EP-4738788-A1 - PACKET PROCESSING METHOD AND APPARATUS, RESOURCE MANAGEMENT METHOD, APPARATUS AND DEVICE

EP4738788A1EP 4738788 A1EP4738788 A1EP 4738788A1EP-4738788-A1

Abstract

The present disclosure provides a packet processing method and apparatus, a resource management method and apparatus, and a device. The packet processing method includes: determining identity information of the member device in the cloud cluster switch; in response to determining that the identity information indicates that the member device is a first NCP and a first protocol packet transmitted by a server is received through a first port, adding local port information of the first port to the first protocol packet to obtain a second protocol packet, and transmitting the second protocol packet to an NCC; and in response to determining that the identity information indicates that the member device is an NCC and the second protocol packet is received, determining global port information of the first port based on the local port information of the first port; establishing a first forwarding entry, where the first forwarding entry includes address information of the server in the first protocol packet and the global port information of the first port; transmitting the first forwarding entry to the first NCP, such that the first NCP, upon receiving a data packet with a destination address being the address information of the server, queries the first forwarding entry based on the destination address to obtain the global port information of the first port, and transmits the data packet through the first port. Through the solution of the present disclosure, forwarding latency is reduced, traffic loss is minimized, and service quality is improved.

Inventors

ZHANG, TAO
FAN, CHI
WANG, WEI

Assignees

New H3C Technologies Co., Ltd.

Dates

Publication Date: 20260506
Application Date: 20240809

Claims (20)

