Bandwidth allocation method and device, electronic equipment and storage medium

Abstract

The invention provides a bandwidth allocation method and device, electronic equipment and a storage medium. The method comprises the following steps: monitoring network state indexes of a plurality of private line links in real time; carrying out heat analysis on service flows transmitted on the plurality of private line links based on the network state indexes, and identifying a heat flow meeting a preset condition; selecting a target link from the plurality of private line links according to the flow characteristics of the heat flow and the network state index; establishing a hardware direct connection forwarding channel on the target link for the heat flow, and synchronizing configuration information of the hardware direct connection forwarding channel to a control component; and receiving a global bandwidth allocation strategy issued by the control component based on the configuration information, and calling the hardware direct connection forwarding channel to execute local flow scheduling for the heat flow according to the global bandwidth allocation strategy and the network state index. Therefore, the forwarding performance and the resource utilization rate of the private line gateway are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer networks, and in particular to a bandwidth allocation method and device, electronic equipment and storage medium. BACKGROUND

[0002] With the deepening of enterprise digital transformation, the hybrid cloud architecture has become a mainstream trend, and more and more businesses rely on dedicated lines to achieve high-speed and stable interconnection between local data centers (IDC) and public cloud virtual private clouds (VPC). As the core hub for realizing cross-cloud and cross-regional network interconnection, the forwarding performance, bandwidth utilization and scheduling efficiency of the dedicated line gateway directly affect the experience of key businesses.

[0003] The prior art usually adopts a software solution based on a central processing unit (CPU) to implement multi-link bandwidth scheduling of the dedicated line gateway. The control plane periodically collects state information of each dedicated line, centrally calculates the traffic scheduling strategy at the software level, and then issues it to the data plane. The data plane performs traffic forwarding through software (such as a kernel protocol stack or a data plane development kit (DPDK)) running on a general server, and uses algorithms such as equal-cost multipath routing (ECMP) or weighted hashing to distribute traffic to multiple links.

[0004] However, this centralized software processing architecture requires the CPU to complete the policy check, network address translation (NAT) and access control list (ACL) filtering of all traffic, so there is a problem of high response delay and serious CPU forwarding bottleneck, which cannot meet the demand for millisecond-level bandwidth scheduling and low-latency forwarding of burst traffic. SUMMARY

[0005] The present application provides a bandwidth allocation method, device, electronic equipment and storage medium to solve the problem of high response delay and serious CPU forwarding bottleneck in the centralized software processing architecture of the prior art.

[0006] In a first aspect, the present application provides a bandwidth allocation method applied to an intelligent network card, comprising: real-time monitoring of network state indicators of multiple dedicated line links; performing heat analysis on traffic flows transmitted on the multiple dedicated line links based on the network status indicators, and identifying hot flows meeting preset conditions; selecting a target link from the multiple dedicated line links according to the traffic characteristics of the hot flows and the network status indicators; establishing a hardware direct connection forwarding channel for the hot flows on the target link, and synchronizing configuration information of the hardware direct connection forwarding channel to a control component; receiving a global bandwidth allocation strategy issued by the control component based on the configuration information, and performing local traffic scheduling for the hot flows by the hardware direct connection forwarding channel according to the global bandwidth allocation strategy and the network status indicators.

[0007] In one possible implementation, the performing heat analysis on traffic flows transmitted on the multiple dedicated line links based on the network status indicators, and identifying hot flows meeting preset conditions, comprises: extracting bandwidth utilization rate data and transmission delay data of each dedicated line link from the network status indicators; statistically calculating byte number of each traffic flow based on the bandwidth utilization rate data, and statistically calculating packet rate of each traffic flow based on the transmission delay data within a sliding time window; determining any traffic flow as a hot flow when the byte number of the traffic flow within the window exceeds a first threshold value, or the packet rate exceeds a second threshold value.

[0008] In one possible implementation, the selecting a target link from the multiple dedicated line links according to the traffic characteristics of the hot flows and the network status indicators, comprises: extracting residual bandwidth and current delay of each dedicated line link from the network status indicators; determining a dedicated line link as a candidate link when the residual bandwidth of the dedicated line link is greater than or equal to a bandwidth condition of the hot flows, and obtaining a candidate link set; determining a candidate link with the lowest current delay in the candidate link set as the target link.

[0009] In one possible implementation, after the performing local traffic scheduling for the hot flows by the hardware direct connection forwarding channel according to the global bandwidth allocation strategy and the network status indicators, the method further comprises: extracting residual available bandwidth of each dedicated line link from the network status indicators; when the residual available bandwidth is lower than a preset threshold value, recalculating bandwidth sharing weights of each traffic flow according to traffic priority levels defined in the global bandwidth allocation strategy, and adjusting flow table output port configuration of the hardware direct connection forwarding channel according to the recalculated bandwidth sharing weights.

[0010] In a possible implementation, the receiving control component issues a global bandwidth allocation strategy based on the configuration information, including: The first channel adopts a low-delay transmission protocol. The second channel adopts a reliable transmission protocol. When it is detected that the version of the locally stored bandwidth allocation strategy is inconsistent with the version of the strategy of the control component, a resynchronization request is initiated to the control component.

[0011] In a possible implementation, the method further includes: When it is detected that the remaining available bandwidth of the target link continuously falls below a failure threshold for a preset length of time, transmission of the hot flow through the hardware direct forwarding channel is terminated, a new hardware direct forwarding channel is established by reselecting a backup link from the multiple dedicated line links, and a channel switching event is reported to the control component.

[0012] In a possible implementation, the method further includes: When none of the backup links meets the bandwidth requirement, the hardware direct forwarding channel binding of the hot flow is released, and encapsulation and forwarding operations of the hot flow are performed by a host central processing unit.

[0013] In a possible implementation, the invoking the hardware direct forwarding channel to perform local traffic scheduling for the hot flow according to the global bandwidth allocation strategy and the network state indicators includes: The physical bandwidth resources of the multiple dedicated line links are virtualized into a unified global bandwidth pool, wherein a real-time total capacity of the global bandwidth pool is dynamically adjusted based on the bandwidth utilization rates of the links in the network state indicators; Based on tenant quota and service priority parameters defined in the global bandwidth allocation strategy, an allocation amount of bandwidth tokens for the hot flow is determined from the global bandwidth pool, and the allocation amount of bandwidth tokens is positively correlated with a real-time remaining bandwidth of a target link of the hot flow; The token bucket parameters in the flow table items corresponding to the hardware direct forwarding channel are configured according to the allocation amount of bandwidth tokens; The hardware direct forwarding channel allocates exclusive bandwidth tokens for the hot flow based on the token bucket parameters and implements token rate limiting forwarding.

[0014] In a second aspect, the present application provides a bandwidth allocation device applied to an intelligent network card, including: A monitoring module is configured to monitor network state indicators of multiple dedicated line links in real time. an identifying module configured to perform heat analysis on traffic flows transmitted on the multiple dedicated line links based on the network status indicators, and identify a hot flow meeting a preset condition; a selecting module configured to select a target link from the multiple dedicated line links according to traffic characteristics of the hot flow and the network status indicators; an establishing module configured to establish a hardware direct forwarding channel for the hot flow on the target link, and synchronize configuration information of the hardware direct forwarding channel to a control component; a scheduling module configured to receive a global bandwidth allocation strategy issued by the control component based on the configuration information, and perform local traffic scheduling on the hot flow by the hardware direct forwarding channel according to the global bandwidth allocation strategy and the network status indicators.

