CDN bandwidth prediction method, device, electronic device, medium and product

Abstract

The present application provides a CDN bandwidth prediction method, device, electronic device, medium and product. According to an example of the present application, the method may include: based on the traffic data of each customer domain name corresponding to the CDN node, each customer domain name is divided into a core domain name and a long-tail domain name, and the bandwidth demand of the core domain name is higher than the bandwidth demand of the long-tail domain name; using a time series prediction model, according to the historical bandwidth corresponding to the core domain name, a first predicted bandwidth is generated; using a time series smoothing method, according to the historical bandwidth corresponding to the long-tail domain name, a second predicted bandwidth is generated, and the first predicted bandwidth and the second predicted bandwidth are used to reflect the predicted average bandwidth of each customer domain name in the next day; according to the predicted average bandwidth in the next day, the overall predicted bandwidth in the prediction time period is determined.

Description

Technical Field

[0001] The present application relates to the field of computer network technology, and in particular to a CDN bandwidth prediction method, device, electronic device, medium and product. Background Art

[0002] With the rapid development of Internet technology, Content Delivery Network (CDN) has become an indispensable part of modern Internet services. It effectively accelerates content transmission and improves user experience by deploying server networks around the world. In the process of operation, CDN service providers face a dynamically changing and highly competitive market environment, in which the reasonable planning and cost control of bandwidth resources are crucial to maintaining business competitiveness and profitability.

[0003] Current bandwidth prediction methods usually only use simple mathematical models to predict bandwidth values, resulting in low bandwidth prediction accuracy. This will not only lead to misjudgment of resource allocation by CDN service providers, but may also affect user experience during peak hours due to insufficient resources, and even cause service interruptions. Summary of the invention

[0004] In order to overcome the problems existing in the related art, the present application provides a CDN bandwidth prediction method, device, electronic device, medium and product.

[0005] According to a first aspect of any embodiment of the present application, a CDN bandwidth prediction method is provided, the method comprising:

[0006] Based on the traffic data of each customer domain name corresponding to the CDN node, the customer domain names are divided into core domain names and long-tail domain names, and the bandwidth requirement of the core domain name is higher than the bandwidth requirement of the long-tail domain name;

[0007] Generate a first predicted bandwidth according to the historical bandwidth corresponding to the core domain name by using a time series prediction model;

[0008] Using a time series smoothing method, generating a second predicted bandwidth according to the historical bandwidth corresponding to the long-tail domain name, wherein the first predicted bandwidth and the second predicted bandwidth are used to reflect the predicted average bandwidth of each customer domain name in the next day;

[0009] The overall predicted bandwidth within the predicted time period is determined according to the predicted average bandwidth within the future day.

[0010] According to a second aspect of any one of the embodiments of the present application, a CDN bandwidth prediction device is provided, the device comprising:

[0011] A domain name division module, for dividing each customer domain name into a core domain name and a long-tail domain name based on traffic data of each customer domain name corresponding to the CDN node, wherein the bandwidth requirement of the core domain name is higher than the bandwidth requirement of the long-tail domain name;

[0012] A first prediction module, configured to generate a first predicted bandwidth according to a historical bandwidth corresponding to the core domain name by using a time series prediction model;

[0013] A second prediction module is used to generate a second predicted bandwidth according to the historical bandwidth corresponding to the long-tail domain name by using a time series smoothing method, wherein the first predicted bandwidth and the second predicted bandwidth are used to reflect the predicted average bandwidth of each customer domain name in the next day;

[0014] The overall prediction module is used to determine the overall predicted bandwidth within the prediction time period according to the predicted average bandwidth within the future day.

[0015] According to a third aspect of any embodiment of the present application, there is provided an electronic device, including:

[0016] processor;

[0017] a memory for storing processor-executable instructions;

[0018] The processor implements the method described in any embodiment of the present application by running the executable instructions.

[0019] According to a fourth aspect of any embodiment of the present application, a computer-readable storage medium is provided, on which computer instructions are stored. When the instructions are executed by a processor, the method described in any embodiment of the present application is implemented.

[0020] According to a fifth aspect of any embodiment of the present application, a computer program product is provided, on which a computer program / instruction is stored, and when the computer program / instruction is executed by a processor, the method described in any embodiment of the present application is implemented.

