Flow control method and device based on feedback distribution and applied to CDN machine room, and medium

Abstract

The invention discloses a feedback distribution-based flow control method and device applied to a CDN machine room, and a medium, and belongs to the technical field of distributed computing systems, and the method comprises the steps: obtaining bandwidth data of each OC and an output port switch in the CDN machine room at a current control interval, and carrying out the preprocessing, the bandwidth data comprises the rate of the OC service business and the rate of the output port switch; carrying out aggregation statistics on the preprocessed bandwidth data of the OC according to an IP (Internet Protocol) or a service; and judging whether the rate of the port switch exceeds a preset threshold value or not, if so, obtaining an overall speed limit value of a next control interval machine room through a feedback control algorithm, and distributing the overall speed limit value to each OC according to the required rate of each OC / various services in the machine room and a configuration strategy. According to the invention, aiming at the problem that the outlet port of the existing CDN machine room exceeds the limit, fine-grained flow regulation and control can be carried out from the CDN machine room level, so that the overall performance of the CDN machine room is improved.

Description

A feedback-based flow control method, device, and medium for CDN data centers. Technical Field

[0001] This invention belongs to the field of distributed computing system technology, specifically relating to a feedback-based traffic control method, device, and medium for CDN data centers. Background Technology

[0002] A Content Delivery Network (CDN) is a distributed service and infrastructure system. By deploying a large number of caching / acceleration nodes at the network edge, it delivers content or services to users nearby, thereby reducing latency, alleviating the load on the origin server, and improving availability and reliability. CDN edge nodes (OC) are often orchestrated on a data center basis. A single data center may have one or more outgoing port switches, depending on its size. All traffic from the data center communicates with the public network through this outgoing port switch. Therefore, the traffic limit through this outgoing port switch is limited by its physical capacity. If the traffic exceeds the switch's capacity, frequent packet loss will occur. Furthermore, besides construction costs, a significant portion of CDN operating costs comes from the cost of purchasing traffic from the network operator accessing the data center's egress. This cost is primarily billed using 95% scheduling, meaning that the actual billed bandwidth is the 95th percentile of traffic within a month. Effectively controlling the traffic at the data center's outgoing ports can significantly reduce packet loss due to exceeding limits and prevent cost increases caused by excessively high 95% percentiles. Therefore, a control system is urgently needed to limit the external traffic of the data center.

[0003] Current CDN systems typically rely on a globally covering DNS scheduling layer for traffic routing. This layer directs user requests to various data centers by writing different IPs to different Local DNS servers. Specifically, this DNS scheduling process can be broken down into three stages: profiling, prediction, and allocation. First, profiling (historical and measurement data such as user / region / network operator / time / content type, etc.) characterizes the behavioral features of each Local DNS server and its user base. Second, predictions are made based on these profiles (e.g., predicting how many requests, bandwidth, and concurrency, as well as latency or hit rate changes, will be introduced in a future time window by writing a certain exit IP to a particular Local DNS server). Finally, in the allocation stage, the prediction results are combined with constraints (data center capacity, link bandwidth, SLA, cost, cache hit rate, load balancing strategies, etc.) to determine which IP or set of IPs and their corresponding weights should be sent to that Local DNS server. DNS-based scheduling allows for macro-level control, rationally distributing data bandwidth across different data centers according to cost and performance aggregation requirements. However, DNS-based scheduling is a relatively coarse-grained scheduling method. Whether it is prediction or allocation, it is difficult to accurately ensure the actual traffic entering the data center, which often leads to the data center exceeding the outgoing port limit, resulting in packet loss and increased costs.