[0015] In one possible implementation, the identifying module is specifically configured to: extract bandwidth utilization rate data and transmission delay data of each dedicated line link from the network status indicators; statistically obtain byte number of each traffic flow based on the bandwidth utilization rate data and packet rate of each traffic flow based on the transmission delay data within a sliding time window; determine a traffic flow as a hot flow when byte number of the traffic flow within the window exceeds a first threshold value, or packet rate of the traffic flow exceeds a second threshold value.

[0016] In one possible implementation, the selecting module is specifically configured to: extract residual bandwidth and current delay of each dedicated line link from the network status indicators; determine a dedicated line link as a candidate link when residual bandwidth of the dedicated line link is greater than or equal to a bandwidth condition of the hot flow, and obtain a candidate link set; determine a candidate link with the lowest current delay in the candidate link set as the target link.

[0017] In one possible implementation, the apparatus further includes an adjusting module configured to: extract residual available bandwidth of each dedicated line link from the network status indicators; when the residual available bandwidth is lower than a preset threshold value, re-calculate bandwidth sharing weights of each traffic flow according to a traffic priority level defined in the global bandwidth allocation strategy, and adjust flow table output port configuration of the hardware direct forwarding channel according to the re-calculated bandwidth sharing weights.

[0018] In one possible implementation, the scheduling module is configured to: receive, through a first channel, an approval result of the control component on the hardware direct forwarding channel, the first channel using a low-delay transmission protocol; receive, through a second channel, complete parameters of the global bandwidth allocation strategy issued by the control component, the second channel using a reliable transmission protocol; initiate a resynchronization request to the control component when detecting that a locally stored bandwidth allocation strategy version is inconsistent with a strategy version of the control component.

[0019] In one possible implementation, the apparatus further includes a reselection module configured to: terminate transmission of the hot flow through the hardware direct forwarding channel when detecting that the remaining available bandwidth of the target link continuously falls below a failure threshold for a preset length of time, and reselect a backup link from the multiple dedicated line links to establish a new hardware direct forwarding channel, and report a channel switching event to the control component.

[0020] In one possible implementation, the apparatus further includes an execution module configured to: when none of the backup links meets the bandwidth requirement, unbind the hardware direct forwarding channel of the hot flow, and perform encapsulation and forwarding operations of the hot flow by a host central processing unit.

[0021] In one possible implementation, the scheduling module is specifically configured to: virtualize physical bandwidth resources of the multiple dedicated line links into a unified global bandwidth pool, wherein a real-time total capacity of the global bandwidth pool is dynamically adjusted based on link bandwidth utilization rates in the network state indicators; determine a bandwidth token allocation quota of the hot flow from the global bandwidth pool based on tenant quota and service priority parameters defined in the global bandwidth allocation strategy, the bandwidth token allocation quota being positively correlated with a real-time remaining bandwidth of a target link of the hot flow; configure a token bucket parameter in a flow table item corresponding to the hardware direct forwarding channel according to the bandwidth token allocation quota; allocate exclusive bandwidth tokens for the hot flow based on the token bucket parameter through the hardware direct forwarding channel and implement token rate limiting forwarding.

[0022] In a third aspect, the present application provides an apparatus, comprising: a processor and a memory, the processor being configured to execute a bandwidth allocation program stored in the memory to implement the bandwidth allocation method of any one of the first aspect.

[0023] In a fourth aspect, the present application provides a storage medium, which stores one or more programs, and the one or more programs are executable by one or more processors to implement the bandwidth allocation method in any one of the first aspect.

[0024] Compared with the prior art, the above technical solution provided by the embodiments of the present application has the following advantages: first, the method provided by the embodiments of the present application periodically collects real-time state indicators of multiple private line links through an intelligent network card, and completes preliminary traffic scheduling at the hardware layer. This hardware-level processing method avoids the delay of the traditional control plane centralized collection, calculation and instruction issuing, and shortens the scheduling response time from seconds to milliseconds. Second, the intelligent network card built-in service flow statistical module quickly identifies target flows with high-frequency access and low-latency sensitivity, and directly guides them to the hardware forwarding channel, bypassing the CPU processing. This mechanism completely avoids the delay and resource competition caused by software processing of high-frequency service flows through the CPU, and fundamentally eliminates the CPU forwarding bottleneck. Thus, the millisecond-level multi-link bandwidth dynamic scheduling and hardware direct connection forwarding of hot flows are realized, and the forwarding performance and resource utilization of the private line gateway are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0025] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and serve to explain the principles of the present application together with the specification.

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, for those skilled in the art, other drawings can also be obtained based on these drawings without any creative effort.

[0027] One or more embodiments are exemplarily illustrated by pictures in the drawings corresponding thereto, and these exemplary illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, unless otherwise specified. The drawings in the drawings do not constitute a proportional limitation.

[0028] Figure 1 An embodiment flowchart of a bandwidth allocation method provided by the embodiments of the present application; Figure 2 An embodiment flowchart of another bandwidth allocation method provided by the embodiments of the present application; Figure 3 A bandwidth allocation system architecture diagram provided by the embodiments of the present application; Figure 4 An embodiment block diagram of a bandwidth allocation device provided by the embodiments of the present application; Figure 5A structural schematic diagram of an electronic device is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0029] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0030] The following disclosure provides many different embodiments, or examples, for implementing different structures of the present application. For the purpose of simplifying the present application, components and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present application. In addition, reference numerals and / or letters can be repeated in different examples in the present application. Such repetition is for the purpose of simplification and clarity, and does not indicate the relationship between the various embodiments and / or settings discussed.

[0031] In order to solve the technical problems that the centralized software processing architecture in the prior art has high response delay, serious CPU forwarding bottleneck, and cannot meet the demand of burst traffic for millisecond-level bandwidth scheduling and low-delay forwarding, the present application provides a bandwidth allocation method, which can realize millisecond-level multi-link bandwidth dynamic scheduling and hot flow hardware direct connection forwarding, and significantly improves the forwarding performance and resource utilization of the private line gateway.

[0032] Figure 1 An embodiment flowchart of a bandwidth allocation method is provided for an embodiment of the present application. As shown in Figure 1 The method comprises the following steps: Step 101, real-time monitoring of network state indicators of a plurality of private line links.

[0033] Private line link: refers to a dedicated physical or virtual communication channel connecting a local data center and a cloud virtual private cloud, providing high reliability and low-delay network connection.

[0034] Network state indicators: include bandwidth utilization, transmission delay, packet loss rate, link jitter and other parameters reflecting real-time quality and performance of the link.

[0035] In the embodiments of the present application, the intelligent network card actively collects real-time network status indicators of each dedicated line link in a millisecond level period through the built-in hardware counter and state monitoring module. The intelligent network card acquires the bandwidth utilization by directly reading the traffic statistics register of the physical port or virtual interface, and acquires the transmission delay by sending a probe packet and calculating the response time, thereby realizing link state awareness without the intervention of the CPU (Central Processing Unit), and providing accurate and low-delay data basis for subsequent traffic scheduling.

[0036] Step 102, based on the network status indicators, performing heat analysis on the service flows transmitted on the plurality of dedicated line links, and identifying hot flows meeting preset conditions.

