Distributed network bandwidth control system

Abstract

The application provides a distributed network bandwidth control system, aiming at solving the problems of coarse granularity, difficult management, poor dynamic adjustment and weak integration of the traditional bandwidth control scheme. The system adopts a "central control platform plus distributed agent program module" architecture, the central control platform provides a Web interface and an API interface, realizes centralized storage, distribution and agent program state monitoring of IP speed limiting strategies (including IP, uplink and downlink bandwidth and priority); the agent program is deployed at the network exit node, calls the Linux TC tool to realize single IP fine-grained traffic shaping, and sends heartbeat synchronization state at regular intervals, and then the agent program module and the related nodes are fed back in this process, forming a complete control closed loop.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network communication, and in particular to a distributed network bandwidth control system. BACKGROUND

[0002] In the field of network communication, bandwidth control is a key means to guarantee network service quality and rationally allocate network resources. With the continuous expansion of network scale and the increasing complexity of business types, the traditional bandwidth control scheme gradually exposes many problems.

[0003] Traditional speed limiting schemes such as overall link speed limiting and QoS strategy can usually only limit bandwidth for the overall link or specific business types, and have a coarse granularity, and thus cannot achieve accurate bandwidth control for a single IP, and cannot meet the bandwidth isolation needs of businesses for individual users or devices. In a large-scale network environment, if each node independently configures a bandwidth strategy, not only is the maintenance cost high, but also it is difficult to guarantee the consistency of the strategies of the nodes, and configuration drift phenomenon is likely to occur.

[0004] At the same time, business traffic has obvious dynamics, and there are peak and low periods, and the traditional scheme cannot support real-time adjustment of bandwidth limitation, cannot adapt to the dynamic changes of business traffic, and has insufficient automated operation and maintenance capabilities. In addition, traditional speed limiting tools such as the tc command line are mostly in command line operation mode, and it is difficult to link with an upper business system, has poor cross-system integration capability, has insufficient programmability and extensibility, and cannot meet the needs of modern network management for automation and intelligence.

[0005] Although commercial bandwidth control products are relatively perfect in function, they have the problem of high cost, usually require an investment of hundreds of thousands of yuan, and are mostly closed-source systems, have a closed interface, are uncontrollable in technology, and are difficult to deeply customize according to actual business needs. The SDN (Software Defined Network) controller scheme supports centralized management and a certain extensibility, but is complex to deploy, has a medium-high cost, and depends on the open source community, and for small and medium-sized network or edge node deployment scenarios, it is too cumbersome and has insufficient flexibility.

[0006] Therefore, there is an urgent need for a distributed network bandwidth control system scheme that is low in cost, self-controllable, lightweight and flexible, high in automation, and easy to implement, to solve the above problems in the prior art. SUMMARY

[0007] The present application aims to overcome the deficiencies of the prior art, and provides a distributed network bandwidth control system to achieve accurate bandwidth control for a single IP, reduce maintenance costs, guarantee strategy consistency, support dynamic adjustment of bandwidth and automated operation and maintenance, improve cross-system integration capability, reduce costs, and achieve technology self-controllability.

[0008] Specifically, the application provides a distributed network bandwidth control system, which comprises a central control platform and a plurality of distributed agent program modules, wherein the central control platform is configured with an API interface module for data interaction and instruction acceptance, the agent program modules are deployed at network exit nodes requiring bandwidth control, starting the agent program modules automatically sends a registration request to the central control platform, the central control platform verifies the identity of the agent program modules and allocates a unique agent identifier to the agent program modules after receiving the registration request, and a communication connection between the central control platform and the agent program modules is established; an external device sends a configuration request of an IP speed limiting strategy to the central control platform by calling the API interface, the central control platform verifies the speed limiting strategy after receiving the request, and the central control platform issues the encrypted IP speed limiting strategy to a target agent program module after the verification, the target agent program module decrypts the encrypted IP speed limiting strategy and parses the strategy content to extract speed limiting information after receiving the encrypted IP speed limiting strategy, and then configures a corresponding flow control rule according to the speed limiting information to execute the IP speed limiting strategy on the external device, the target agent program module collects state information of a local network where the external device is located at a preset time interval, then encapsulates the state information and sends the state information to the central control platform in the form of a heartbeat message, and the central control platform parses the heartbeat message to monitor the strategy execution state of the target agent program module after receiving the heartbeat message, wherein the target agent program module dynamically adjusts the initial time interval by real-time feedback according to the load rate and interaction with the central control platform through a basic threshold dynamic adaptation mechanism.