[0004] Furthermore, the dynamic and diverse nature of traffic, primarily including short-term traffic surges and large file transfers, makes outbound bandwidth load more severe and time-varying. The reason for these two types of traffic can be understood as follows: the traffic data predicted for the next moment based on traffic profiles from the previous period is an empirical value from previous profiles. However, traffic characteristics are dynamically changing. For example, sudden bursts of regionally popular traffic or a sudden increase in the proportion of large file requests can cause the empirically based profile value to be lower than the actual traffic value. This can lead to sudden over-limits in data centers that are already operating at peak times or under high pressure, resulting in frequent packet loss and increased costs. Therefore, a data center-based rate limiting system is needed to smooth outbound traffic and reduce the risk of over-limits. Summary of the Invention

[0005] This invention addresses the shortcomings of existing technologies by providing a feedback-based traffic control method, device, and medium for CDN data centers. It enables fine-grained traffic regulation at the CDN data center level to address the issue of excessive outgoing ports in existing CDN data centers, thereby improving the overall performance of CDN data centers.

[0006] This invention provides the following technical solution: Firstly, a feedback-based traffic control method for CDN data centers is provided, comprising the following steps: Step 1: In the current control interval, bandwidth data of each OC and outgoing port switch within the CDN data center is acquired and preprocessed. The bandwidth data includes the rate of the OC service and the rate of the outgoing port switch; Step 2: The preprocessed bandwidth data of the OC is aggregated and statistically analyzed by IP or service to obtain the required rate of each OC or the required rate of each type of service under the current control interval; Step 3: It is determined whether the rate of the outgoing port switch exceeds a preset threshold. If it does, the overall rate limit value of the data center in the next control interval is obtained through a feedback control algorithm, and the overall rate limit value is allocated to each OC according to the required rate and configuration strategy of each OC / type of service within the data center; Step 4: Rate limit configuration is pushed to each OC to enable each OC to perform rate limit operations.

[0007] Optionally, in step 1, the bandwidth data of the OC under the current control interval is bandwidth data in the dimensions of timestamp-IP-host-speed, where timestamp is the timestamp within the current control interval, IP is the OC identifier, host is the Uniform Resource Locator of the service served by the OC within the current control interval, and OC speed is the rate, representing the bandwidth value of the service served by the OC within the current control interval; the bandwidth data of the outgoing port switch under the current control interval is bandwidth data in the dimensions of timestamp-name-speed, where timestamp is the timestamp within the current control interval, name is the data center identifier, and outgoing port switch speed is the rate, representing the bandwidth value of the entire data center to the outside world within the current control interval.

[0008] Optionally, in step 1, the preprocessing includes: filtering abnormal and expired data based on the validity of the timestamp and rate fields of the bandwidth data. Specifically, based on the received bandwidth data of the OC and outgoing port switch, if the time corresponding to the timestamp is not within the current control interval, the current bandwidth data is filtered; otherwise, the speed dimension data is checked. If the rate is less than or equal to 0, it is illegal data and is filtered; otherwise, it is normal data and is cached.

[0009] Optionally, step 2 specifically involves: traversing the bandwidth data of each OC within the current time interval, locating the data center to which the IP of each OC belongs by searching the IP-data center name database, and classifying the bandwidth data of all OCs according to the data center; for each data center, accumulating the rate data corresponding to OCs with the same IP to obtain the required rate of each OC under the current control interval; for each data center, accumulating the rate data of services with the same service to obtain the required rate of each type of service under the current control interval.

[0010] Optionally, in step 3, the overall speed limit value of the computer room for the next control interval is obtained through a feedback control algorithm, specifically by obtaining the output of the feedback control algorithm in the current control interval according to the following formula. ;

[0011] in, For proportional gain, For differential gain, and In the current control interval and the previous control interval Within, the amount by which the outgoing port switch rate exceeds a preset threshold; the following formula maps the output of the feedback control algorithm in the current control interval to the overall room rate limit value for the next control interval. ; ;in, This is the preset maximum speed limit offset. This is the overall speed limit value for the computer room in the previous control interval. This is the overall speed limit value for the current control interval of the computer room.