[0037] Service flow: refers to a data packet sequence in a communication session uniquely identified by five tuples of source IP, destination IP, protocol type, source port and destination port.

[0038] Heat analysis: refers to a process of statistically and evaluating the traffic characteristics of service flows, mainly including byte number, packet number, packet sending rate and other indicators within a unit time.

[0039] Hot flow: refers to a service flow whose traffic data exceeds a preset threshold (such as byte number greater than 50MB or packet rate higher than 2000 packets / second) within a set sliding time window (for example, 5 seconds).

[0040] In the embodiments of the present application, step 102 can specifically include the following steps: extracting bandwidth utilization data and transmission delay data of each dedicated line link from the network status indicators; within a sliding time window, based on the bandwidth utilization data, counting the byte number of each service flow, and based on the transmission delay data, counting the packet rate of each service flow; when the byte number of any one service flow within the window exceeds a first threshold, or the packet rate exceeds a second threshold, determining the service flow as a hot flow.

[0041] Bandwidth utilization data: refers to the percentage of the actual bandwidth used by a certain private line link in a unit of time to the theoretical maximum bandwidth, which is a key indicator to measure the load of the link. Transmission delay data: refers to the time required for a data packet to be transmitted from the sending end to the receiving end, usually measured in milliseconds (ms), used to measure the delay performance of the link. Sliding time window: refers to a fixed length of time interval (such as 5 seconds) that slides with time, used to limit the time range of statistical calculation, to ensure the timeliness of the analysis. Byte number: refers to the total amount of data transmitted by a certain traffic flow within the statistical time window, measured in bytes. Packet rate: refers to the number of data packets passed by a certain traffic flow in a unit of time, usually measured in pps (packets per second). First threshold value: the byte number threshold value (such as 50MB) for determining whether a traffic flow is a hot flow, which is identified as a hot flow if it exceeds this value. Second threshold value: the packet rate threshold value (such as 2000 packets per second) for determining whether a traffic flow is a hot flow, which is identified as a hot flow if it exceeds this value.

[0042] In this scheme, first, the real-time bandwidth utilization data and transmission delay data of each private line link are extracted from the network state indicators; then, within the set sliding time window (such as 5 seconds), the total byte number of each traffic flow transmitted in this time period is calculated based on the bandwidth utilization data, and the real-time packet rate of each traffic flow is calculated based on the transmission delay data; finally, when it is detected that the byte number of any one traffic flow in the current sliding time window exceeds the first threshold value (such as 50MB), or its packet rate exceeds the second threshold value (such as 2000 packets per second), the traffic flow is determined as a hot flow. Through this identification mechanism based on multi-dimensional indicators and dynamic threshold values, the application can quickly and accurately identify the key traffic flow that is sensitive to delay or occupies high bandwidth from a large number of traffic flows, providing accurate target objects for subsequent hardware direct forwarding, thereby significantly improving the processing efficiency of the private gateway and the business experience.

[0043] Step 103, selecting a target link from a plurality of private line links according to the traffic characteristics of the hot flow and the network state indicators.

[0044] Traffic characteristics: refer to the network transmission demand characteristics of the hot flow, including the required bandwidth, tolerable delay, SLA (Service Level Agreement) requirements, etc.

[0045] Target link: refers to the private channel selected from the available links that is most suitable for transmitting the hot flow according to the demand of the hot flow and the real-time state of the plurality of private line links.

[0046] In the embodiment of the present application, S103 can specifically include the following steps: extracting the residual bandwidth and current latency of each dedicated line link from the network state indicators; in the case that the residual bandwidth of any dedicated line link is greater than or equal to the bandwidth condition of the hot flow, determining the dedicated line link as a candidate link to obtain a candidate link set; and determining the candidate link with the lowest current latency in the candidate link set as a target link.

[0047] Residual bandwidth: refers to the bandwidth resource that is not currently occupied and can be allocated in a dedicated line link, usually measured in Mbps or Gbps. Current latency: refers to the real-time delay time required for data packets to be transmitted on a dedicated line link, usually measured in milliseconds (ms), reflecting the instant communication quality of the link. Bandwidth condition: refers to the bandwidth resource required for normal transmission of the target hot flow, usually pre-set according to the service type or quality of service requirement. Candidate link set: refers to the set of available links selected from all dedicated line links, whose residual bandwidth meets the demand of the hot flow. Target link: refers to the dedicated line link finally selected for transmitting the hot flow, which is the optimal path selected from the candidate link set according to the pre-set strategy.

[0048] In this scheme, first, the residual bandwidth and current latency data of each dedicated line link are extracted from the real-time collected network state indicators; then, the residual bandwidth of each link is compared with the bandwidth demand of the target hot flow, and those dedicated line links with residual bandwidth greater than or equal to the bandwidth demand of the hot flow are selected as candidate links to form a candidate link set; finally, the link with the lowest current latency is selected from the candidate link set and determined as the target link, which is the final forwarding path of the hot flow. Through this multi-level screening mechanism based on real-time link state, the present application can dynamically and efficiently select the optimal forwarding path for the hot flow, which not only fully guarantees the bandwidth demand of the hot flow, but also maximally reduces the transmission delay, thereby significantly improving the link utilization rate of the dedicated gateway and the communication quality of the key service.

[0049] Step 104, establishing a hardware direct forwarding channel for the hot flow on the target link, and synchronizing the configuration information of the hardware direct forwarding channel to the control component.

[0050] Hardware direct forwarding channel: refers to a high-speed forwarding path created on an intelligent network card, which is realized through a flow table item or a hardware forwarding table, so that the data packets matching the flow table item are directly forwarded within the network card without being uploaded to the host CPU for processing.

[0051] Control component: refers to a centralized management software system running on the control plane, responsible for global policy management, resource coordination and state monitoring.

[0052] In the embodiments of the present application, the intelligent network card creates a hardware flow table item for the hot flow and specifies that it is forwarded from the target link. The flow table item contains a matching rule (five-tuple) and a forwarding action (output port or tunnel encapsulation parameter). At the same time, the intelligent network card reports the configuration information (including flow identifier, target link, allocated bandwidth, etc.) of the created direct forwarding channel to the control component in real time through a high-speed communication channel (such as RDMA (Remote Direct Memory Access) or PCIe (Peripheral Component Interconnect Express) channel) to ensure synchronization of the global state of the control plane.

[0053] In step 105, the control component based on the global bandwidth allocation strategy issued by the configuration information, and according to the global bandwidth allocation strategy and the network state index, calls the hardware direct forwarding channel to perform local traffic scheduling for the hot flow.

[0054] Global bandwidth allocation strategy: refers to the bandwidth management strategy formulated by the control component, including total bandwidth pool allocation, basic quota of each tenant or service, priority rule, etc.

[0055] Local traffic scheduling: refers to the intelligent network card performing fine-grained management operations such as rate control and priority scheduling on the traffic passing through the hardware direct forwarding channel according to the global strategy and real-time network state.

[0056] In the embodiments of the present application, after the control component receives the direct channel information reported by the intelligent network card, it audits or adjusts according to the global bandwidth allocation strategy, and issues the strategy decision (such as bandwidth limit, priority mark) to the intelligent network card. After the intelligent network card receives the strategy, it combines the real-time collected network state index, and performs local rate control and scheduling on the hot flow in the direct channel through the hardware traffic scheduler (such as using token bucket or queue scheduling mechanism), to ensure that it conforms to the global strategy and fully utilizes the link resources.