[0021] The technical solution provided by this application may have the following beneficial effects:

[0022] According to the above embodiments, each customer domain name is divided into a core domain name and a long-tail domain name based on the traffic data of each customer domain name corresponding to the CDN node. A time series prediction model is used to generate a first predicted bandwidth according to the historical bandwidth corresponding to the core domain name. A time series smoothing method is used to generate a second predicted bandwidth according to the historical bandwidth corresponding to the long-tail domain name. The overall predicted bandwidth within the predicted time period is determined based on the predicted average bandwidth in the next day. Since the traffic of the core domain name is larger and more stable, a more accurate time series prediction model can be used to capture its traffic pattern and trend. For long-tail domain names with small and scattered traffic, the time series smoothing method can be used to quickly predict bandwidth data and reduce calculation time. Different prediction methods are used for domain names with different characteristics. While improving prediction accuracy, the calculation complexity is reduced, which helps CDN operators to better control costs and manage budgets.

[0023] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings, which are incorporated in the specification and constitute a part of this application, illustrate embodiments consistent with the application and, together with the description, serve to explain the principles of the application.

[0025] Figure 1 is a flow chart of a CDN bandwidth prediction method according to an exemplary embodiment of the present application;

[0026] Figure 2 This is a flow chart of a CDN bandwidth prediction method combining weekly seasonal characteristics according to an exemplary embodiment of the present application;

[0027] Figure 3 is a flow chart of a CDN bandwidth prediction method combined with holiday characteristics according to an exemplary embodiment of the present application;

[0028] Figure 4 is a flowchart of another CDN bandwidth prediction method according to an exemplary embodiment of the present application;

[0029] Figure 5 is a structural schematic diagram of an electronic device according to an exemplary embodiment of the present application;

[0030] Figure 6 It is a block diagram of a CDN bandwidth prediction device according to an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0031] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0032] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0033] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0034] Current bandwidth prediction methods usually only use simple mathematical models to predict bandwidth values, resulting in low accuracy of bandwidth prediction, which will affect the reasonable allocation of CDN resources and further affect the operating costs of CDN.

[0035] In order to solve the above problems, the present application proposes a CDN bandwidth prediction method. To further illustrate the present application, the following embodiments are provided:

[0036] See also Figure 1 , Figure 1 This is a flow chart of a CDN bandwidth prediction method according to an exemplary embodiment of the present application. The CDN bandwidth prediction method can be applied to a CDN management system, and the CDN management system can be a central big data cluster in the CDN. The method can include the following steps:

[0037] Step 102: Based on the traffic data of each customer domain name corresponding to the CDN node, each customer domain name is divided into a core domain name and a long-tail domain name, and the bandwidth demand of the core domain name is higher than the bandwidth demand of the long-tail domain name.

[0038] In this step, on each CDN node, the traffic data of each customer domain name corresponding to each CDN node is collected, and the traffic data is aggregated to the CDN management system for aggregation and processing. The CDN management system can perform preliminary aggregation and processing on the aggregated traffic data of each customer domain name, merge the traffic data of the same customer domain name and the same time period, and calculate the total traffic, etc.

[0039] The CDN management system sets a preset period (such as five minutes) as the granularity for data aggregation, converts traffic data into bandwidth data, and generates summary data corresponding to the traffic data for each CDN node.

[0040] The CDN node may be an edge host. The summary data is data summarized at a preset period granularity, and may include customer domain names, business areas, and bandwidth data of traffic data. The customer domain name is used to identify the source of traffic data. The business area is information such as geographic areas and operator attributes derived from analysis of the IP address of the traffic. The bandwidth data is the average bandwidth within a preset period, which is used to reflect the rate and efficiency of network transmission.

[0041] The CDN management system can pre-process bandwidth data, identify and process abnormal values ​​and missing values ​​of bandwidth data, and identify and confirm abnormal fluctuations such as bandwidth bursts and bandwidth drops. After pre-processing, based on the bandwidth data corresponding to the traffic data, each customer domain name is divided into core domain names with higher bandwidth requirements and long-tail domain names with lower bandwidth requirements.

[0042] Among them, core domain names are domain names with large traffic, frequent and stable visits, and have higher demands for bandwidth resources. Long-tail domain names are domain names with smaller visits, more dispersed traffic and larger numbers, and have relatively lower demands for bandwidth resources. The bandwidth demand of core domain names is higher than that of long-tail domain names.

