the step of allocating computational resources comprises increasing a first portion of the dynamically allocated computational resources that are assigned to the first thread group during an upcoming monitoring interval.

### Claim 10 (depends on 8)

The non-transitory tangible computer readable storage medium of claim 8, wherein if a number of slots of shared global memory containing data for processing by a second of the thread groups has decreased since the prior monitoring interval, the step of allocating computational resources comprises decreasing a portion of the dynamically allocated computational resources that are assigned to the second thread group during an upcoming monitoring interval.

### Claim 11

A method of dynamic CPU allocation based on feature usage profiles of shared global memory, comprising:
waiting for expiration of a monitoring interval;
upon expiration of the monitoring interval, determining by a global memory monitor a feature usage profile of shared global memory for the monitoring interval, the feature usage profile of shared global memory being based on a number of slots of shared global memory containing data for processing by each of a plurality of thread groups at the end of the monitoring interval, each thread group being responsible for processing a respective type of memory access operation;
allocating computational resources from a dynamically allocated portion of computational resources to the plurality of thread groups based on the feature usage profile of shared global memory; and
dynamically allocating the dynamically allocated portion of computational resources over time according to evolving feature usage profiles of shared global memory by iterating the steps of waiting for expiration of the monitoring interval, determining the feature usage profile of shred global memory, and allocating computational resources after each subsequent monitoring interval.

### Claim 12 (depends on 11)

The method of claim 11, wherein:
a first of the plurality of thread groups is responsible for processing read hit memory access operations;
a second of the plurality of thread groups is responsible for processing read miss memory access operations;
a third of the plurality of thread groups is responsible for processing write destage memory access operations; and
a fourth of the plurality of thread groups is responsible for processing local replication (LREP) write memory access operations.

### Claim 13 (depends on 11)

The method of claim 11, further comprising providing a fixed minimum allocation of the computational resources to each of the thread groups that are not part of the dynamically allocated portion of the computational resources.

### Claim 14 (depends on 11)

The method of claim 11, wherein the step of allocating computational resources from a dynamically allocated portion of computational resources to the plurality of thread groups comprises comparing the number of slots of shared global memory containing data for processing by each of a plurality of thread groups with a plurality of thresholds.

### Claim 15 (depends on 14)

The method of claim 14, wherein the plurality of thresholds comprise respective promotion and demotion thresholds for each of the respective thread groups.

### Claim 16 (depends on 15)

The method of claim 15, wherein if a number of slots of shared global memory containing data for processing by a first of the thread groups is above a first promotion threshold for the first thread group, the step of allocating computational resources comprises increasing a first portion of the dynamically allocated computational resources that are assigned to the first thread group during an upcoming monitoring interval.

### Claim 17 (depends on 15)

The method of claim 15, wherein if a number of slots of shared global memory containing data for processing by a second of the thread groups is below a first demotion threshold for the second thread group, the step of allocating computational resources comprises decreasing a portion of the dynamically allocated computational resources that are assigned to the second thread group during an upcoming monitoring interval.

### Claim 18 (depends on 11)

The method of claim 11, wherein the step of allocating computational resources from a dynamically allocated portion of computational resources to the plurality of thread groups comprises comparing the number of slots of shared global memory containing data for processing by each of a plurality of thread groups a number of slots of shared global memory containing data for processing by each of the plurality of thread groups during a prior monitoring interval.

### Claim 19 (depends on 18)

The method of claim 18, wherein if a number of slots of shared global memory containing data for processing by a first of the thread groups has increased since the prior monitoring interval, the step of allocating computational resources comprises increasing a first portion of the dynamically allocated computational resources that are assigned to the first thread group during an upcoming monitoring interval.

### Claim 20 (depends on 18)

The method of claim 18, wherein if a number of slots of shared global memory containing data for processing by a second of the thread groups has decreased since the prior monitoring interval, the step of allocating computational resources comprises decreasing a portion of the dynamically allocated computational resources that are assigned to the second thread group during an upcoming monitoring interval.
