# US 2026/0186925 A1 · Distributed computing system and fault handling method for distributed computing system

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

- Number: US 2026/0186925 A1
- Document ID: US20260186925A1
- Status: Published application
- Kind: A1
- Application number: 19/548,164
- Filed: 2026-02-24
- Published: 2026-07-02
- Earliest claimed: 2023-08-25
- Applicants: Huawei Technologies Co., Ltd.
- Inventors: Huan Chen
- Web page: https://patentgrep.com/patent/US20260186925A1

## CPC

- G06F 11/2025: {using centralised failover control functionality}
- G06F 11/2028: {eliminating a faulty processor or activating a spare}
- G06F 11/2041: {with more than one idle spare processing component}
- H04Q 11/0062: {Network aspects}
- G06F 2201/805: Real-time
- H04Q 2011/0081: {Fault tolerance; Redundancy; Recovery; Reconfigurability}
- H04Q 2213/13166: Fault prevention
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- H04Q: SELECTING (switches, relays, selectors H01H; wireless communication networks H04W)

## Abstract

A distributed computing system and a fault handling method for a distributed computing system. In the distributed computing system, both a computing node group and a backup node access a switching device, and when detecting that a fault occurs in the computing node group, a management node in the distributed computing system may configure the switching device to establish a data exchange path between the backup node and the computing node group.

## Claims

### Claim at position 1

1-18. (canceled)

### Claim 19

A distributed computing system, comprising:
a management node;
a switching device;
a computing node group; and
a backup node,
wherein the switching device is physically connected to the computing node group and the backup node separately, each computing node in the computing node group comprises a plurality of accelerators, and the backup node comprises a plurality of backup accelerators, and
the management node is configured to:
detect that a fault occurs in the computing node group, and determine a faulty device in which the fault occurs, wherein the faulty device is a faulty computing node or a faulty accelerator in the computing node group, and
configure the switching device to establish a data exchange path between the backup node and the computing node group via the switching device, wherein the data exchange path enables data exchange between the backup node and the computing node group; and
the backup node is configured to replace the faulty device to execute a computing task.

### Claim 20 (depends on 19)

The system according to claim 19, wherein the switching device is an optical switch, the optical switch comprises a plurality of optical fiber ports, and the plurality of optical fiber ports are configured to access the backup node and the computing node group.

### Claim 21 (depends on 20)

The system according to claim 20, wherein the management node is configured to:
create an optical path between a first optical fiber port and a second optical fiber port on the optical switch, to establish an optical path between the backup node and the computing node group via the optical switch, wherein
the optical path switches an optical signal between the backup node and the computing node group, the first optical fiber port is an optical fiber port for accessing the backup node, the second optical fiber port is an optical fiber port for accessing the computing node group, and the optical signal carries data corresponding to the computing task.

### Claim 22 (depends on 20)

The system according to claim 20, wherein computing nodes in the computing node group are connected to each other via a first switching component, at least one port of the first switching component is configured to access the optical switch, and the first switching component is a switch or a switching chip.

### Claim 23 (depends on 20)

The system according to claim 20, wherein the backup accelerators in the backup node are connected to each other via a second switching component, at least one port of the second switching component is configured to access the optical switch, and the second switching component is a switch or a switching chip.

### Claim 24 (depends on 19)

The system according to claim 19, wherein the management node is configured to perform at least one of the following:
configuring the switching device, when the faulty device comprises a first quantity of faulty computing nodes, to establish a first data exchange path between a first quantity of backup nodes and at least one first node via the switching device, wherein the first data exchange path enables data exchange between the first quantity of backup nodes and the at least one first node, and the first node is a node other than the faulty computing node in the computing node group;
configuring the switching device, when the faulty device comprises a second quantity of faulty accelerators, to establish a second data exchange path between a second quantity of backup accelerators in at least one backup node and at least one first accelerator via the switching device, wherein the second data exchange path enables data exchange between the second quantity of backup accelerators and the at least one first accelerator, and the first accelerator is an accelerator other than the faulty accelerator in the computing node group; or
if the faulty device comprises a third quantity of faulty accelerators, determining a fourth quantity of nodes to which the third quantity of faulty accelerators belong, and configuring the switching device, to establish a third data exchange path between a fourth quantity of backup nodes and the at least one first node via the switching device, wherein the third data exchange path enables data exchange between the fourth quantity of backup nodes and the at least one first node.