[0057] The technical scheme provided by the embodiments of the present application firstly collects real-time state indexes of multiple private line links periodically through an intelligent network card, and completes preliminary traffic scheduling at a hardware layer. This hardware-level processing mode avoids the delay of the traditional control plane centralized collection, calculation and instruction issuing, and shortens the scheduling response time from seconds to milliseconds. Secondly, the intelligent network card built-in service flow statistics module quickly identifies target flows with high-frequency access and low-latency sensitivity, and directly guides them to the hardware forwarding channel, bypassing the CPU processing. This mechanism completely avoids the delay and resource competition caused by CPU software processing for high-frequency service flows, and fundamentally eliminates the CPU forwarding bottleneck. Thus, millisecond-level multi-link bandwidth dynamic scheduling and hardware direct connection forwarding of hot flows are realized, and the forwarding performance and resource utilization of the private line gateway are significantly improved.

[0058] In an embodiment, the receiving control component issues a global bandwidth allocation strategy based on the configuration information, which can specifically include the following steps: receiving, through a first channel, an approval result of the control component for the hardware direct connection forwarding channel, the first channel using a low-latency transmission protocol; receiving, through a second channel, complete parameters of the global bandwidth allocation strategy issued by the control component, the second channel using a reliable transmission protocol; and initiating a resynchronization request to the control component when detecting that the version of the locally stored bandwidth allocation strategy is inconsistent with the strategy version of the control component.

[0059] The first channel refers to a special low-latency communication link established between the intelligent network card and the control component, which is used to transmit key instructions or approval results that need to be responded quickly.

[0060] The low-latency transmission protocol refers to a communication protocol such as UDP (User Datagram Protocol) or QUIC (Quick UDP Internet Connections) that is designed to reduce transmission delay.

[0061] The approval result refers to the audit conclusion of the control component for the hardware direct connection forwarding channel creation request reported by the intelligent network card, and its core content is approval or rejection. If approved, the specific constraints of this approval will be carried in the result, such as: including the traffic type whitelist allowed to be directly connected; the maximum allowed direct connection duration; the minimum link health threshold.

[0062] The second channel refers to a high-reliability communication link established between the intelligent network card and the control component, which is used to transmit policy and configuration data that need to be guaranteed for integrity and consistency.

[0063] Reliable transport protocol: refers to a communication protocol such as HTTP / 2 (Hypertext Transfer Protocol version 2) or gRPC (Google Remote Procedure Call) that provides reliability guarantees.

[0064] Complete parameters of the global bandwidth allocation strategy: refer to all configuration information of the policies such as bandwidth pool allocation, tenant quota, and priority rules formulated by the control plane.

[0065] Policy version: refers to a version number used to identify the update status of the global bandwidth allocation strategy, which is usually assigned and managed by the control component, and is used to maintain the consistency of the policies among multiple nodes.

[0066] Resynchronization request: refers to a request for policy resynchronization initiated by the intelligent network card to the control component when the intelligent network card detects that the local policy is inconsistent with the policy version of the control component.

[0067] In this embodiment, first, the intelligent network card receives the fast approval result of the control component on the reported hardware direct forwarding channel configuration through the first channel (using a low-latency transmission protocol such as UDP or QUIC); second, it receives the complete parameter set of the global bandwidth allocation strategy issued by the control component through the second channel (using a reliable transmission protocol such as gRPC); finally, the intelligent network card compares the local and server policy version numbers after receiving the policy, and automatically initiates a resynchronization request to the control component when detecting that the versions are inconsistent, to ensure the consistency of the local policy and the global policy. This scheme uses a dual-channel coordinated policy issuance and version synchronization mechanism to achieve low-latency response to critical instructions, while ensuring reliable transmission and consistency maintenance of bulk policy data, thereby improving system response speed while ensuring the accuracy and reliability of global resource scheduling.

[0068] Figure 2 Another embodiment flowchart of the bandwidth allocation method provided in the embodiments of the present application. Figure 2 The flowchart shown in Figure 1 Based on the flowchart shown, the following steps are included: Step 201: Virtualize the physical bandwidth resources of multiple dedicated line links into a unified global bandwidth pool, wherein the real-time total capacity of the global bandwidth pool is dynamically adjusted based on the link bandwidth utilization rates in the network status indicators.

[0069] Step 202: Based on the tenant quota and service priority parameters defined in the global bandwidth allocation strategy, determine the bandwidth token allocation quota of the hot flow from the global bandwidth pool, which is positively correlated with the real-time residual bandwidth of the target link where the hot flow is located.

[0070] Step 203, configuring the token bucket parameters in the flow table item corresponding to the hardware direct forwarding channel according to the bandwidth token allocation quota.

[0071] Step 204, allocating exclusive bandwidth tokens for the hot flow and implementing token rate limiting forwarding based on the token bucket parameters through the hardware direct forwarding channel.

[0072] For ease of understanding, the following unified description is made for steps 201-204: Global bandwidth pool: refers to a unified logical bandwidth resource pool formed by integrating the physical bandwidth resources of multiple dedicated line links through virtualization technology, the total capacity of which is dynamically calculated according to the real-time bandwidth utilization of each link, rather than simply fixed summation.

[0073] Tenant quota: refers to a basic bandwidth guarantee value defined in advance in the global bandwidth allocation strategy and allocated to a specific tenant or service.

[0074] Business priority parameter: refers to a weight or level indicator defined in the strategy to distinguish the importance of the business, and high-priority businesses can be allocated bandwidth resources in priority.

[0075] Bandwidth token allocation quota: refers to the upper limit value of the allocable bandwidth calculated for the hot flow according to the global strategy and real-time link state, expressed in token quantity, which is positively related to the current remaining bandwidth of the target link.

[0076] Token bucket parameters: refers to a set of parameters configured in the flow table item for controlling traffic, usually including token generation rate (rate) and bucket capacity (burst).

[0077] Token rate limiting forwarding: refers to generating tokens according to the token bucket parameters, and data packets can be forwarded only after obtaining tokens, thereby achieving accurate control of bandwidth.

[0078] In the embodiments of the present application, first, the intelligent network card periodically collects the bandwidth utilization data of each dedicated line link, and dynamically calculates the real-time total capacity of the global bandwidth pool through a weighting algorithm (such as weighted summation based on the actual residual bandwidth of the link). The capacity value will be adjusted with the change of the link state. Then, the local scheduling module on the intelligent network card first queries the pre-deployed global strategy to obtain the quota and service priority of the tenant to which the hot flow belongs, and then combines the real-time residual bandwidth of the target link (obtained by reading the hardware register) to calculate the accurate bandwidth token allocation quota by using a weighted allocation algorithm (such as: allocation quota = tenant basic quota × priority weight × (current link residual bandwidth / link total bandwidth)). Next, the intelligent network card driver converts the calculated bandwidth token allocation quota into token bucket parameters (such as token generation rate = allocation quota, bucket capacity = allocation quota × buffer coefficient), and configures these parameters into the flow table item corresponding to the hardware direct forwarding channel through the call of the network card hardware programming interface (such as NIC Firmware API). Finally, the hardware token bucket counter on the intelligent network card generates tokens continuously according to the parameters configured in the flow table item. When the data packet arrives, the hardware looks up the corresponding flow table item and tries to obtain the token. Only the data packet that successfully obtains the token will be immediately forwarded, otherwise it will be cached or discarded, thereby realizing the token rate limiting forwarding at line speed.