[0009] Preferably, the preset mode of the initial time interval is that the central control platform allocates an initial time interval within a basic interval [MinT, MaxT] according to the node type of the network exit node If the network exit node is a core node, a first initial time interval T is allocated If the network exit node is a normal access node, a second initial time interval T' is allocated If the network exit node is a light edge node, a third initial time interval T'' is allocated , wherein the first initial time interval T , the second initial time interval T' , and the third initial time interval T'' are within the basic interval [MinT, MaxT] and satisfy T < T' < T'' .

[0010] Preferably, in the core gateway node, the number of CPU cores is greater than eight and the gateway node memory is greater than or equal to 32G; in the general access node, the number of CPU cores is four to eight and the gateway node memory is between 8GB and 32GB; in the light edge node, the number of CPU cores is less than four and the gateway node memory is less than 8GB.

[0011] Preferably, the "target agent module dynamically adjusts the initial time interval through a basic threshold mechanism according to the load rate and interacts with the central control platform to feedback and adjust the initial time interval in real time" comprises: the target agent module obtains the initial time interval Then, the load rate A of the network exit node is calculated, and the percentage constants a and b are set, wherein b>a, if the load rate A of the network exit node is less than a, it indicates that the network exit node is in low load, and then the initial time interval is adjusted to T = T + (a-T) / (a-b) * (1+A) ; if the load rate A of the network exit node satisfies a<=A<=b, it indicates that the network exit node is in medium load, and in this case, the initial time interval is kept unchanged; if the load rate A of the network exit node is greater than b, it indicates that the network exit node is in high load, and then the initial time interval is adjusted to T = T - (T-b) / (b-a) * (1+A). b.

[0012] Preferably, the calculation formula of the load rate is: load rate = current active TCP connection number / node maximum bearing connection number*100%.

[0013] Optionally, the time interval basic interval is [MinT, MaxT] and is [10 seconds, 180 seconds].

[0014] In summary, the application discloses a kind of distributed network bandwidth control systems, to solve the problems of traditional bandwidth control scheme, such as coarse granularity, difficult management, poor dynamic adjustment and weak integration. The system uses the architecture of "central control platform + distributed agent module", the central control platform provides Web interface and API interface, realizes the centralized storage, issue and agent state monitoring of IP speed limiting strategy (including IP, uplink and downlink bandwidth, priority);Agent program is deployed in network exit node, and single IP fine-grained traffic shaping is realized by calling Linux TC tool, and heartbeat synchronization state is sent regularly, and then strategy feedback is carried out on agent module and related nodes in the process, forming a complete control loop. BRIEF DESCRIPTION OF DRAWINGS

[0015] ​In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will describe the drawings needed to be used in the embodiments or prior art description. Obviously, the technical solutions described in the description in combination with the drawings are only some embodiments of the present application, and for those skilled in the art, other embodiments and drawings can be obtained from the embodiments shown in the drawings without creative labor.

[0016] Figure 1 A basic architecture diagram of the distributed network bandwidth control system according to the present application is shown.

[0017] Figure 2 A flowchart of network state information adjustment in the distributed network bandwidth control system according to the present application is shown. DETAILED DESCRIPTION

[0018] The technical solutions of the embodiments of the present application will be described in detail below in combination with the drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments described in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0019] In general, to solve the prior art, the present application proposes a distributed network bandwidth control system based on the architecture of “central control platform + distributed Agent (agent program module)”.

[0020] Figure 1 A basic architecture diagram of the distributed network bandwidth control system according to the present application is shown.

[0021] As shown in Figure 1 , in the operation process of the system, the central control platform is first initialized and configured. The central control platform builds a Web operation interface for visual operation by management personnel, including creation, editing, deletion of IP speed limiting strategy and viewing of system state.

