# US 2025/0156094 A1 · Storage scheduling method, storage scheduling apparatus, storage scheduling device and computer-readable storage medium

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

- Number: US 2025/0156094 A1
- Document ID: US20250156094A1
- Status: Published application
- Kind: A1
- Application number: 19/021,785
- Filed: 2025-01-15
- Published: 2025-05-15
- Earliest claimed: 2022-09-29
- Assignees: Sanechips Technology Co., Ltd.
- Applicants: Sanechips Technology Co., Ltd.
- Inventors: Yingjie HU; Jinfeng Shi; Rongrong Shen; Yunlong Shi
- Web page: https://patentgrep.com/patent/US20250156094A1

## CPC

- G06F 3/0635: {by changing the path, e.g. traffic rerouting, path reconfiguration}
- G06F 9/546: {Message passing systems or structures, e.g. queues}
- G06F 3/0613: {in relation to throughput}
- G06F 3/0683: {Plurality of storage devices}
- G06F 12/0888: using selective caching, e.g. bypass
- G06F 13/1673: {using buffers}
- G06F 2209/548: Queue
- G06F 2212/1024: Latency reduction
- G06F 2212/1032: Reliability improvement, data loss prevention, degraded operation etc
- G06F 2212/1044: Space efficiency improvement
- G06F 2212/502: using adaptive policy
- Y02D 10/00: Energy efficient computing, e.g. low power processors, power management or thermal management
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- Y02D: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE

## Abstract

Disclosed are a storage scheduling method, a storage scheduling apparatus, a storage scheduling device and a computer-readable storage medium. The method includes: obtaining a current first amount of space used of an on-chip memory, and obtaining a current write bandwidth of an off-chip memory; determining a second queue from each first queue stored in the on-chip memory in response to that the first amount of space used is greater than a preset first threshold value and the write bandwidth is less than a preset second threshold value; the queue congestion degree is evaluated according to a preset congestion degree evaluation index; setting a storage state of the second queue to an off-chip state; and storing a target message in a memory indicated by a storage state of a queue where the target message is located.

## Claims

### Claim 1

A storage scheduling method, comprising:
obtaining a current first amount of space used of an on-chip memory, and obtaining a current write bandwidth of an off-chip memory;
determining a second queue from each first queue stored in the on-chip memory in response to that the first amount of space used is greater than a preset first threshold value and the write bandwidth is less than a preset second threshold value; wherein the first queue is a queue whose storage state is set to an on-chip state and only occupies an on-chip memory space among each queue stored in the on-chip memory, and the second queue is a queue with a first preset number of names at the front in response to sorting each first queue from high to low according to a queue congestion degree; the queue congestion degree is evaluated according to a preset congestion degree evaluation index;
setting a storage state of the second queue to an off-chip state; and
storing a target message in a memory indicated by a storage state of a queue where the target message is located, wherein the target message is a newly queued and unsaved message.

### Claim 2 (depends on 1)

The storage scheduling method according to claim 1, wherein the congestion evaluation index is an amount of space used of the queue, and the determining the second queue from each first queue stored in the on-chip memory comprises:
obtaining a second amount of space used respectively corresponding to each first queue; and
taking a queue with the first preset number of names at the front in response to sorting each first queue from large to small according to the second amount of space used as the second queue.

### Claim 3 (depends on 1)

The storage scheduling method according to claim 1, wherein after the obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory, the method further comprises:
determining a fourth queue from each third queue stored in the off-chip memory in response to that the write bandwidth is greater than or equal to a preset third threshold; wherein the third threshold is greater than or equal to the second threshold, the third queue is a queue whose storage state is set to the off-chip state among each queue stored in the off-chip memory, and the fourth queue is a queue with a second preset number of names at the front among each third queue in response to sorting from high to low according to the queue congestion degree; and
setting a storage state of the fourth queue to the on-chip state.