[0043] Step 104: Generate a first predicted bandwidth according to the historical bandwidth corresponding to the core domain name by using the time series prediction model.

[0044] In this step, the CDN management system can obtain the historical bandwidth of the core domain name in the past period of time, and use a time series prediction model such as a seasonal autoregressive integrated moving average with eXogenous regressors (SARIMAX) to generate a first predicted bandwidth according to the historical bandwidth corresponding to the core domain name.

[0045] The CDN management system uses historical data to train the SARIMAX model and evaluates the model's prediction performance through methods such as cross-validation. The historical bandwidth corresponding to the core domain name is used as the input data of the trained SARIMAX model to obtain the first predicted bandwidth for the next day generated by the SARIMAX model. The SARIMAX model can take into account the seasonality, autoregression, and moving average of the time series, and can better handle data with periodic changes.

[0046] The historical bandwidth is the average bandwidth in the past period of time with a preset period as the granularity, for example, the average bandwidth of the core domain name in the past seven days with a granularity of five minutes. The first predicted bandwidth is the average bandwidth of the core domain name in the next day with a preset period as the granularity, for example, the average bandwidth in the next day with a granularity of five minutes.

[0047] Step 106: Using a time series smoothing method, a second predicted bandwidth is generated according to the historical bandwidth corresponding to the long-tail domain name. The first predicted bandwidth and the second predicted bandwidth are used to reflect the predicted average bandwidth of each customer domain name in the next day.

[0048] In this step, for the sake of system performance, the CDN management system can obtain the historical bandwidth of the long-tail domain name in the past period of time, and use time-series smoothing methods such as weighted moving average method and exponential smoothing that consume less computing resources to generate a second predicted bandwidth based on the historical bandwidth corresponding to the long-tail domain name.

[0049] Taking the weighted moving average method as an example, the CDN management system can determine the weighted moving average period based on the volatility of historical bandwidth and predicted demand, and assign different weights to historical bandwidth at different time points based on the timeliness of historical bandwidth. Use the weighted moving average formula to calculate the weighted moving average for the next day based on historical bandwidth and assigned weights.

[0050] The second predicted bandwidth is the average bandwidth of the long-tail domain name in the next day with a preset period as the granularity, for example, the average bandwidth of the long-tail domain name in the next day with a granularity of five minutes. The first predicted bandwidth and the second predicted bandwidth are used to reflect the predicted average bandwidth of each customer domain name in the next day.

[0051] The first predicted bandwidth can reflect the predicted average bandwidth of the core domain name in the next day, and the second predicted bandwidth can reflect the predicted average bandwidth of the long-tail domain name in the next day. The predicted average bandwidth is the average bandwidth of the core domain name and the long-tail domain name in the next day, which is aggregated at a preset period.

[0052] Step 108: Determine the overall predicted bandwidth within the predicted time period according to the predicted average bandwidth within the next day.

[0053] In this step, the CDN management system determines the overall predicted bandwidth within the prediction period based on the predicted average bandwidth within the next day. For example, the predicted average bandwidth within the next day is copied according to the number of days in the prediction period to obtain the overall predicted bandwidth within the prediction period; for another example, the overall predicted bandwidth within the prediction period is determined based on the change trend of historical bandwidth, weekly seasonality within the prediction period, and holiday characteristics.

[0054] The forecast time period is a future time period, which can be the next billing cycle or a time period set according to forecast demand. The forecast time period includes at least one day in the future. The overall forecast bandwidth is the average bandwidth aggregated with the preset cycle as the granularity within the forecast time period.

[0055] In one embodiment, the CDN management system can summarize the traffic data of each customer domain name based on the summary period, and obtain the regional share data of each summary period in the historical time period. According to the regional share data of each summary period in the historical time period, the average bandwidth of each preset period in the overall predicted bandwidth is allocated to the service area.

[0056] Specifically, the average bandwidth of each preset period is allocated according to the regional share of the corresponding summary period in the historical time period. For example, if the regional share of a customer domain name in region A is 60% and the regional share in region B is 40% in a certain summary period in the historical time period, then of the average bandwidth of the corresponding preset period in the forecast time period, 60% of the average bandwidth is allocated to region A and 40% of the average bandwidth is allocated to region B.