A packet processing method, applied to a member device in a cloud cluster switch, wherein the member device comprises an NCC, an NCF or an NCP, and the method comprises: determining identity information of the member device in the cloud cluster switch; in response to determining that the identity information indicates that the member device is a first NCP and a first protocol packet transmitted by a server is received through a first port, adding local port information of the first port to the first protocol packet to obtain a second protocol packet, and transmitting the second protocol packet to an NCC; and in response to determining that the identity information indicates that the member device is an NCC and the second protocol packet is received, determining global port information of the first port based on the local port information of the first port; establishing a first forwarding entry, wherein the first forwarding entry comprises address information of the server in the first protocol packet and the global port information of the first port; and transmitting the first forwarding entry to the first NCP, such that the first NCP, upon receiving a data packet with a destination address being the address information of the server, queries the first forwarding entry based on the destination address to obtain the global port information of the first port, and transmits the data packet through the first port.
The method according to claim 1, wherein before determining the global port information of the first port based on the local port information of the first port, the method further comprises: in response to determining that the identity information indicates that the member device is a first NCP, establishing a first internal channel with the NCC, and transmitting a first registration message to the NCC through the first internal channel, wherein the first registration message comprises local port information of the first port; and in response to determining that the identity information indicates that the member device is an NCC and the first registration message is received, configuring the global port information for the first port, and recording, in a port mapping table, a corresponding relationship between the local port information of the first port and the global port information of the first port.
The method according to claim 1, wherein after establishing the first forwarding entry, the method further comprises: transmitting the first forwarding entry to a second NCP; and in response to determining that the identity information indicates that the member device is a second NCP and the data packet with the destination address being the address information of the server is received, querying the first forwarding entry based on the destination address to obtain the global port information of the first port, encapsulating the global port information of the first port into the data packet, and transmitting an encapsulated data packet to an NCF.
The method according to claim 3, wherein after determining the global port information of the first port based on the local port information of the first port, the method further comprises: in response to determining that the identity information indicates that the member device is an NCC, determining global port information of a second port of an NCF, wherein the NCF is connected to the first NCP through the second port; establishing a second forwarding entry and transmitting the second forwarding entry to the NCF, wherein the second forwarding entry comprises the global port information of the first port and the global port information of the second port; and in response to determining that the identity information indicates that the member device is an NCF and a received data packet comprises the global port information of the first port, querying the second forwarding entry to obtain the global port information of the second port, and transmitting the data packet through the second port.
The method according to claim 4, wherein before determining the global port information of the second port of the NCF, the method further comprises: in response to determining that the identity information indicates that the member device is an NCF, establishing a second internal channel with the NCC, and transmitting a second registration message to the NCC through the second internal channel, wherein the second registration message comprises local port information of the second port and information of a first NCP connected to the second port; and in response to determining that the identity information indicates that the member device is an NCC and the second registration message is received, configuring the global port information for the second port, and recording, in a port mapping table, a corresponding relationship between the local port information of the second port, the global port information of the second port, and the information of the first NCP.
The method according to claim 1, wherein the NCP comprises an application container and a routing container, and the NCC comprises an application container and a routing container, wherein the application container of the first NCP is for receiving the first protocol packet and transmitting the first protocol packet to the routing container of the first NCP, and the routing container of the first NCP is for adding the local port information of the first port to the first protocol packet to obtain the second protocol packet, and transmitting the second protocol packet to the routing container of the NCC; and the routing container of the NCC is for receiving the second protocol packet and transmitting the second protocol packet to the application container of the NCC, and the application container of the NCC is for determining the global port information of the first port, establishing the first forwarding entry, and transmitting the first forwarding entry to the application container of the first NCP.
The method according to any one of claims 1-6, wherein when a first member device joins the cloud cluster switch, the method further comprises: in response to determining that the identity information indicates that the member device is an NCC, establishing a third internal channel with the first member device and receiving a third registration message transmitted by the first member device through the third internal channel; in response to determining that the first member device is a second NCP and the third registration message is received, transmitting the first forwarding entry to the second NCP; and in response to determining that the first member device is an NCF and the third registration message is received, transmitting the second forwarding entry to the NCF.
The method according to any one of claims 1-6, wherein when a second member device leaves the cloud cluster switch, the method further comprises: in response to determining that the identity information indicates that the member device is an NCC and the second member device is a first NCP, obtaining a third forwarding entry and a fourth forwarding entry corresponding to the second member device, wherein the third forwarding entry comprises the address information of a server and global port information of a port on the second member device, and the fourth forwarding entry comprises the global port information of the port on the second member device and global port information of a port on the NCF; transmitting a first delete instruction to all second NCPs, wherein the first delete instruction is configured to cause each second NCP to delete the third forwarding entry corresponding to the second member device; and transmitting a second delete instruction to all NCFs, wherein the second delete instruction is configured to cause each NCF to delete the fourth forwarding entry corresponding to the second member device.
The method according to any one of claims 1-6, further comprising: in response to determining that the identity information indicates that the member device is an NCP and the NCP is a first NCP or a second NCP, periodically transmitting a first heartbeat message to the NCC through a first internal channel, wherein the first heartbeat message comprises resource information of the NCP; in response to determining that the identity information indicates that the member device is an NCC, receiving the first heartbeat message; and in response to determining that the resource information of the NCP indicates that remaining available resources of the NCP are less than a first threshold, displaying a first alarm information, wherein the first alarm information is configured to prompt a scale-out for NCPs of the cloud cluster switch; or in response to determining that the resource information of the NCP indicates that remaining available resources of the NCP are greater than a second threshold, displaying a second alarm information, wherein the second alarm information is configured to prompt a scale-in for NCPs of the cloud cluster switch.
The method according to any one of claims 1-6, further comprising: in response to determining that the identity information indicates that the member device is an NCF, periodically transmitting a second heartbeat message to the NCC through a second internal channel, wherein the second heartbeat message comprises resource information of the NCF; in response to determining that the identity information indicates that the member device is an NCC, receiving the second heartbeat message; and in response to determining that the resource information of the NCF indicates that remaining available resources of the NCF are less than a third threshold, displaying a third alarm information, wherein the third alarm information is configured to prompt a scale-out for the NCF of the cloud cluster switch; or in response to determining that the resource information of the NCF indicates that remaining available resources of the NCF are greater than a fourth threshold, displaying a fourth alarm information, wherein the fourth alarm information is configured to prompt a scale-in for the NCF of the cloud cluster switch.