### Claim 25 (depends on 19)

The system according to claim 19, wherein data exchange between computing nodes in the computing node group is performed based on a first bandwidth; and
the backup node is configured to perform data exchange with the computing node group through the data exchange path based on the first bandwidth, replacing the faulty device to execute the computing task.

### Claim 26 (depends on 25)

The system according to claim 25, wherein the system comprises a plurality of computing node groups, each computing node group comprises a same quantity of computing nodes, data exchange between the computing node groups is performed based on a second bandwidth, and the first bandwidth is greater than or equal to the second bandwidth.

### Claim 27 (depends on 26)

The system according to claim 26, wherein bandwidths for data exchange between computing nodes in different computing node groups are different, and the bandwidths for data exchange between computing nodes in different computing node groups are all greater than or equal to the second bandwidth.

### Claim 28 (depends on 25)

The system according to claim 25, wherein each computing node in the computing node group comprises a same quantity of accelerators, data exchange between the accelerators is performed based on a third bandwidth, and the first bandwidth is less than or equal to the third bandwidth.

### Claim 29 (depends on 28)

The system according to claim 28, wherein data exchange between the backup accelerators in the backup node is performed based on the third bandwidth.

### Claim 30 (depends on 19)

The system according to claim 19, wherein the distributed computing system is configured to execute a distributed parallel training task of an AI model, and the faulty device is configured to execute a subtask of the distributed parallel training task.

### Claim 31

A fault handling method for a distributed computing system, applied to a management node in the distributed computing system, wherein the system further comprises a switching device, a computing node group, and a backup node, the switching device is physically connected to the computing node group and the backup node separately, each computing node in the computing node group comprises a plurality of accelerators, the backup node comprises a plurality of backup accelerators, and the method comprises:
detecting that a fault occurs in the computing node group, and determining a faulty device in which the fault occurs, wherein the faulty device is a faulty computing node or a faulty accelerator in the computing node group; and
configuring the switching device, to establish a data exchange path between the backup node and the computing node group via the switching device, wherein the data exchange path enables data exchange between the backup node and the computing node group, to cause the backup node to replace the faulty device to execute a computing task.

### Claim 32 (depends on 31)

The method according to claim 31, wherein the switching device is an optical switch, the optical switch comprises a plurality of optical fiber ports, the plurality of optical fiber ports are configured to access the backup node and the computing node group, and configuring the switching device to establish the data exchange path between the backup node and the computing node group via the switching device comprises:
creating an optical path between a first optical fiber port and a second optical fiber port on the optical switch, to establish an optical path between the backup node and the computing node group via the optical switch, wherein
the optical path switches an optical signal between the backup node and the computing node group, to cause the backup node to replace the faulty device to execute the computing task, the first optical fiber port is an optical fiber port for accessing the backup node, the second optical fiber port is an optical fiber port for accessing the computing node group, and the optical signal carries data corresponding to the computing task.