[0022] Then, the central control platform builds a strategy database for storing IP speed limiting strategy information, including target IP address, uplink bandwidth limit value, downlink bandwidth limit value, strategy priority and strategy effective time range.

[0023] Subsequently, the central control platform configures an API interface module supporting RESTful API protocol for receiving API call requests from external systems to realize data interaction and instruction reception with external systems.

[0024] Next, Agent deployment and registration are performed. First, the Agent program is deployed on network egress nodes requiring bandwidth control, such as gateways, routers, and virtual machine hosts. After the Agent program is started, it automatically sends a registration request to the central control platform. The registration request carries the device information of the node where the Agent is located, including the node IP address, device model, operating system version, and network interface information.

[0025] Subsequently, the central control platform receives the registration request from the Agent, verifies the Agent's identity, and once the verification is successful, includes the Agent in the management scope, assigns it a unique Agent identifier, and establishes a communication connection between the central control platform and the Agent.

[0026] The next step is policy distribution and execution. IP rate limiting policies are created through the central control platform's interface, or external devices send IP rate limiting policy configuration requests by calling the central control platform's API. Upon receiving the IP rate limiting policy configuration request, the central control platform verifies the policy's legality and completeness. Verification includes checking if the target IP address format is correct, if the bandwidth limit is within a reasonable range, and if the policy priority is set appropriately.

[0027] After verification, the central control platform determines the target agent for which the IP rate limiting policy needs to be issued based on the management scope of each agent, and encrypts the IP rate limiting policy according to the preset encryption method.

[0028] Subsequently, the central control platform sends the encrypted IP rate limiting policy to the target agent, while recording information such as the policy sending time, target agent identifier, and policy content.

[0029] The target agent receives the encrypted IP rate limiting policy issued by the central control platform. It first decrypts the policy, then parses the policy content to extract key rate limiting information such as the target IP address, uplink / downlink bandwidth limit value, and policy priority.

[0030] Next, the Agent configures the corresponding traffic control rules based on the parsed policy information. For example, it calls the Linux TC tool and uses the htb (Hierarchical Token Bucket) and htbfairqueue algorithms to configure the corresponding traffic control rules, thereby achieving per-IP traffic shaping for the target IP and thus completing the execution of the IP rate limiting policy.

[0031] The next step is state monitoring and heartbeat interaction. In this step, the Agent collects local network state information and policy execution state information at a preset time interval (such as every 30 seconds), the network state information includes the CPU usage of the node, the memory occupancy, the throughput of the network interface, the packet loss rate, the delay, etc., the policy execution state information includes the list of executed IP speed limiting policies, the actual bandwidth control effect of each policy (such as the actual uplink / downlink bandwidth value), whether the policy is normally effective, etc.

[0032] Figure 2 A flowchart of network state information adjustment under the center heartbeat interaction of the distributed network bandwidth control system according to the application is shown.

[0033] As Figure 2 shown, the adjustment of the above-mentioned time interval is particularly important in the overall system. The initial time interval (denoted as ) of the Agent collecting network state and policy execution state is not a fixed value, but is determined by the mechanism of “basic threshold + dynamic adaptation”, taking into account the monitoring accuracy and system resource consumption.

[0034] Firstly, the central control platform presets the time interval basic interval for the Agent as [MinT, MaxT] (for example, [10 seconds, 180 seconds]), which is based on the performance characteristics of the Linux kernel, the timeliness requirements of network state data, and the Agent resource consumption test:

[0035] The setting of the lower limit MinT (for example, 10 seconds): to avoid too short interval causing Agent to collect data frequently, causing node CPU occupancy rate to exceed a certain percentage, for example, 5% (thousands of flow scenarios), while ensuring that network anomalies with millisecond-level fluctuations (such as instantaneous packet loss) can be captured;

[0036] The setting of the upper limit MaxT (for example, 180 seconds): to avoid too long interval causing the central control platform to fail to discover Agent offline or policy execution anomaly (such as bandwidth overrun lasting for 3 minutes without being detected) in time, meeting the high availability requirement.