[0057] The historical time period is a time range in the past, which is used to collect and analyze the traffic data of the customer domain name in the time period. It can be a past billing cycle, or a specified past two months, three months, etc. The length of the historical time period can be the same as the length of the forecast time period.

[0058] The regional share data is used to reflect the distribution of traffic data of each customer domain name in different business areas. The summary period is the time unit for summarizing the business area of ​​traffic data, and the preset period is the time unit for analyzing the bandwidth data of traffic data. The duration of the summary period is longer than the duration of the preset period.

[0059] As mentioned above, by obtaining the regional share data of each summary period within the historical time period, we can understand the distribution of traffic data of each customer domain name in different business areas. According to the regional share data of each summary period within the historical time period, the average bandwidth of each preset period in the overall predicted bandwidth is allocated to the business area, which can achieve more refined CDN resource management and allocation.

[0060] The CDN bandwidth prediction method of this embodiment divides each customer domain name into a core domain name and a long-tail domain name based on the traffic data of each customer domain name corresponding to the CDN node, uses a time series prediction model to generate a first predicted bandwidth according to the historical bandwidth corresponding to the core domain name, uses a time series smoothing method to generate a second predicted bandwidth according to the historical bandwidth corresponding to the long-tail domain name, and determines the overall predicted bandwidth within the prediction time period based on the predicted average bandwidth in the next day. Since the traffic of the core domain name is larger and more stable, a more accurate time series prediction model can be used to capture its traffic pattern and trend. For long-tail domain names with small and scattered traffic, the use of a time series smoothing method can quickly predict bandwidth data and reduce calculation time. Different prediction methods are used for domain names with different characteristics, which improves prediction accuracy while reducing calculation complexity, which helps CDN operators to better control costs and manage budgets.

[0061] In the above-mentioned embodiment, it is introduced that each customer domain name is divided into a core domain name with a higher bandwidth demand and a long-tail domain name with a lower bandwidth demand, and different prediction methods are used to predict the predicted average bandwidth of the core domain name and the long-tail domain name in the next day, and determine the overall predicted bandwidth in the prediction time period. In the following embodiment, the determination process of the overall predicted bandwidth will be described in more detail, and can be applied to any of the above embodiments.

[0062] In one embodiment, see Figure 2 , Figure 2 This is a flow chart of a CDN bandwidth prediction method combined with weekly seasonal characteristics according to an exemplary embodiment of the present application. The method may include the following steps:

[0063] Step 202: Obtain the daily bandwidth peak value of each customer domain name in a historical week.

[0064] In this step, the CDN management system can pre-add a weekly feature identifier from Monday to Sunday to the bandwidth data. The CDN management system obtains the daily bandwidth peak value of each customer domain name in the historical week based on the bandwidth data and weekly feature identifiers summarized in the historical week.

[0065] The daily bandwidth peak is the highest bandwidth peak value from Monday to Sunday in the historical week, which is used to analyze the periodic changes in bandwidth demand.

[0066] Step 204: The average value of the daily bandwidth peak values ​​of multiple days in the historical week is taken as the weekly base peak value.

[0067] In this step, the CDN management system takes the average of the daily bandwidth peaks of multiple days in the historical week as the weekly benchmark peak, adds the daily bandwidth peaks from Monday to Sunday in the historical week, and then divides it by seven to get the weekly benchmark peak. The weekly benchmark peak is used to reflect the average bandwidth peak level of the customer domain name in a week.

[0068] Step 206: Calculate the ratio of each daily bandwidth peak value to the weekly base peak value to obtain the daily characteristic ratio within the historical week.

[0069] In this step, the CDN management system calculates the ratio of each daily bandwidth peak to the weekly benchmark peak for each day in the historical week, and obtains the daily characteristic ratio of the seven days in the historical week. The daily characteristic ratio is the ratio of the daily bandwidth peak to the weekly benchmark peak on a certain day in the historical week, which is used to reflect the intensity of the bandwidth demand on that day relative to the average level.

[0070] Step 208: averaging the daily feature ratios in multiple historical weeks to obtain an averaged daily feature ratio.

[0071] In this step, the CDN management system averages the daily feature ratios in multiple historical weeks to obtain the averaged daily feature ratio. For example, the daily feature ratios of Mondays in multiple historical weeks are obtained, and the daily feature ratios of Mondays in multiple historical weeks are averaged to obtain the averaged daily feature ratio of Mondays.