[0012] Optionally, in step 3, the overall rate limit value is allocated to each OC according to the demand rate and configuration strategy of each OC. Specifically, the demand rate of all OCs in the data center is smoothed using the EWMA method, and the smoothed demand rate of each OC and the overall rate limit value of the data center are used as initial values. The allocation rate of each OC is obtained through iterative loop. The allocation rate of each OC is obtained through iterative loop, which specifically includes the following steps: Step a: Select the current demand rate of all OCs in the demand list. The minimum value in is denoted as ,like Then assign to each OC The rate is determined, and step b is executed; otherwise, each OC is allocated... Then stop iterating and output the allocation rates of all open-ended choices; where, This represents the current number of Open Centers (OCs) in the data center. Let be the overall data center speed limit value for the current iteration step; Step b: Let ,Will The corresponding objective (OC) should be removed from the requirements list. To update the list of requirements, and at the same time To update the current demand rate for each OC; Step c: Determine if the requirement is met. or If the condition is met, the iteration stops. For each OC, the rates allocated in each iteration are summed to obtain the allocated rate output of the OC. If the condition is not met, the process returns to step a.

[0013] Optionally, it also includes: allocating the service rates of all services of the OC in order of service priority according to the allocation rate of the OC, until the allocation rate obtained by the OC is completely allocated.

[0014] Optionally, in step 3, the overall rate limit value is allocated to each OC according to the demand rate and configuration strategy of various services. Specifically, this includes: smoothing the demand rate of various services in the data center using the EWMA method to obtain the smoothed demand rate of each service; assigning priorities to various services, where the priority of unrestricted services is 0, and all other restricted services are assigned m-1 priorities, where m is the total number of priorities; and calculating the total demand of services with priority 0. : , For the smoothed demand rate, if Then the rate of services corresponding to all priorities is not limited. Then, the rate of services with priority 0 is not limited, and it is set to... Rate allocation for restricted services This sets the overall rate limit for the data center. When allocating rates for restricted services, rates are allocated to all levels of restricted services in order of service priority until the overall rate limit for the data center is fully allocated. For services with unallocated rates, the rate limit is set to the minimum guaranteed rate. For each type of restricted service, the average rate limit is obtained based on its allocated rate limit and the number of corresponding Open Controllers (OCs), thus outputting the allocated rate for each restricted service in each OC.

[0015] In a second aspect, a computer device is provided, including a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the steps of the feedback-based allocation flow control method for CDN data centers as described in any one of the first aspects.

[0016] Thirdly, a computer-readable storage medium is provided for storing a computer program; when executed by a processor, the computer program implements the steps of the feedback-based allocation flow control method for CDN data centers as described in any one of the first aspects.

[0017] Compared with existing technologies, the beneficial effects of this invention are as follows: The feedback control strategy of this invention improves the sensitivity and fine-grainedness of CDN data centers in controlling outgoing port bandwidth, avoiding the coarse-grainedness and high latency of current DNS-based traffic scheduling methods. It provides efficient monitoring capabilities for outgoing port overruns and enhances response capabilities to overruns. Furthermore, the bandwidth allocation strategy of this invention effectively identifies the actual needs within the CDN data center, providing stable and demand-appropriate rate limiting capabilities for each operational center (OC). Ultimately, this ensures that the overall external bandwidth of the CDN data center is stably limited within a specified limit, improving overall performance. Attached Figure Description

[0018] Figure 1 is a flowchart of the feedback-based traffic control method for CDN data centers according to the present invention; Figure 2 is a schematic diagram of the data preprocessing process of the present invention; Figure 3 is a schematic diagram of the data aggregation and statistics of the present invention; Figure 4 is a schematic diagram of the feedback control algorithm of the present invention; Figure 5 is a schematic diagram of traffic allocation of the present invention; Figure 6 is a schematic diagram of the rate limiting configuration distribution of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be used to limit the scope of protection of the present invention. It should be noted that the term "comprising" and any variations thereof in the specification, claims and the above-mentioned drawings of the present invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or devices.

[0020] Example 1, as shown in Figure 1, describes a feedback-based traffic control method applied to a CDN data center, comprising the following steps: Step 1: In the current control interval, acquire and preprocess the bandwidth data of each OC and outgoing port switch within the CDN data center. The bandwidth data includes the rate of the OC service and the rate of the outgoing port switch.

[0021] Step 2: Aggregate and statistically analyze the bandwidth data of the preprocessed OCs by IP or service to obtain the required rate of each OC or the required rate of each type of service under the current control interval.