A resource management method, comprising: obtaining target resource data of a cloud cluster switch, wherein the target resource data comprises one or more of network element-level resource data, network-level resource data, application-level resource data, or service resource data; obtaining initial demand resource information of a user, wherein the initial demand resource information represents expected occupied resources; determining, based on the target resource data and the initial demand resource information, target demand resource information of the user, wherein the target demand resource information represents resources actually allocated to the user; and transmitting the target demand resource information to the cloud cluster switch, such that the cloud cluster switch allocates available resources to the user based on the target demand resource information when receiving a data packet for the user, and processes the data packet based on the available resources.
The method according to claim 11, wherein the network-level resource data comprises resource data of the cloud cluster switch, and the resource data of the cloud cluster switch comprises overall network utilization rate, network capacity, and network health of the cloud cluster switch; and the network element-level resource data comprises resource data of each member device in the cloud cluster switch, wherein a member device in the cloud cluster switch is an NCC, an NCF, or an NCP, wherein the resource data of the member device comprises a physical resource utilization rate, a logical resource utilization rate, and a health-level of the member device.
The method according to claim 11, wherein the application-level resource data comprises network quality when the cloud cluster switch forwards a data packet, wherein the network quality comprises one or more of a packet loss rate, latency, or jitter; wherein obtaining the application-level resource data corresponding to the cloud cluster switch comprises: receiving an RPC protocol packet transmitted by an RPC module of a member device within the cloud cluster switch, wherein the RPC protocol packet comprises a monitoring indicator counted by the member device, wherein the member device is an NCC, an NCF, or an NCP, and the member device comprises an SQA module, an eMDI module, and the RPC module, wherein the SQA module transmits a specified type of a target traffic characteristic to the eMDI module, wherein the specified type comprises an audio type and/or a video type; wherein when the eMDI module receives a data packet that matches the target traffic characteristic, the eMDI module counts a monitoring indicator of the data packet and transmits the monitoring indicator to the SQA module, the SQA module transmits the monitoring indicator to the RPC module, and the RPC module generates the RPC protocol packet containing the monitoring indicator; determining, based on the monitoring indicator counted by each member device, the application-level resource data.
The method according to claim 11, wherein the service resource data comprises resource data of a server coupled with the cloud cluster switch; in a case that the server is a GPU, the service resource data comprises computing performance of the GPU and/or memory bandwidth of the GPU; wherein the computing performance represents a theoretical computing capability of the GPU when processing data of different accuracies; and the memory bandwidth represents a data transmitting speed provided by the GPU when processing large-scale data.
The method according to any one of claims 11-14, wherein determining the target demand resource information of the user comprises: in response to determining that the initial demand resource information comprises expected occupied resources for the cloud cluster switch, and the target resource data indicates that available resources of the cloud cluster switch are greater than a first resource threshold, determining that the target demand resource information is greater than the initial demand resource information, such that available resources allocated to the user are greater than the expected occupied resources; or in response to determining that the target resource data indicates that available resources of the cloud cluster switch are less than a second resource threshold, determining that the target demand resource information is less than the initial demand resource information, such that available resources allocated to the user are less than the expected occupied resources.
The method according to any one of claims 11-14, wherein determining the target demand resource information of the user comprises: in response to determining that the initial demand resource information comprises expected occupied resources for a server coupled with the cloud cluster switch, and the target resource data indicates that the available resources of the server are greater than a third resource threshold, determining that the target demand resource information is greater than the initial demand resource information, such that available resources allocated to the user are greater than the expected occupied resources; or in response to determining that the target resource data indicates that available resources of the server are less than a fourth resource threshold, determining that the target demand resource information is less than the initial demand resource information, such that available resources allocated to the user are less than the expected occupied resources.
A packet processing apparatus, applied to a member device in a cloud cluster switch, wherein the member device comprises an NCC, an NCF or an NCP, and the apparatus comprises: a determining module, a receiving module, a processing module, and a transmitting module; wherein the determining module is configured to determine identity information of the member device in the cloud cluster switch; wherein in a case that the identity information indicates that the member device is a first NCP, the receiving module is configured to receive a first protocol packet transmitted by a server through a first port; the processing module is configured to add local port information of the first port into the first protocol packet to obtain a second protocol packet; the transmitting module is configured to transmit the second protocol packet to an NCC; wherein in a case that the identity information indicates that the member device is an NCC, the receiving module is configured to receive the second protocol packet; the processing module is configured to determine global port information of the first port based on the local port information of the first port; and establish a first forwarding entry, wherein the first forwarding entry comprises address information of the server in the first protocol packet and the global port information of the first port; and the transmitting module is configured to transmit the first forwarding entry to the first NCP, such that upon receiving a data packet with a destination address being the address information of the server, the first NCP queries the first forwarding entry based on the destination address to obtain the global port information of the first port, and transmits the data packet through the first port.
A resource management apparatus, comprising: an obtaining module, configured to obtain target resource data of a cloud cluster switch, wherein the target resource data comprises one or more of network element-level resource data, network-level resource data, application-level resource data, or service resource data; and obtain initial demand resource information of a user, wherein the initial demand resource information represents expected occupied resources; a determining module, configured to determine, based on the target resource data and the initial demand resource information, target demand resource information of the user, wherein the target demand resource information represents resources actually allocated to the user; and a processing module, configured to transmit the target demand resource information to the cloud cluster switch, such that the cloud cluster switch allocates available resources to the user based on the target demand resource information when receiving a data packet for the user, and processes the data packet based on the available resources.
A member device in a cloud cluster switch, wherein the member device is an NCC, an NCF or an NCP, and comprises: one or more processors and one or more machine-readable storage media, wherein the one or more machine-readable storage media store machine executable instruction that can be executed by the one or more processors, and the one or more processors execute the machine executable instructions to implement the method according to any one of claims 1-10.
A management device, comprising: one or more processors and one or more machine-readable storage media, wherein the one or more machine-readable storage media store machine executable instructions that can be executed by the one or more processors, and the one or more processors execute the machine executable instructions to implement the method according to any one of claims 11-16.