[0072] Step 210: Determine the overall predicted bandwidth by combining the predicted average bandwidth and the averaged daily feature ratio.

[0073] In this step, the CDN management system combines the predicted average bandwidth for the next day and the average daily feature ratio, multiplies the predicted average bandwidth for the next day by the average daily feature ratio, and combines the average bandwidth of each day in the prediction time period to form an overall predicted bandwidth.

[0074] As described above, by obtaining the daily bandwidth peak value of each customer domain name in the historical week, taking the average of the daily bandwidth peak values ​​of multiple days in the historical week as the weekly base peak value, respectively calculating the ratio of each daily bandwidth peak value to the weekly base peak value, and obtaining the daily feature ratio in the historical week, the traffic characteristics of the customer domain name from Monday to Sunday can be captured. The daily feature ratios in multiple historical weeks are respectively averaged to obtain the averaged daily feature ratio, which can reduce the impact of accidental factors on the prediction results. The overall predicted bandwidth is determined by combining the predicted average bandwidth and the averaged daily feature ratio. There is no need to use complex models for training and prediction, and the weekly seasonal bandwidth demand of the customer domain name can be simply, quickly and comprehensively reflected, thereby improving the accuracy of bandwidth prediction.

[0075] In one embodiment, see Figure 3 , Figure 3 This is a flowchart of a CDN bandwidth prediction method combined with holiday characteristics according to an exemplary embodiment of the present application. The method may include the following steps:

[0076] Step 302: Obtain the bandwidth peak values ​​of each customer domain name over multiple days during historical holidays.

[0077] In this step, the CDN management system can pre-add holiday feature identifiers such as the Mid-Autumn Festival and National Day to the bandwidth data. The CDN management system obtains the multi-day bandwidth peak value of each customer domain name during the historical holidays based on the bandwidth data and holiday feature identifiers summarized in the historical week. The multi-day bandwidth peak value during the historical holidays can include the bandwidth peak value on the first day of the holiday, the bandwidth peak value in the middle period, and the bandwidth peak value on the last day of the holiday.

[0078] Step 304: Based on the holiday benchmark peak value, obtain the first day characteristic ratio, the middle day characteristic ratio and the last day characteristic ratio of the historical holiday.

[0079] In this step, the CDN management system uses the ratio of the bandwidth peak value on the first day of the holiday to the holiday benchmark peak value as the first-day characteristic ratio of the historical holiday, and uses the ratio of the bandwidth peak value on the last day of the holiday to the holiday benchmark peak value as the last-day characteristic ratio of the historical holiday.

[0080] If the number of historical holidays is greater than three days, calculate the ratio of the bandwidth peak value of each day in the middle period to the holiday benchmark peak value, and take the average of the ratios of the bandwidth peak value of each day in the middle period to the holiday benchmark peak value as the middle characteristic ratio of the historical holiday.

[0081] If the number of days of the historical holiday is equal to three days, the ratio of the bandwidth peak in the middle period to the holiday benchmark peak is used as the middle-segment characteristic ratio in the historical holiday.

[0082] The holiday base peak is the bandwidth peak on the last day before the historical holiday. The first-day characteristic ratio is the ratio of the bandwidth peak on the first day of the holiday to the holiday base peak, which is used to reflect the relative strength of bandwidth demand at the beginning of the holiday.

[0083] The mid-period characteristic ratio is the average ratio of the bandwidth peak in the middle of the holiday to the holiday base peak, which is used to reflect the relative strength of bandwidth demand in the middle of the holiday. The last-day characteristic ratio is the ratio of the bandwidth peak on the last day of the holiday to the holiday base peak, which is used to reflect the relative strength of bandwidth demand at the end of the holiday.

[0084] Step 306: Average the first-day characteristic ratio, the middle-day characteristic ratio, and the last-day characteristic ratio of multiple historical holidays respectively.

[0085] In this step, the CDN management system averages the first-day characteristic ratio, middle-day characteristic ratio and last-day characteristic ratio of multiple historical holidays such as Mid-Autumn Festival and National Day to obtain the averaged first-day characteristic ratio, middle-day characteristic ratio and last-day characteristic ratio.