[0079] The scheme realizes the fine management and fair allocation of bandwidth resources through the dynamic resource virtualization and token bucket rate limiting mechanism based on hardware implementation as described above, which not only guarantees the bandwidth demand of high-priority hot flow, but also avoids the excessive occupation of resources by a single flow, significantly improves the multi-link bandwidth utilization rate while ensuring the quality and fairness of network services.

[0080] In another embodiment, after step 105, the following steps can also be included: extracting the residual available bandwidth of each dedicated line link from the network state indicators; when the residual available bandwidth is lower than a preset threshold, recalculating the bandwidth sharing weight of each service flow according to the service priority level defined in the global bandwidth allocation strategy, and adjusting the flow table output port configuration of the hardware direct forwarding channel according to the recalculated bandwidth sharing weight.

[0081] Residual available bandwidth: refers to the bandwidth resource of a certain private line link that is not currently allocated, which is usually calculated by subtracting the currently allocated bandwidth from the total bandwidth of the link, and is expressed in Mbps or Gbps. Programmable module: refers to a hardware component in an intelligent network card that supports logical definition through P4, eBPF or other programming languages, used to implement flexible traffic processing strategies. Service priority level: refers to the classification of the importance of services defined in the global bandwidth allocation strategy, usually represented by a numerical value (such as 1-10), with the higher the numerical value, the higher the priority. Bandwidth sharing weight: refers to the proportion coefficient of bandwidth allocation for different flows according to service priority, used to determine the bandwidth share that each flow can obtain when resources are competed. Flow table output port configuration: refers to the action parameter in the intelligent network card flow table for specifying the forwarding port of the message, which can change the output link of the flow by modifying this configuration.

[0082] In this embodiment, first, the residual available bandwidth of each private line link is monitored in real time, and when it is detected that the residual available bandwidth of a certain link is lower than a preset threshold (such as 10% of the total bandwidth of the link), the dynamic adjustment mechanism is triggered; then, according to the service priority level defined in the global bandwidth allocation strategy, the bandwidth sharing weight of each service flow is recalculated using a weighted allocation algorithm (such as: weight = base weight x priority coefficient); finally, according to the recalculated weight, the traffic is dynamically redistributed to the link with more residual bandwidth by modifying the output port action parameter (such as adjusting the hash weight or directly specifying the output port) of the relevant flow table entry in the intelligent network card hardware flow table, realizing the optimal allocation of bandwidth resources.

[0083] This scheme introduces a dynamic weight adjustment mechanism based on real-time bandwidth availability and service priority, which can automatically and quickly schedule traffic when bandwidth resources are tight, ensuring the quality of service of high-priority services and maximizing the utilization efficiency of multi-link bandwidth resources, and enhancing the adaptive ability of the private line gateway in burst scenarios.

[0084] In yet another embodiment, the method can further include the steps of: when it is detected that the residual available bandwidth of the target link is continuously lower than the failure threshold for a preset duration, terminating the transmission of the hot flow through the hardware direct forwarding channel, and reselecting a backup link from the multiple private line links to establish a new hardware direct forwarding channel, and reporting the channel switching event to the control component.

[0085] Fault threshold: refers to a pre-set bandwidth threshold (e.g., less than 5% of the total capacity of the link) for determining whether the link has failed, which is usually lower than the regular load adjustment threshold. Pre-set duration: refers to a time window (e.g., 3 consecutive detection periods) for confirming the continuous abnormality of the link to avoid false switching caused by transient fluctuations. Backup link: refers to a pre-planned or dynamically selected alternative path that can be used to carry the traffic of the failed link among multiple dedicated line links. Channel switching event: refers to a notification message recording the details of link switching, including the original link identifier, the new link identifier, the switching timestamp, and the affected flow identifier, etc.

[0086] In this embodiment, the intelligent network card continuously monitors the remaining available bandwidth of the target link. When it is detected that the value is continuously lower than the fault threshold for a pre-set duration (e.g., less than 5% for 3 consecutive sampling periods), it is determined that the link has failed or is severely congested. Then, the intelligent network card immediately terminates the transmission of hot flows through the original hardware direct forwarding channel, and re-selects the optimal backup link (selection criteria include remaining bandwidth, latency, and priority matching degree) from other dedicated line links according to the real-time network state, and quickly establishes a new hardware direct forwarding channel based on the backup link. Finally, the intelligent network card reports the channel switching event to the control component through a reliable channel (e.g., gRPC) for updating the global state and conducting audit analysis.

[0087] This scheme introduces a link fault rapid detection and automatic switching mechanism based on hardware monitoring, which can complete fault perception and traffic rerouting within milliseconds, greatly improving the reliability and business continuity of the dedicated gateway. At the same time, by timely reporting the switching event, the accuracy of the global state of the control plane is guaranteed, and high availability protection in the fault scenario is realized.

[0088] In addition, in another embodiment, the method can further include the following steps: when all backup links do not meet the bandwidth requirement, the hardware direct forwarding channel binding of the hot flow is released, and the encapsulation and forwarding operations of the hot flow are performed by the host central processor.

[0089] Backup link: refers to a dedicated line link other than the currently used target link, which meets the basic connectivity and is allowed for business transmission by policy. Bandwidth requirement: refers to the minimum bandwidth value required to ensure normal transmission of a specific hot flow, which is usually pre-set based on the type of service or quality of service requirement. Hardware direct forwarding channel binding: refers to the hardware flow table item established and activated on the intelligent network card for the forwarding of a specific hot flow, which makes the traffic bypass the host CPU. Host central processing unit: refers to the server host CPU running a general operating system, responsible for performing software protocol stack processing, encapsulation and forwarding, etc. Encapsulation and forwarding operation: refers to the operation of processing data packets through software, such as tunnel encapsulation (such as VXLAN, IPsec), route query, policy check, etc., and sending to the next hop.

[0090] In this embodiment, when the intelligent network card detects that the remaining available bandwidth of all backup links cannot meet the minimum bandwidth requirement of the hot flow, the disaster recovery fallback process is triggered: the intelligent network card first unbinds the hot flow from the hardware direct forwarding channel (i.e. deletes or deactivates the corresponding hardware flow table item), and then notifies the host CPU through an interrupt or a message; the host CPU takes over the processing task of the hot flow, performs tunnel encapsulation, route query and forwarding operation on data packets through the kernel network protocol stack or the user mode forwarding framework, to ensure the continuous transmission of traffic.

[0091] This scheme introduces an intelligent disaster recovery fallback mechanism, which can still guarantee the connectivity of critical services in extreme resource shortage scenarios, realizes a flexible processing architecture of "hardware acceleration as the main and software forwarding as the auxiliary", and effectively improves the disaster recovery capability and business continuity of the system.

[0092] Figure 3 A bandwidth allocation system architecture diagram is provided for the embodiments of the present application. As shown in Figure 3 , the architecture includes the following components: 1. VPC (Virtual Private Cloud) Function: used to carry business instances in the cloud, including virtual machines, containers, etc., and provides a private Layer 2 or Layer 3 network environment.