### Claim 4 (depends on 3)

The storage scheduling method according to claim 3, wherein the congestion evaluation index is an amount of space used of the queue, and the determining the fourth queue from each third queue stored in the off-chip memory comprises:
obtaining a third amount of space used corresponding to each third queue respectively; the third amount of space used is sum of the amount of space used of the third queue in the on-chip memory and the amount of space used of the third queue in the off-chip memory; and
taking a queue with the second preset number of names at the front in response to sorting each third queue from small to large according to the third amount of space used as the fourth queue.

### Claim 5 (depends on 1)

The storage scheduling method according to claim 1, wherein the obtaining the current write bandwidth of the off-chip memory comprises:
obtaining a currently recorded write bandwidth of the off-chip memory;
the storage scheduling method further comprises:
counting a first message length of a message written to the off-chip memory within the time window at an end of a time window; wherein a length of the time window is a preset duration;
dividing the first message length by the preset duration, and adopting a calculation result to update the currently recorded write bandwidth of the off-chip memory; and
moving the time window backward on a time axis by a preset step length, and returning to execute the counting total number of bytes of the message written to the off-chip memory within the time window at the end of the time window.

### Claim 6 (depends on 5)

The storage scheduling method according to claim 5, wherein the preset duration is a preset multiple of the preset step length, and the counting the first message length of the message written to the off-chip memory within the time window comprises:
accumulating a second message lengths corresponding to multiple small windows within the time window, and obtaining the first message length; wherein a length of the small window is the preset step length, and the second message length is a length of the message written to the off-chip memory within the small window obtained by counting at an end of a corresponding small window.

### Claim 7 (depends on 1)

The storage scheduling method according to claim 1, wherein the obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory comprises:
obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory in response to receiving the target message.

### Claim 8 (depends on 2)

The storage scheduling method according to claim 2, wherein the obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory comprises:
obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory in response to receiving the target message.

### Claim 9 (depends on 3)

The storage scheduling method according to claim 3, wherein the obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory comprises:
obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory in response to receiving the target message.

### Claim 10 (depends on 4)

The storage scheduling method according to claim 4, wherein the obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory comprises:
obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory in response to receiving the target message.

### Claim 11 (depends on 5)

The storage scheduling method according to claim 5, wherein the obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory comprises:
obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory in response to receiving the target message.

### Claim 12 (depends on 6)

The storage scheduling method according to claim 6, wherein the obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory comprises:
obtaining the current first amount of space used of the on-chip memory, and obtaining the current write bandwidth of the off-chip memory in response to receiving the target message.

### Claim 13

A storage scheduling apparatus, comprising:
an obtaining module, configured for obtaining a current first amount of space used of an on-chip memory and obtaining a current write bandwidth of an off-chip memory;
a determination module, configured for determining a second queue from each first queue stored in the on-chip memory in response to that the first amount of space used is greater than a preset first threshold value and the write bandwidth is less than a preset second threshold value; wherein the first queue is a queue whose storage state is set to an on-chip state and only occupies an on-chip memory space among each queue stored in the on-chip memory, and the second queue is a queue with a first preset number of names at the front in response to sorting each first queue from high to low according to a queue congestion degree; the queue congestion degree is evaluated according to a preset congestion degree evaluation index;
a setting module, configured for setting a storage state of the second queue to an off-chip state; and
a storage module, configured for storing a target message in a memory indicated by a storage state of a queue where the target message is located, wherein the target message is a newly queued and unsaved message.

### Claim 14 (depends on 1)

A storage scheduling device, comprising:
a memory;
a processor; and
a storage scheduling program stored in the memory and executable on the processor,
wherein the storage scheduling method according to claim 1 is implemented when the storage scheduling program is executed by the processor.

### Claim 15 (depends on 1)

A non-transitory computer-readable storage medium, wherein a storage scheduling program is stored on the computer-readable storage medium, and the storage scheduling method according to claim 1 is implemented when the storage scheduling program is executed by the processor.