[0086] For example, the bandwidth peak values ​​on the first day of the Mid-Autumn Festival and the National Day holidays are obtained, and the bandwidth peak values ​​on the first day of the Mid-Autumn Festival and the National Day holidays are averaged to obtain the averaged bandwidth peak value on the first day of the holiday.

[0087] Step 308: Determine the overall predicted bandwidth by combining the predicted average bandwidth with the averaged first-day feature ratio, middle-day feature ratio, and last-day feature ratio.

[0088] In this step, if the forecast time period includes holidays, the CDN management system combines the forecast average bandwidth in the future day with the average first-day feature ratio, middle-day feature ratio, and last-day feature ratio, multiplies the forecast average bandwidth in the future day by the average first-day feature ratio, middle-day feature ratio, and last-day feature ratio, and combines the average bandwidth of each day in the forecast time period to form an overall forecast bandwidth.

[0089] As described above, by dividing the holidays into different time periods and calculating the characteristic proportions of different time periods respectively, the traffic characteristics during the holidays can be captured more focusedly, reducing the impact of data fluctuations on the prediction results, and helping to more accurately identify traffic changes in different time periods. Averaging the characteristic proportions of the first day, the middle day and the last day of multiple historical holidays can reduce the impact of accidental factors on the prediction results. The overall predicted bandwidth can be determined by combining the predicted average bandwidth and the averaged characteristic proportions of the first day, the middle day and the last day. There is no need to use complex models for training and prediction, and it can simply, quickly and comprehensively reflect the bandwidth demand of the holiday characteristics of the customer domain name, thereby improving the accuracy of bandwidth prediction.

[0090] In the above-mentioned embodiment, it is introduced that the overall predicted bandwidth within the complete prediction time period is generated by combining the weekly seasonal characteristics and holiday characteristics of the prediction time period based on the predicted average bandwidth within the next day. In the following embodiment, the processing process of future events within the prediction time period will be described in detail and can be applied to any of the above embodiments.

[0091] In one embodiment, the future event may include a burst expectation requirement. The customer may set a burst expectation requirement for a specific customer domain name according to their business needs. The CDN management system obtains the burst expectation requirement corresponding to each customer domain name, and based on the bandwidth growth multiple indicated by the burst expectation requirement, amplifies the overall predicted bandwidth within the specified time period to obtain the overall predicted bandwidth that meets the burst expectation requirement.

[0092] The burst expectation requirement is used to indicate the bandwidth growth multiple within a specified time period so that the CDN management system can prepare in advance to ensure the stability and availability of the service. The specified time period is the time period in which the bandwidth growth may occur, and the specified time period can be within the forecast time period.

[0093] Taking the prediction period as one month in the future and the specified period as the fifth to tenth days in the future as an example, if the customer sets the bandwidth growth multiple of the burst expectation requirement as 2 times, the CDN management system will multiply the average bandwidth from the fifth to the tenth day in the future by 2, and apply the amplified average bandwidth to the overall predicted bandwidth in the specified time period to obtain the adjusted overall predicted bandwidth.

[0094] As described above, by obtaining the expected burst requirements corresponding to each customer domain name, the average bandwidth in a specified time period is amplified based on the bandwidth growth multiple to obtain the overall predicted bandwidth that meets the expected burst requirements. This allows for personalized bandwidth prediction for each customer, thereby improving the accuracy of the prediction.

[0095] In one embodiment, the future event may include a checkout requirement. The customer may set a checkout requirement for a specific customer domain name according to their business needs. The CDN management system obtains the checkout requirement corresponding to each customer domain name, divides the specified 95% peak value indicated by the checkout requirement by the predicted 95% peak value of the overall predicted bandwidth, and calculates the scaling factor.

[0096] The calculated scaling factor is used to scale the overall predicted bandwidth in proportion. The average bandwidth of each preset period in the overall predicted bandwidth is multiplied by the scaling factor to obtain the overall predicted bandwidth that meets the checkout requirements. The predicted 95 peak value of the overall predicted bandwidth that meets the checkout requirements is the same as the specified 95 peak value.

[0097] The billing requirement is used to indicate the expected specified 95 peak value. The specified 95 peak value is the expected bandwidth peak value set by the customer for a specific customer domain name, which is used to represent the bandwidth usage that the customer expects not to exceed 95% of the time. The predicted 95 peak value is the maximum value of the remaining 95% average bandwidth after removing the first 5% average bandwidth from the overall predicted bandwidth.