[0093] Key features: communication between internal instances does not need to go through the public network; when accessing IDC or other VPCs is needed, traffic must be routed and policy processed through the transit gateway (TGW).

[0094] Association with SmartNIC: the physical network card of each cloud host can use an intelligent network card (SmartNIC), so that traffic detection, acceleration and direct forwarding, etc. can be completed locally on the host side; when interfacing with the TGW, the outgoing traffic can be processed by the SmartNIC first and then sent out.

[0095] 2. IDC (Internet Data Center, Internet Data Center) Function: Host enterprise local business systems and databases, usually interconnected with the cloud through a dedicated line.

[0096] Key Features: Provide low-latency, high-bandwidth dedicated line channels; when accessing the cloud VPC, routing and policy control must be performed through the TGW.

[0097] Association with SmartNIC: SmartNIC can be deployed in IDC boundary network devices or cloud access devices to achieve fast path selection and traffic acceleration.

[0098] 3. TGW (Transit Gateway) Function: As the core hub for intercommunication between VPC and IDC, it is responsible for cross-network three-layer forwarding, network address translation (NAT), access control list (ACL) filtering, route aggregation, etc.

[0099] Key Capabilities: NAT Function: Supports source address translation (SNAT) to convert VPC private network addresses to IDC hybrid addresses and destination address translation (DNAT) to map IDC addresses to VPC addresses; ACL Control: Supports access control policies based on four-layer granularity (source / destination IP, port number, etc.); High Availability Architecture: Supports primary / backup or multi-active mode deployment to ensure service reliability.

[0100] Association with SmartNIC: SmartNIC can be mounted inside the TGW to perform part of the forwarding logic, hot flow direct processing, and acceleration channel establishment in the network card hardware; for the data path between VPC and IDC, the "control plane direct path" mode can be used, with SmartNIC directly completing packet forwarding.

[0101] 4. SmartNIC (Smart Network Interface Card) Function: Directly perform packet processing functions on the network card side, reducing the involvement of the central processing unit (CPU) and achieving microsecond-level traffic forwarding and policy execution.

[0102] Main functional modules: traffic access and analysis module: receives traffic from VPC or IDC, analyzes its layer 2, layer 3, and layer 4 header information, and performs flow table matching; hot flow detection module: uses data structures such as LRU (Least Recently Used), Sketch, or small hash table to record the byte count and packet rate of active flows, and determines whether it is a hot flow (Hot Flow) within a sliding time window (such as 5 seconds); local bandwidth splicing and allocation module: maintains real-time indicators of each available link (including utilization and remaining bandwidth), allocates tokens (Token) to flows according to weight or priority, and achieves millisecond-level response; direct connection cache table: when the flow meets the direct connection condition (identified as a hot flow, link health meets the standard, and the policy allows), a direct connection flow table entry is established, and subsequent matching packets are directly forwarded or tunnel encapsulated at the SmartNIC layer, without the need to report to the CPU; fallback and failure handling: when direct connection fails or the policy is revoked, the traffic automatically falls back to the CPU or normal path for processing; control plane interaction module: reports hot flow events through a low-latency channel (such as UDP or QUIC protocol), receives policy and flow table distribution in batches through a reliable channel (such as gRPC protocol), and supports policy version number comparison to avoid race conditions.

[0103] 5、Control Plane Function: responsible for unified management of policies, traffic scheduling, and event processing.

[0104] Core functions: global policy management: including bandwidth allocation (Bandwidth Pool Allocation, BWP allocation), priority model definition, and tenant quota management; event approval and automated processing: automatically or manually approving hot flow events reported by SmartNIC, and issuing direct connection or flow limiting policies; policy synchronization mechanism: supports rapid response to sudden hot flow events through low-latency links, and periodically synchronizes full policies to all SmartNICs through gRPC protocol; version control mechanism: assigns version number fields to all policies and flow table entries, and when SmartNIC detects version inconsistency, it can perform rollback or re-synchronization request; link health monitoring: aggregates link utilization, packet loss rate, and other indicators reported by each SmartNIC, providing decision basis for bandwidth splicing and traffic scheduling.

[0105] 6、Key algorithms and processes A. Hot flow detection Objective: quickly identify sessions (i.e. "hot flow") with high traffic proportion or sudden increase in frequency within a short period of time, and convert them to direct (DIRECT) mode or preferentially allocate local tokens.

[0106] Method (implemented in SmartNIC): Maintain an LRU cache, space-saving Sketch, or small hash table structure, and increase its count at each flow update; within a sliding time window W (e.g., 5 seconds), if the byte count of a flow is greater than a threshold THRESH_BYTE or the packet rate is greater than a threshold THRESH_PKT, mark it as a hot flow; when a flow is marked as a hot flow, trigger a local rate boost request and report a "hot event" to the control plane (Control).

[0107] B. Local fast bandwidth aggregation and allocation Objective: Aggregate bandwidth of a set of candidate links and allocate tokens for flows in SmartNIC, achieving millisecond-level response.

[0108] Design: SmartNIC maintains the metrics of N visible links (LinkMetric for each); SmartNIC allocates tokens for flows based on priority queues, and the allocation strategy can use Weighted Fair Share (according to tenant or flow weight) or Strict Priority (high priority flows are allocated first); for hot flows, SmartNIC attempts to allocate a fast token rate (token_rate), which is the minimum value of the requested rate and the available link capacity share; Local algorithm (simplified flow): Collect the utilization (link_i.utilization) and capacity (capacity_i) of each link; calculate the total remaining bandwidth: total_free = Σ[capacity_i × (1 - utilization_i)]; if the total remaining bandwidth is greater than or equal to the required rate (total_free ≥ required_rate), allocate the required bandwidth; otherwise, scale the allocation proportionally according to the priority.

[0109] C. Direct cache table establishment and flow forwarding (Direct Path) Condition determination (example): The flow is identified as a hot flow; the control plane policy allows the flow to be directly connected locally (e.g., without intermediate translation, audit, or specific ACL check); the local link and target path meet the delay and bandwidth constraints.

[0110] Table building process: SmartNIC generates flow entries (action as DIRECT) and broadcasts or issues them to cloud edge SmartNIC (across multi-point scenarios); subsequent matching packets are directly forwarded or tunnel encapsulated on the network card according to the action defined by the flow entry, without reporting to the CPU; fallback mechanism: if the flow entry expires, the network is abnormal, or the control plane receives a recovery instruction, the entry is immediately deleted, and subsequent traffic falls back to the CPU path processing.

[0111] D. Control plane coordination and policy distribution Control plane responsibilities: responsible for long-term policy management, including bandwidth pool allocation, priority model, and tenant quota; approval or approval (configurable automatic threshold policy to reduce manual intervention) for emergency events (such as "hot flow").

[0112] Distribution mechanism: low-latency channel: SmartNIC reports hot flow events to the control plane through UDP or QUIC protocol, and the control plane can quickly return approval (approve) or rejection (reject) results; reliable channel: full or batch synchronization of policies uses gRPC protocol transmission to ensure consistency; policy versioning: all policies and flow allocations carry version number fields, and when SmartNIC receives inconsistent policies, rollback or re-synchronization requests are executed to avoid race conditions.