[0022] Step 3: Determine whether the port switch rate exceeds the preset threshold. If it does, obtain the overall rate limit value of the computer room for the next control interval through the feedback control algorithm, and allocate the overall rate limit value to each OC / type of service in the computer room according to the required rate and configuration strategy of each OC / type of service in the computer room.

[0023] Step 4: Push the rate limiting configuration to each OC so that each OC can perform the rate limiting operation.

[0024] In this embodiment, the bandwidth data of the OC under the current control interval is bandwidth data in the dimensions of timestamp-IP-host-speed, where timestamp is the timestamp within the current control interval, IP is the OC identifier, host is the Uniform Resource Locator of the service served by the OC within the current control interval, and OC speed is the rate, representing the bandwidth value of the service served by the OC within the current control interval; the bandwidth data of the outgoing port switch under the current control interval is bandwidth data in the dimensions of timestamp-name-speed, where timestamp is the timestamp within the current control interval, name is the data center identifier, and outgoing port switch speed is the rate, representing the bandwidth value of the entire data center to the outside world within the current control interval.

[0025] In this embodiment, the OC connected to the central controller calculates the bandwidth (speed) of the service it provides at specified time intervals (intervals), using the OC's service as the granularity. The midpoint of this statistical time interval is used as the timestamp, the OC's IP address as the OC identifier, and the service's URL (Uniform Resource Locator) as the service identifier (host), ultimately forming bandwidth data in the dimensions of timestamp-IP-host-speed. For outgoing port switches, the entire data center's outgoing bandwidth (speed) is calculated at specified time intervals (intervals), and the midpoint of this statistical time interval is used as the timestamp, with the data center's name as the data center identifier, ultimately forming bandwidth data in the dimensions of timestamp-name-speed. When the interval ends, the OC and the switch report their respective bandwidth data for the current time interval to the central controller. The central controller executes steps 1-4.

[0026] As shown in Figure 2, the preprocessing includes filtering abnormal and expired data based on the validity of the timestamp and rate fields of the bandwidth data. Specifically, based on the bandwidth data of the received OC and outgoing port switch, if the time corresponding to the timestamp is not within the current control interval, the current bandwidth data is filtered; otherwise, the speed dimension data is checked. If the rate is less than or equal to 0, it is considered illegal data and is filtered; otherwise, it is considered normal data and is cached.

[0027] More specifically, a timer is maintained, running at intervals. For received bandwidth data from the OC and outgoing port switches, if the timestamp of the bandwidth data is not within the current time interval maintained by the central controller, the bandwidth data is filtered out; otherwise, if it is within the time interval, the speed field is checked. If the speed is less than or equal to 0, it is considered invalid data and is filtered out. The remaining data is considered normal data and stored in the data cache.

[0028] In this embodiment, as shown in Figure 3, step 2 specifically includes the following sub-steps: S21: Traverse the bandwidth data of each OC within the current time interval, locate the data center to which the IP of each OC belongs by searching the IP-data center name database, and classify the bandwidth data of all OCs according to the data center.

[0029] S22: For each data center, the rate data corresponding to the OCs with the same IP are summed to obtain the required rate of each OC under the current control interval, i.e., data center-IP-speed.

[0030] S23: For each data center, the rate data of the same service are accumulated to obtain the required rate of each type of service under the current control interval, i.e., data center-host-speed data.

[0031] In this embodiment, as shown in Figure 4, it is determined whether the rate of the outgoing port switch exceeds a preset threshold. If it does not exceed the preset threshold, no flow control is required; if it does, regulation is necessary. That is, to ensure the service quality of some operations, the rate of some open-source (OC) switches needs to be regulated to ensure that the overall outgoing port bandwidth of the data center does not exceed the preset limit. This invention employs a feedback control algorithm. Given that the data center is in a state below the limit for a long time, the main purpose of rate limiting is to effectively limit short-term over-limits. Therefore, the feedback control algorithm only uses differential and differential terms to avoid integral storms and feedback delay problems.