[0098] When future events include both burst expectation requirements and billing requirements, the CDN management system can prioritize burst expectation requirements and locally weight the overall predicted bandwidth based on the specified time period and bandwidth growth multiple indicated by the bandwidth growth multiple. Then process the billing requirements and scale the overall predicted bandwidth in proportion based on the specified 95 peak value indicated by the billing requirements to obtain an overall predicted bandwidth that meets both burst expectation requirements and billing requirements.

[0099] As described above, by obtaining the checkout requirements corresponding to each customer domain name, the overall predicted bandwidth is scaled proportionally to obtain the overall predicted bandwidth that meets the checkout requirements, ensuring that the predicted 95% peak value of the scaled overall predicted bandwidth is the same as the specified 95% peak value, so that the overall predicted bandwidth accurately matches the customer's expectations, thereby avoiding additional costs or insufficient resources due to inaccurate bandwidth prediction.

[0100] In the above-mentioned embodiment, it is introduced that the overall predicted bandwidth that meets the customer's expectations is generated by locally weighted amplification and proportional scaling of the average bandwidth in the predicted time period. In the following embodiment, the prediction process of the predicted average bandwidth in the next day will be described in more detail, and can be applied to any of the above embodiments.

[0101] In one embodiment, the CDN management system can obtain bandwidth data corresponding to the traffic data based on the traffic data of each customer domain name, ignoring the bandwidth data of historical days with abnormal fluctuations such as sudden increase in customer volume and sudden decrease in customer volume in the historical time period. For each customer domain name, the bandwidth peak value of each day in the historical time period is determined based on the bandwidth data in the historical time period. Based on the bandwidth peak value determined for each day in the historical time period, the average bandwidth peak value of each customer domain name in the historical time period is determined.

[0102] Based on the bandwidth peak average, sort each customer domain name to obtain the bandwidth peak average from high to low or from low to high sorting results. According to the sorting results, the customer domain names that meet the preset sorting conditions are regarded as core domain names, and the customer domain names that do not meet the preset sorting conditions are regarded as long-tail domain names.

[0103] The bandwidth peak average is the average value of the bandwidth peak of a customer domain name every day in the historical time period, which is used to reflect the average maximum bandwidth demand of the customer domain name in the historical time period. The preset sorting condition is a set of rules or standards used to sort customer domain names.

[0104] Exemplarily, the preset sorting condition may be that the ranking exceeds a preset percentage, and the customer domain names whose ranking exceeds the preset percentage are taken as core domain names, and the customer domain names whose ranking does not exceed the preset percentage are taken as long-tail domain names; the preset sorting condition may also be that the bandwidth peak average value exceeds a preset threshold, and the customer domain names whose bandwidth peak average value exceeds the preset threshold are taken as core domain names, and the customer domain names whose bandwidth peak average value does not exceed the preset threshold are taken as long-tail domain names. The embodiments of the present application do not impose any restrictions on this.

[0105] As described above, by determining the average bandwidth peak value of each customer domain name in the historical time period based on the traffic data of each customer domain name, and sorting each customer domain name based on the bandwidth peak value, according to the sorting result, the customer domain names that meet the preset sorting conditions are taken as core domain names, and the customer domain names that do not meet the preset sorting conditions are taken as long-tail domain names. In this way, the core domain names with high bandwidth requirements and the long-tail domain names with low bandwidth requirements in the historical time period can be accurately identified, so as to facilitate the subsequent use of appropriate prediction methods for domain names with different characteristics.

[0106] In one embodiment, when determining the average bandwidth peak value, the CDN management system may set different weights for the bandwidth peak values ​​of multiple days based on the statistical dates of the bandwidth peak values ​​of multiple days in the historical time period, according to the distance between the statistical dates and the current time. The closer the statistical dates are to the current time, the higher the weight. The size of the weight can reflect the degree of influence of the bandwidth peak values ​​of multiple days in the historical time period on the bandwidth peak average value.

[0107] Use the weighted average formula to perform a weighted average of the bandwidth peaks for multiple days in the historical time period to obtain the average bandwidth peak value of each domain name.