[0037] After the Agent is deployed, it first reports the “node type” (determined by node hardware configuration and network role) of the node where it is located to the central control platform through registration, and the central control platform allocates the initial time interval for the Agent within the basic interval [MinT, MaxT] according to the node type.

[0038] The specific adaptation mode is: for the core gateway node, for example, the CPU core number is greater than eight and the gateway node memory is greater than or equal to 32 GB, which means that it is an enterprise export gateway or a cloud service provider edge gateway (often carrying ten thousand level IP traffic), and the initial time interval allocated is = 30 seconds, ensuring that the node performance is sufficient, and high-frequency monitoring is required to ensure the stability of the core link.

[0039] For ordinary access nodes, for example, the CPU core number is four to eight and the gateway node memory is between 8 GB and 32 GB, which means it is a department-level gateway or a virtual machine host, carrying less than one thousand level IP traffic, and the initial time interval allocated is = 60 seconds, ensuring that the node performance is moderate, balancing monitoring accuracy and resource consumption.

[0040] For light edge nodes, for example, the CPU core number is less than four and the gateway node memory is less than 8 GB, which means it is a small branch gateway or an IoT device aggregation node, carrying less than one hundred level IP traffic, and the initial time interval allocated is = 120 seconds, in which case the node performance is limited, and the Agent resource occupation is reduced in priority to avoid affecting business.

[0041] The Agent starts and obtains the initial time interval After that, the node "initial network load rate" is calculated synchronously when collecting data for the first time (calculation formula: load rate = current active TCP connection number / node maximum carrying connection number x 100%), and the is calibrated again according to the load rate.

[0042] Specifically, if the load rate A < a (a is a percentage numerical constant, for example, a = 30%), it is considered low load, and the is adjusted to time interval T= (1+A), for example, the percentage numerical constant a = 30%, and the load rate A = 20%, and the initial time interval = 30 seconds, then T = 30 (1+20%) = 36 seconds, and then the calculated time interval T replaces the original time interval , further reducing resource consumption.

[0043] If the load rate A satisfies a ≤ A ≤ b (b is another percentage numerical constant, b > a, for example, b = 70%), for example, if 30% ≤ A ≤ 70%, it is considered medium load, in which case T= , that is, the initial interval time is kept unchanged.

[0044] If the load rate A satisfies A > b, it is considered high load, and Down to T= b, for example A=80%>70%, is considered high load, set =60 seconds, T=60 70%=42 seconds, and the calculated time interval T is replaced with the original time interval to enhance the monitoring frequency to capture bandwidth anomalies under high load.

[0045] The Agent encapsulates the collected state information in a preset format and sends it to the control platform in the form of a heartbeat message. The control platform receives the heartbeat message sent by the Agent, parses the state information therein, and monitors the survival state and policy execution state of the Agent.

[0046] If the control platform does not receive the heartbeat message of a certain Agent within a preset time threshold (e.g., 90 seconds), it is determined that the Agent may have a fault, and the control platform immediately triggers an alarm mechanism to send alarm information to the administrator through SMS, email, or system notification, etc. The alarm information includes the identification of the faulty Agent, the node information, and the fault occurrence time, etc.

[0047] If the control platform discovers through parsing the heartbeat message that the IP rate limiting policy execution effect on a certain Agent does not meet the expectation, such as the actual bandwidth control error exceeds 5%, the control platform will automatically generate a policy adjustment suggestion, or according to the preset automatic adjustment rules, issue a policy adjustment instruction to the Agent, and the Agent receives the instruction and adjusts the corresponding traffic control rules to ensure that the bandwidth control effect meets the expectation.

[0048] Subsequently, the step of policy dynamic adjustment and update is performed. When the business requirement changes and the IP rate limiting policy needs to be adjusted, the administrator can edit the existing IP rate limiting policy through the Web operation interface of the control platform, or the external system sends a policy adjustment request by calling the API interface of the control platform. After receiving the policy adjustment request, the control platform verifies the adjusted policy according to the policy verification process described above, and after verification, issues the adjusted policy to the corresponding Agent.

[0049] The Agent receives the adjusted policy, updates the local traffic control rules according to the policy execution process described above, realizes the dynamic adjustment of the policy, and after the policy is changed, completes the issuance and application within a certain period of time (e.g., 1 second).