### Claim 33 (depends on 31)

The method according to claim 31, wherein configuring the switching device, to establish the data exchange path between the backup node and the computing node group via the switching device comprises at least one of the following:
configuring the switching device, if the faulty device comprises a first quantity of faulty computing nodes, to establish a first data exchange path between a first quantity of backup nodes and at least one first node via the switching device, wherein the first data exchange path enables data exchange between the first quantity of backup nodes and the at least one first node, to cause the first quantity of backup nodes to replace the first quantity of faulty computing nodes to execute the computing task, and the first node is a node other than the faulty computing node in the computing node group;
configuring the switching device, if the faulty device comprises a second quantity of faulty accelerators, to establish a second data exchange path between a second quantity of backup accelerators in at least one backup node and at least one first accelerator via the switching device, wherein the second data exchange enables data exchange between the second quantity of backup accelerators and the at least one first accelerator, to cause the second quantity of backup accelerators to replace the second quantity of faulty accelerators to execute the computing task, and the first accelerator is an accelerator other than the faulty accelerator in the computing node group; or
if the faulty device comprises a third quantity of faulty accelerators, determining a fourth quantity of nodes to which the third quantity of faulty accelerators belong, and configuring the switching device, to establish a third data exchange path between a fourth quantity of backup nodes and the at least one first node via the switching device, wherein the third data exchange path enables data exchange between the fourth quantity of backup nodes and the at least one first node, to cause the fourth quantity of backup nodes to replace the node to which the third quantity of faulty accelerators belong to execute the computing task.

### Claim 34

A management device, wherein the management device comprises at least one processor and a non-transitory memory, and the at least one processor is configured to execute at least one segment of program code stored in the memory, to cause the management device to perform a fault handling method of:
detecting that a fault occurs in a computing node group, and determining a faulty device in which the fault occurs, wherein the faulty device is a faulty computing node or a faulty accelerator in the computing node group, wherein a switching device is physically connected separately to the computing node group and a backup node, each computing node in the computing node group comprises a plurality of accelerators, and the backup node comprises a plurality of backup accelerators; and
configuring the switching device, to establish a data exchange path between the backup node and the computing node group via the switching device, wherein the data exchange path enables data exchange between the backup node and the computing node group, to cause the backup node to replace the faulty device to execute a computing task.

### Claim 35 (depends on 34)

The management device according to claim 34, wherein the switching device is an optical switch, the optical switch comprises a plurality of optical fiber ports, the plurality of optical fiber ports are configured to access the backup node and the computing node group, and configuring the switching device to establish the data exchange path between the backup node and the computing node group via the switching device comprises:
creating an optical path between a first optical fiber port and a second optical fiber port on the optical switch, to establish an optical path between the backup node and the computing node group via the optical switch, wherein
the optical path switches an optical signal between the backup node and the computing node group, to cause the backup node to replace the faulty device to execute the computing task, the first optical fiber port is an optical fiber port for accessing the backup node, the second optical fiber port is an optical fiber port for accessing the computing node group, and the optical signal carries data corresponding to the computing task.

### Claim 36 (depends on 34)

The management device according to claim 34, wherein configuring the switching device, to establish the data exchange path between the backup node and the computing node group via the switching device comprises at least one of the following:
configuring the switching device, if the faulty device comprises a first quantity of faulty computing nodes, to establish a first data exchange path between a first quantity of backup nodes and at least one first node via the switching device, wherein the first data exchange path enables data exchange between the first quantity of backup nodes and the at least one first node, to cause the first quantity of backup nodes to replace the first quantity of faulty computing nodes to execute the computing task, and the first node is a node other than the faulty computing node in the computing node group;
configuring the switching device, if the faulty device comprises a second quantity of faulty accelerators, to establish a second data exchange path between a second quantity of backup accelerators in at least one backup node and at least one first accelerator via the switching device, wherein the second data exchange path enables data exchange between the second quantity of backup accelerators and the at least one first accelerator, to cause the second quantity of backup accelerators to replace the second quantity of faulty accelerators to execute the computing task, and the first accelerator is an accelerator other than the faulty accelerator in the computing node group; or
if the faulty device comprises a third quantity of faulty accelerators, determining a fourth quantity of nodes to which the third quantity of faulty accelerators belong, and configuring the switching device, to establish a third data exchange path between a fourth quantity of backup nodes and the at least one first node via the switching device, wherein the third data exchange path enables data exchange between the fourth quantity of backup nodes and the at least one first node, to cause the fourth quantity of backup nodes to replace the node to which the third quantity of faulty accelerators belong to execute the computing task.