[0032] The overall speed limit value of the computer room for the next control interval is obtained through a feedback control algorithm, specifically: S31: The output of the feedback control algorithm in the current control interval is obtained according to the following formula. ;

[0033] in, For proportional gain, For differential gain, and In the current control interval and the previous control interval Within, the amount by which the outgoing port switch rate exceeds the preset threshold.

[0034] S32: The output of the feedback control algorithm in the current control interval is mapped to the overall room speed limit value in the next control interval using the following formula. ; ;in, This is the preset maximum speed limit offset. This is the overall speed limit value for the computer room in the previous control interval. This is the overall speed limit value for the current control interval of the computer room. In other words, to avoid speed fluctuations in the computer room caused by excessively large single adjustments, this invention configures a maximum speed limit offset. .

[0035] In this embodiment, step S33: The overall rate limit value is allocated to each OC according to the demand rate and configuration strategy of each OC, or the overall rate limit value is allocated to each OC according to the demand rate and configuration strategy of various services.

[0036] 1. Allocate the overall rate limit to each OC based on their required rate and configuration strategy.

[0037] In this embodiment, as shown in Figure 5, the overall rate limit value is allocated to each OC according to the demand rate and configuration strategy of each OC. Specifically, the demand rate of all OCs in the data center is smoothed by the EWMA method, and the smoothed demand rate of each OC and the overall rate limit value of the data center are used as initial values. The allocation rate of each OC is obtained by iterative loop.

[0038] Specifically, smoothing is used to address fluctuations in bandwidth statistics. The formula for smoothing is: ;in, This represents the rate of OC after smoothing in the previous control interval. This is the weighted average coefficient, typically taken as 0.5-0.7. The required rate under the current control interval, The required rate for OC after smoothing.

[0039] After smoothing, let the number of open-source computers in the computer room be... , to all OC Sort the requirements in ascending order, and then repeat the following steps in a loop: Step a: Select all current requirement rates of all requirements in the requirement list. The minimum value in is denoted as ,like Then assign to each OC The rate is determined, and step b is executed; otherwise, each OC is allocated... Then stop iterating and output the allocation rates of all open-ended choices; where, This represents the current number of Open Centers (OCs) in the data center. Let be the overall data center speed limit value for the current iteration step; Step b: Let ,Will The corresponding objective (OC) should be removed from the requirements list. To update the list of requirements, and at the same time To update the current demand rate for each OC; Step c: Determine if the requirement is met. or If the condition is met, the iteration stops. For each OC, the rates allocated in each iteration are summed to obtain the allocated rate output of the OC. If the condition is not met, the process returns to step a.

[0040] After the loop ends, the rate limiting configuration for each OC will be adjusted accordingly. Stored in the cache, waiting to be pushed to Objective-C upon expiration.

[0041] Furthermore, it also includes: allocating the service rates of all services in the OC according to the obtained allocation rate, in the order of service priority, until the allocated rate of the OC is completely allocated.

[0042] Specifically, high-priority strict guarantees can be provided for different services based on optional configurations. That is, services with high bandwidth requirements are given priority, while services with low customer perception have lower bandwidth guarantee priority. The specific process is as follows: 1. For all services under this IP, query the priority of the corresponding host (through the configuration center), and the configuration center returns the corresponding host priority list. 2. Sort in reverse order of priority, and record the number of hosts after sorting as the number of hosts for this IP. 3. Execute the following steps in a loop: (1) Take out the sorted The first record in is denoted as ,if Then allocate the full bandwidth requirement to that host. (2) Let And remove the host from the demand list, meaning that the demand for that host has been met, and update each value in the demand list with the remaining demand according to the allocated amount. (3) For step 1 In this situation, The bandwidth is allocated to that host, then the loop ends and lower-priority hosts no longer receive bandwidth. After the loop ends, the rate-limiting configuration is stored in the cache to await the completion of all rate-limiting configuration calculations.

[0043] 2. Allocate the overall rate limit to each OC according to the demand rate and configuration strategy of various services.

[0044] In this embodiment, the specific steps include: Step I: Smoothing the demand rates of various services in the data center using the EWMA method to obtain the smoothed demand rates of various services; the smoothing method is described above.