[0050] If it is needed to delete an IP speed limiting policy, the administrator sends a policy deletion request through a web operation interface or an external system through an API interface, the central control platform receives the request, and sends a policy deletion instruction to the corresponding agent, the agent receives the instruction, and deletes the corresponding traffic control rule, and stops the bandwidth limitation on the target IP.

[0051] In general, the application discloses a kind of distributed network bandwidth control systems, to solve the problems of traditional bandwidth control scheme, such as coarse granularity, difficult management, poor dynamic adjustment and weak integration. The system uses the architecture of "central control platform + distributed agent program module", the central control platform provides Web interface and API interface, realizes the centralized storage, issue and agent program state monitoring of IP speed limiting policy (including IP, uplink and downlink bandwidth, priority);Agent program is deployed in network exit node, and single IP fine-grained traffic shaping is realized by calling Linux TC tool, and heartbeat synchronization state is sent regularly, and then policy feedback is carried out on agent program module and related nodes in the process, forming a complete control loop.

[0052] The above only describes exemplary embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A distributed network bandwidth control system, characterized by, The system comprises a central control platform and a plurality of distributed agent modules, wherein, The central control platform is configured with an API interface module for data interaction and instruction acceptance, An agent module is deployed at a network exit node requiring bandwidth control, and starting the agent module automatically sends a registration request to the central control platform, the central control platform verifies the identity of the agent module and allocates a unique agent identifier to it after receiving the registration request, and establishes a communication connection between the central control platform and the agent module; An external device sends a configuration request of an IP speed limit policy to the central control platform by calling the API interface, the central control platform verifies the speed limit policy after receiving the request, and after verification, the central control platform issues the encrypted IP speed limit policy to the target agent module, The target agent module receives the encrypted IP speed limit policy, decrypts it and parses the policy content to extract the speed limit information, and then configures the corresponding traffic control rule according to the speed limit information to execute the IP speed limit policy on the external device, The target agent module collects state information of the local network where the external device is located at a preset time interval, then encapsulates the state information and sends it to the central control platform in the form of a heartbeat message, and the central control platform parses the heartbeat message to monitor the policy execution state of the target agent module after receiving it, The target agent module adjusts the initial time interval in real time through the basic threshold dynamic adaptation mechanism according to the load rate and interacts with the central control platform to feedback, The target agent module adjusts the initial time interval in real time through the basic threshold dynamic adaptation mechanism according to the load rate and interacts with the central control platform to feedback, The target agent module obtains an initial time interval After, the load rate A of the network exit node is calculated, and the percentage constants a and b are set, where b > a, If the load rate A of the network egress node is less than a, which means that the network egress node is in low load, the initial time interval T is set to T = T + T0. Up-regulation is ; If the load rate A of the network egress node satisfies a ≤ A ≤ b, it means that the network egress node is in medium load, then the initial time interval is kept in this case unchanged; If the load rate A of the network egress node is greater than b, it indicates that the network egress node is in high load, and the initial time interval is reduced to .

2. The system of claim 1, wherein, The preset mode of the initial time interval is: The central control platform allocates an initial time interval within the base interval [MinT, MaxT] according to the node type of the network exit node , If the network egress node is a core nexus node, then assign a first initial time interval , If the network egress node is a normal access node, a second initial time interval is assigned , If the network egress node is a lightweight edge node, a third initial time interval is assigned , wherein the first initial time interval , the second initial time interval , the third initial time interval are all within the base interval [MinT, MaxT] and satisfy .

3. The system of claim 2, wherein, In the core gateway node, the number of CPU cores is greater than eight and the gateway node memory is ≥32G; in the ordinary access node, the number of CPU cores is four to eight and the gateway node memory is between 8GB and 32GB; in the light edge node, the number of CPU cores is less than four and the gateway node memory is less than 8GB.

4. The system of claim 1, wherein, The formula for calculating the load rate is: load rate = current active TCP connection number / node maximum load connection number × 100%.

5. The system of claim 2, wherein, The time interval basic interval is [MinT, MaxT] = [10 seconds, 180 seconds].

Images