Description

TECHNICAL FIELD The present disclosure relates to the technical field of communications, and in particular to packet processing methods and apparatuses, resource management methods and apparatuses, and devices. BACKGROUND Data Center (DC) is a facility for integrated storage, processing, and distribution of data. DC includes a server, a storage device, and a network device, etc., which is used to provide various computing and storage services. Data Center Interconnect (DCI) is a facility for connecting DCs. DCI is used to achieve high-speed, reliable, and secure communication between DCs, to meet the high bandwidth and low latency requirements for data and applications. With rapid development of the Internet, cloud service, big data and AI technology, the traffic of a DC network rapidly increases, which causes tremendous pressure on the network device of the DC network. For example, the development and application of Artificial Intelligence Generated Content (AIGC) has a characteristic of high computational amount, where from text and image generation to audio and video generation, the computational amount and traffic grow exponentially. For a network device in a DC network, the network device includes a main control board, an interface board, and a network board. Due to the hardware structure limitation of the network device, that is, the number of main control boards, interface boards, and network boards carried is limited, and it is not possible to expand the capacity of the main control boards, interface boards, and network boards. Based on this, when there is a significant increase in traffic, the network device cannot meet a forwarding requirement of a large amount of traffic, resulting in problems such as increased latency and traffic loss. SUMMARY The present disclosure provides a packet processing method, applied to a member device in a cloud cluster switch, where the member device includes an NCC, an NCF or an NCP, and the method includes: determining identity information of the member device in the cloud cluster switch;in response to determining that the identity information indicates that the member device is a first NCP and a first protocol packet transmitted by a server is received through a first port, adding local port information of the first port to the first protocol packet to obtain a second protocol packet, and transmitting the second protocol packet to an NCC; andin response to determining that the identity information indicates that the member device is an NCC and the second protocol packet is received, determining global port information of the first port based on the local port information of the first port; establishing a first forwarding entry, where the first forwarding entry includes address information of the server in the first protocol packet and the global port information of the first port; transmitting the first forwarding entry to the first NCP, such that the first NCP, upon receiving a data packet with a destination address being the address information of the server, queries the first forwarding entry based on the destination address to obtain the global port information of the first port, and transmits the data packet through the first port. The present disclosure provides a resource management method including: obtaining target resource data of a cloud cluster switch, where the target resource data includes one or more of network element-level resource data, network-level resource data, application-level resource data, or service resource data;obtaining initial demand resource information of a user, where the initial demand resource information represents expected occupied resources;determining, based on the target resource data and the initial demand resource information, target demand resource information of the user, where the target demand resource information represents resources actually allocated to the user; andtransmitting the target demand resource information to the cloud cluster switch, such that the cloud cluster switch allocates available resources to the user based on the target demand resource information when receiving a data packet for the user, and processes the data packet based on the available resources. The present disclosure provides a packet processing apparatus, applied to a member device in a cloud cluster switch, where the member device includes an NCC, an NCF or an NCP, and the apparatus includes: a determining module, a receiving module, a processing module, and a transmitting module; wherethe determining module is configured to determine identity information of the member device in the cloud cluster switch;where in a case that the identity information indicates that the member device is a first NCP,the receiving module is configured to receive a first protocol packet transmitted by a server through a first port;the processing module is configured to add local port information of the first port into the first protocol packet to obtain a s