[0045] Step II: Assign priorities to various types of services. Unrestricted services have a priority of 0, and all other restricted services are assigned m-1 priorities, where m is the total number of priorities.

[0046] Step III: Calculate the total demand for services with priority 0. : , For the smoothed demand rate, if Then the rate of services corresponding to all priorities is not limited. Then, the rate of services with priority 0 is not limited, and it is set to... Rate allocation for restricted services This is the overall speed limit for the computer room.

[0047] Step IV: When allocating rates for restricted services, rate allocation is performed on all levels of restricted services in order of service priority until the overall rate limit value of the data center is allocated. For services with unallocated rates, the rate limit value is set to the minimum guaranteed value.

[0048] Step V: For each type of restricted service, based on its assigned rate limit value and the number of corresponding Open Rates (OCs), obtain the average rate limit for each type of restricted service, and then output the assigned rate for each restricted service for each OC.

[0049] Specifically, step IV is: Let , record the current The host requirements are: .like Then allocate to this business. The speed limit value makes Repeat the current steps until the overall speed limit value for the computer room has been allocated.

[0050] In this embodiment, as shown in Figure 6, the central controller maintains a speed limit configuration list based on OC (Operational Center) and speed limit value. Once the speed limit configuration for the current period is calculated, the central controller cyclically pushes the corresponding speed limit configuration to all OC nodes. The OC nodes receive... As the overall machine speed limit, the receiver This serves as the rate limit for different services within the machine.

[0051] Example 2: This invention provides a computer device, including a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the steps of the above-described flow control method based on feedback allocation applied to CDN data centers.

[0052] For more detailed information on the above methods, please refer to the relevant content disclosed in the foregoing embodiments, which will not be repeated here.

[0053] Example 3: This invention provides a computer-readable storage medium for storing a computer program; when the computer program is executed by a processor, it implements the steps of the above-described feedback-based flow control method applied to CDN data centers.

[0054] For more detailed information on the above methods, please refer to the relevant content disclosed in the foregoing embodiments, which will not be repeated here.

[0055] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The devices and storage media disclosed in the embodiments are described simply because they correspond to the methods disclosed in the embodiments; relevant parts can be referred to the method section.

[0056] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.

[0057] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A feedback-based traffic control method for CDN data centers, characterized in that, The process includes the following steps: Step 1: In the current control interval, acquire and preprocess the bandwidth data of each OC and outgoing port switch in the CDN data center. The bandwidth data includes the rate of the OC service and the rate of the outgoing port switch. Step 2: Aggregate and statistically analyze the preprocessed OC bandwidth data by IP or service to obtain the required rate of each OC or the required rate of each type of service under the current control interval. Step 3: Determine whether the rate of the outgoing port switch exceeds a preset threshold. If it does, obtain the overall rate limit value of the data center for the next control interval through a feedback control algorithm, and allocate the overall rate limit value to each OC according to the required rate and configuration strategy of each OC / type of service in the data center. Step 4: Push the rate limit configuration to each OC so that each OC can perform the rate limit operation.

2. The feedback-based traffic control method for CDN data centers according to claim 1, characterized in that, In step 1, the bandwidth data of the OC under the current control interval is bandwidth data in the dimensions of timestamp-IP-host-speed, where timestamp is the timestamp within the current control interval, IP is the OC identifier, host is the Uniform Resource Locator of the service served by the OC within the current control interval, and OC speed is the rate, representing the bandwidth value of the service served by the OC within the current control interval; the bandwidth data of the outgoing port switch under the current control interval is bandwidth data in the dimensions of timestamp-name-speed, where timestamp is the timestamp within the current control interval, name is the data center identifier, and outgoing port switch speed is the rate, representing the bandwidth value of the entire data center to the outside world within the current control interval.

3. The feedback-based traffic control method for CDN data centers according to claim 2, characterized in that, In step 1, the preprocessing includes filtering abnormal and expired data based on the validity of the timestamp and rate fields of the bandwidth data. Specifically, based on the bandwidth data of the received OC and outgoing port switch, if the time corresponding to the timestamp is not within the current control interval, the current bandwidth data is filtered; otherwise, the speed dimension data is checked. If the rate is less than or equal to 0, it is illegal data and is filtered; otherwise, it is normal data and is cached.