[0113] 7、Example A. Scenario setup (example values) Deploy 3 bearer links to the cloud edge on the IDC side: Link-A (10 Gbps), Link-B (5 Gbps), Link-C (1 Gbps); the total capacity of the global bandwidth pool (BWP) is set by the control plane to 12 Gbps (can be oversold to 14 Gbps), and is allocated according to tenants, such as TenantX's basic allocation of 300 Mbps priority quota; the default minimum rate (min_rate) of the SmartNIC local token bucket is 100 Mbps, and the burst rate (burst) is 200 Mbps; hot flow judgment threshold: trigger when the flow is ≥50 MB or the packet number is ≥2000 in 5 seconds.

[0114] B. Running process (timeline) System startup and synchronization: the control plane distributes bandwidth pool allocation, tenant priority, and initial policies to IDC SmartNIC and cloud edge SmartNIC (full synchronization); SmartNIC locally establishes initial token bucket and link monitoring tasks.

[0115] Normal traffic distribution (non-hotspot flow): After the ordinary flow enters the IDC SmartNIC, it is matched as a regular flow, distributed to multiple links according to the hash or ECMP algorithm, or sent to the cloud central gateway after encapsulation; the CPU only processes the traffic that needs policy checking.

[0116] Sudden burst flow: A database backup flow transmits 120 MB of data in 5 seconds, which is marked as a hot flow by the SmartNIC hot flow detection module, and a flow table entry is created locally (the default action is DIRECT, and the token rate request is 400 Mbps); the SmartNIC calculates the current total remaining bandwidth (for example, Link-A has 2 Gbps, Link-B has 1 Gbps, and Link-C has 0.2 Gbps, totaling 3.2 Gbps), which can meet the request; the SmartNIC immediately locks the flow in direct mode locally and sends subsequent traffic fragments in proportion on Link-A / B / C (or preferentially selects Link-A, B); at the same time, the SmartNIC reports the hot flow event (including flow identification, observed rate and requested rate) to the control plane; the control plane automatically approves (or adjusts the temporary quota of the bandwidth pool if necessary) according to the global policy, and if approved, updates the metadata database and returns the version number to the SmartNIC (if a higher quota is required, it may trigger a global resource rebalancing process, which is usually completed in seconds).

[0117] Traffic persistence and fallback: If a link (such as Link-B) fails, the SmartNIC immediately recalculates and redistributes the fragments of the flow to the remaining links; if the remaining capacity is insufficient, the SmartNIC reduces the token rate and temporarily reports the excess traffic to the cloud central gateway (CPU) for degradation processing; when the hot flow is idle for more than the valid period (such as 30 seconds), the SmartNIC automatically recovers the flow table entry and reports the statistical information to the control plane for auditing and billing.

[0118] The application constructs an innovative architecture of "hardware acceleration + control plane cooperation + local decision" by deeply integrating an intelligent network card in a private line gateway architecture, and realizes millisecond-level dynamic scheduling of multi-link bandwidth resources and hardware direct forwarding of hot flow services. This method effectively solves the problems of high response delay, prominent CPU forwarding bottleneck, lack of service awareness ability and other problems in traditional software solutions, significantly improves the forwarding performance, resource utilization and service experience of the private line gateway in a hybrid cloud scenario, and provides a high-performance and high-reliability cloud network interconnection solution for enterprises.

[0119] Figure 4 An embodiment block diagram of a bandwidth allocation device provided by the embodiment of the application is shown in FIG. 1. Figure 4 As shown in the figure, the device comprises: The monitoring module 41 is configured to monitor network state indicators of a plurality of dedicated line links in real time. The identification module 42 is configured to perform heat analysis on traffic flows transmitted on the plurality of dedicated line links based on the network state indicators, and identify a hot flow satisfying a preset condition. The selection module 43 is configured to select a target link from the plurality of dedicated line links according to traffic characteristics of the hot flow and the network state indicators. The establishment module 44 is configured to establish a hardware direct connection forwarding channel for the hot flow on the target link, and synchronize configuration information of the hardware direct connection forwarding channel to a control component. The scheduling module 45 is configured to receive a global bandwidth allocation strategy issued by the control component based on the configuration information, and perform local traffic scheduling for the hot flow by the hardware direct connection forwarding channel according to the global bandwidth allocation strategy and the network state indicators.

[0120] In one possible implementation, the identification module is specifically configured to: extract bandwidth utilization rate data and transmission delay data of each dedicated line link from the network state indicators; statistically obtain byte number of each traffic flow based on the bandwidth utilization rate data and packet rate of each traffic flow based on the transmission delay data within a sliding time window; determine any traffic flow as a hot flow when the byte number of the traffic flow within the window exceeds a first threshold value or the packet rate exceeds a second threshold value.

[0121] In one possible implementation, the selection module is specifically configured to: extract residual bandwidth and current delay of each dedicated line link from the network state indicators; determine a dedicated line link as a candidate link to obtain a candidate link set when the residual bandwidth of the dedicated line link is greater than or equal to a bandwidth condition of the hot flow; determine a candidate link with the lowest current delay in the candidate link set as the target link.

[0122] In one possible implementation, the apparatus further includes an adjustment module configured to: extract residual available bandwidth of each dedicated line link from the network state indicators; when the residual available bandwidth is lower than a preset threshold value, re-calculate bandwidth sharing weights of each traffic flow according to a traffic priority level defined in the global bandwidth allocation strategy, and adjust a flow table output port configuration of the hardware direct connection forwarding channel according to the re-calculated bandwidth sharing weights.

[0123] In one possible implementation, the scheduling module is configured to: receive the approval result of the control component on the hardware direct forwarding channel through a first channel, the first channel using a low-delay transmission protocol; receive the complete parameters of the global bandwidth allocation strategy issued by the control component through a second channel, the second channel using a reliable transmission protocol; when detecting that the version of the locally stored bandwidth allocation strategy is inconsistent with the version of the strategy of the control component, initiate a resynchronization request to the control component.

[0124] In one possible implementation, the apparatus further includes a reselection module configured to: when detecting that the remaining available bandwidth of the target link continuously falls below a failure threshold for a preset length of time, terminate the transmission of the hot flow through the hardware direct forwarding channel, and reselect a backup link from the multiple dedicated line links to establish a new hardware direct forwarding channel, and report a channel switching event to the control component.

[0125] In one possible implementation, the apparatus further includes an execution module configured to: when none of the backup links meets the bandwidth requirement, unbind the hardware direct forwarding channel of the hot flow, and perform the packaging and forwarding operation of the hot flow through a host central processing unit.

[0126] In one possible implementation, the scheduling module is specifically configured to: virtualize the physical bandwidth resources of the multiple dedicated line links into a unified global bandwidth pool, wherein the real-time total capacity of the global bandwidth pool is dynamically adjusted based on the link bandwidth utilization rates in the network state indicators; determine the bandwidth token allocation quota of the hot flow from the global bandwidth pool based on the tenant quota and service priority parameters defined in the global bandwidth allocation strategy, the bandwidth token allocation quota being positively correlated with the real-time remaining bandwidth of the target link where the hot flow is located; configure the token bucket parameters in the flow table items corresponding to the hardware direct forwarding channel according to the bandwidth token allocation quota; allocate exclusive bandwidth tokens for the hot flow based on the token bucket parameters through the hardware direct forwarding channel and implement token rate-limiting forwarding.

[0127] As shown in Figure 5 The embodiments of the present application provide a device, which comprises a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112 and the memory 113 complete mutual communication through the communication bus 114, The memory 113 is used for storing a computer program. In an embodiment of the present application, the processor 111, when executing the program stored in the memory 113, implements the bandwidth allocation method provided by any one of the foregoing method embodiments, including: monitoring network state indexes of a plurality of dedicated line links in real time; performing heat analysis on service flows transmitted on the plurality of dedicated line links based on the network state indexes, and identifying hot flows satisfying preset conditions; selecting a target link from the plurality of dedicated line links according to traffic characteristics of the hot flows and the network state indexes; establishing a hardware direct connection forwarding channel for the hot flows on the target link, and synchronizing configuration information of the hardware direct connection forwarding channel to a control component; receiving a global bandwidth allocation strategy issued by the control component based on the configuration information, and performing local traffic scheduling for the hot flows by the hardware direct connection forwarding channel according to the global bandwidth allocation strategy and the network state indexes.

[0128] The present application also provides a computer readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the bandwidth allocation method provided by any one of the foregoing method embodiments.

[0129] The apparatus embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment.

[0130] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be implemented by means of software plus a general hardware platform, and of course can also be implemented by hardware. Based on such understanding, the above technical solutions or the part that contributes to the related art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in the various embodiments or some parts of the embodiments.

[0131] It is to be understood that the terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. The terms "comprises", "comprising", "includes", "including" and "has" are inclusive and therefore specify the presence of stated features, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof. The method steps, processes, and operations described herein are not to be construed as necessarily requiring their performance in the particular order

[0132] The above description is that of current embodiments of the application. Various modifications and changes can be made thereto without departing from the spirit and scope of the application as set forth. The scope of the application is not to be limited to the exact details shown above.

Claims

1. A bandwidth allocation method, characterized in that, Applied to smart network interface cards, the method includes: Real-time monitoring of network status indicators for multiple dedicated line links; Based on the network status indicators, the heat analysis of service flows transmitted on multiple leased links is performed to identify hot flows that meet preset conditions. Based on the flow characteristics of the heat flow and the network status indicators, a target link is selected from multiple leased links; A hardware direct forwarding channel is established for the heat flow on the target link, and the configuration information of the hardware direct forwarding channel is synchronized to the control component; The receiving control component sends a global bandwidth allocation policy based on the configuration information, and according to the global bandwidth allocation policy and the network status indicators, calls the hardware direct forwarding channel to perform local traffic scheduling for the hot flow.

2. The method according to claim 1, characterized in that, The step of performing heat analysis on service flows transmitted on multiple leased links based on the network status indicators to identify hot flows that meet preset conditions includes: Extract bandwidth utilization data and transmission latency data for each dedicated line link from the network status indicators; Within the sliding time window, the number of bytes for each service stream is calculated based on the bandwidth utilization data, and the packet rate for each service stream is calculated based on the transmission delay data. When the number of bytes in any service flow within the window exceeds a first threshold, or the packet rate exceeds a second threshold, the service flow is determined to be a hot flow.

3. The method according to claim 1, characterized in that, The step of selecting a target link from multiple leased links based on the flow characteristics of the heat flow and the network status indicators includes: Extract the remaining bandwidth and current latency of each dedicated line link from the network status indicators; If the remaining bandwidth of any leased link is greater than or equal to the bandwidth of the heat flow, the leased link is identified as a candidate link, and a candidate link set is obtained. The candidate link with the lowest current latency in the candidate link set is determined as the target link.

4. The method according to claim 1, characterized in that, After invoking the hardware direct forwarding channel to perform local traffic scheduling for the hot flow based on the global bandwidth allocation strategy and the network status indicators, the method further includes: Extract the remaining available bandwidth of each leased link from the network status indicators; When the remaining available bandwidth is lower than a preset threshold, the bandwidth sharing weight of each service flow is recalculated according to the service priority level defined in the global bandwidth allocation strategy, and the flow table output port configuration of the hardware direct forwarding channel is adjusted according to the recalculated bandwidth sharing weight.

5. The method according to claim 1, characterized in that, The global bandwidth allocation strategy issued by the receiving control component based on the configuration information includes: The control component receives the approval result of the hardware direct forwarding channel through the first channel, and the first channel adopts a low-latency transmission protocol. The system receives the complete parameters of the global bandwidth allocation strategy issued by the control component through a second channel, which employs a reliable transmission protocol. When a discrepancy is detected between the bandwidth allocation policy version of the local storage and the policy version of the control component, a resynchronization request is initiated to the control component.

6. The method according to claim 1, characterized in that, The method further includes: When the remaining available bandwidth of the target link is detected to be continuously lower than the fault threshold for a preset time, the transmission of the hot flow through the hardware direct forwarding channel is terminated, and a new hardware direct forwarding channel is established by reselecting a backup link from multiple leased links, and the channel switching event is reported to the control component.

7. The method according to claim 6, characterized in that, The method further includes: When all backup links fail to meet bandwidth requirements, the hardware direct-connection forwarding channel binding of the hot flow is released, and the encapsulation and forwarding operation of the hot flow is performed by the host central processing unit.

8. The method according to claim 1, characterized in that, The step of invoking the hardware direct forwarding channel to perform local traffic scheduling for the hot flow based on the global bandwidth allocation strategy and the network status indicators includes: The physical bandwidth resources of multiple dedicated line links are virtualized into a unified global bandwidth pool, wherein the real-time total capacity of the global bandwidth pool is dynamically adjusted based on the bandwidth utilization of each link in the network status indicators. Based on the tenant quota and service priority parameters defined in the global bandwidth allocation strategy, the bandwidth token allocation quota for the hot flow is determined from the global bandwidth pool. The bandwidth token allocation quota is positively correlated with the real-time remaining bandwidth of the target link where the hot flow is located. Configure the token bucket parameters in the flow table entry corresponding to the hardware direct forwarding channel according to the bandwidth token allocation quota; The hardware direct forwarding channel allocates dedicated bandwidth tokens to the hot flow based on the token bucket parameters and implements token rate limiting forwarding.

9. A bandwidth allocation device, characterized in that, The device, applied to a smart network interface card, includes: The monitoring module is used to monitor the network status indicators of multiple leased links in real time. The identification module is used to perform heat analysis on the service flows transmitted on multiple leased links based on the network status indicators, and identify the hot flows that meet the preset conditions. The selection module is used to select a target link from multiple leased links based on the flow characteristics of the heat flow and the network status indicators. A module is established to create a hardware direct forwarding channel for the heat flow on the target link and to synchronize the configuration information of the hardware direct forwarding channel to the control component. The scheduling module is used to receive the global bandwidth allocation policy issued by the control component based on the configuration information, and, according to the global bandwidth allocation policy and the network status indicators, call the hardware direct forwarding channel to perform local traffic scheduling for the hot flow.

10. An electronic device, characterized in that, include: A processor and a memory, the processor being configured to execute a bandwidth allocation program stored in the memory to implement the bandwidth allocation method according to any one of claims 1-8.

11. A storage medium, characterized in that, The storage medium stores one or more programs, which can be executed by one or more processors to implement the bandwidth allocation method according to any one of claims 1-8.

Images