4. The feedback-based traffic control method for CDN data centers according to claim 1, characterized in that, Step 2 specifically involves: traversing the bandwidth data of each OC within the current time interval, locating the data center to which the IP of each OC belongs by searching the IP-data center name database, and classifying the bandwidth data of all OCs according to the data center; for each data center, accumulating the rate data corresponding to OCs with the same IP to obtain the required rate of each OC under the current control interval; and for each data center, accumulating the rate data of services with the same service to obtain the required rate of each type of service under the current control interval.

5. The feedback-based traffic control method for CDN data centers according to claim 1, characterized in that, In step 3, the overall speed limit value of the computer room for the next control interval is obtained through a feedback control algorithm. Specifically, the output of the feedback control algorithm in the current control interval is obtained according to the following formula. ； in, For proportional gain, For differential gain, and In the current control interval and the previous control interval Within, the amount by which the outgoing port switch rate exceeds a preset threshold; the following formula maps the output of the feedback control algorithm in the current control interval to the overall room rate limit value for the next control interval. ； ;in, This is the preset maximum speed limit offset. This is the overall speed limit value for the computer room in the previous control interval. This is the overall speed limit value for the current control interval of the computer room.

6. The feedback-based traffic control method for CDN data centers according to claim 1, characterized in that, In step 3, the overall rate limit is allocated to each OC based on their demand rate and configuration strategy. Specifically, the demand rates of all OCs in the data center are smoothed using the EWMA method, and the smoothed demand rates of each OC and the overall rate limit of the data center are used as initial values. The allocation rate of each OC is obtained through iterative looping. The allocation rate of each OC is obtained through iterative looping, which includes the following steps: Step a: Select the current demand rate of all OCs in the demand list. The minimum value in is denoted as ,like Then assign to each OC The rate is determined, and step b is executed; otherwise, each OC is allocated... Then stop iterating and output the allocation rates of all open-ended choices; where, This represents the current number of Open Centers (OCs) in the data center. Let be the overall data center speed limit value for the current iteration step; Step b: Let ,Will The corresponding objective (OC) should be removed from the requirements list. To update the list of requirements, and at the same time To update the current demand rate for each OC; Step c: Determine if the requirement is met. or If the condition is met, the iteration stops. For each OC, the rates allocated in each iteration are summed to obtain the allocated rate output of the OC. If the condition is not met, the process returns to step a.

7. The traffic control method based on feedback allocation applied to CDN data centers according to claim 6, characterized in that, Also includes: Based on the allocated rate of the OC, the rate of all services of the OC is allocated in order of service priority until the allocated rate of the OC is completely allocated.

8. The traffic control method based on feedback allocation applied to CDN data centers according to claim 4, characterized in that, In step 3, the overall rate limit value is allocated to each OC according to the demand rate and configuration strategy of various services. Specifically, this includes: smoothing the demand rate of various services in the data center using the EWMA method to obtain the smoothed demand rate of each service; assigning priorities to various services, where the priority of unrestricted services is 0, and all other restricted services are assigned m-1 priorities, where m is the total number of priorities; and calculating the total demand of services with priority 0. ： ， For the smoothed demand rate, if Then the rate of services corresponding to all priorities is not limited. Then, the rate of services with priority 0 is not limited, and it is set to... Rate allocation for restricted services This sets the overall rate limit for the data center. When allocating rates for restricted services, rates are allocated to all levels of restricted services in order of service priority until the overall rate limit for the data center is fully allocated. For services with unallocated rates, the rate limit is set to the minimum guaranteed rate. For each type of restricted service, the average rate limit is obtained based on its allocated rate limit and the number of corresponding Open Controllers (OCs), thus outputting the allocated rate for each restricted service in each OC.

9. A computer device, characterized in that, It includes a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the steps of the feedback-based flow control method for CDN data centers as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, Used to store computer programs; when executed by a processor, the computer programs implement the steps of the feedback-based allocation flow control method for CDN data centers as described in any one of claims 1-8.

Images