[0108] As described above, by setting different weights for the bandwidth peak values ​​of multiple days based on the statistical dates of the bandwidth peak values ​​of multiple days in the historical time period, and performing weighted averaging on the bandwidth peak values ​​of multiple days to obtain the bandwidth peak average value, and by setting higher weights for adjacent bandwidth peak values, the bandwidth peak average value can more accurately reflect the recent changing trend of bandwidth demand, so that the divided core domain names and long-tail domain names are more in line with the latest bandwidth demand.

[0109] In one embodiment, the predicted average bandwidth may include the average bandwidth of each preset period. Before determining the complete overall predicted bandwidth in the prediction time period, the CDN management system may divide the bandwidth peak average by the maximum value of the average bandwidth of each preset period for each customer domain name to calculate the scaling factor.

[0110] The average bandwidth of each preset period corresponding to the customer domain name is scaled proportionally, and the average bandwidth value of each preset period is multiplied by the scaling factor to obtain the scaled predicted average bandwidth. The maximum value of the average bandwidth of each preset period in the scaled predicted average bandwidth is the same as the bandwidth peak average of each customer domain name.

[0111] As described above, by proportionally scaling the average bandwidth of each preset period corresponding to each customer domain name, the scaled predicted average bandwidth is obtained, ensuring that in the scaled predicted average bandwidth, the relative relationship of the average bandwidth of each preset period remains unchanged, and at the same time, the maximum average bandwidth is matched with the bandwidth peak average value, and the maximum value of the average bandwidth of each preset period after scaling is set as the bandwidth peak average value, which can ensure that the overall predicted bandwidth will not be distorted during the prediction process.

[0112] To further introduce the CDN bandwidth prediction process, Figure 4 A flowchart of another CDN bandwidth prediction method is shown, and the method may include the following steps:

[0113] Step 402: Summarize the traffic data of each customer domain name corresponding to the CDN node to generate bandwidth data.

[0114] In this step, the CDN node collects the traffic data of each customer domain name and aggregates the traffic data to the CDN management system. Taking the preset period of five minutes as an example, the CDN management system aggregates the traffic data of each customer domain name corresponding to the CDN node and generates "customer domain name-business area-bandwidth data" with a five-minute granularity.

[0115] Step 404: pre-process the bandwidth data.

[0116] In this step, the CDN management system identifies and processes abnormal values ​​and missing values ​​in the bandwidth data, identifies abnormal fluctuations in the bandwidth data, and adds weekly feature identifiers from Monday to Sunday and holiday feature identifiers to the bandwidth data.

[0117] Taking the aggregation cycle of one hour as an example, the CDN management system aggregates the regional share data with a granularity of one hour based on the "customer domain name-business area-bandwidth data" with a granularity of five minutes.

[0118] Step 406: Based on the ranking results of each customer domain name, each customer domain name is divided into a core domain name and a long-tail domain name.

[0119] In this step, taking the previous billing cycle as an example, the CDN management system ignores the bandwidth data of abnormally fluctuating historical days in the aggregated bandwidth data, and determines the bandwidth peak value of each day in the previous billing cycle based on the bandwidth data of each customer domain name in the previous billing cycle. The bandwidth peak values ​​of each day in the previous billing cycle are averaged to obtain the bandwidth peak average value.

[0120] Based on the bandwidth peak average value, each customer domain name is sorted to obtain the ranking result of the bandwidth peak average value of each customer domain name. According to the sorting result, the customer domain name that meets the preset sorting conditions is regarded as the core domain name, and the customer domain name that does not meet the preset sorting conditions is regarded as the long-tail domain name.

[0121] Step 408: Calculate the daily feature ratio of the historical week.

[0122] In this step, the CDN management system obtains the bandwidth peak value of each customer domain name every day in the historical week, and takes the average of the daily bandwidth peak values ​​of multiple days in the historical week as the weekly benchmark peak value. For each day in the historical week, the ratio of each daily bandwidth peak value to the weekly benchmark peak value is calculated to obtain the daily feature ratio of the seven days in the historical week. The daily feature ratios of multiple historical weeks are averaged to obtain the averaged daily feature ratio.

[0123] Step 410: Calculate the first-day characteristic ratio, the middle-day characteristic ratio, and the last-day characteristic ratio of historical holidays.

[0124] In this step, the CDN management system obtains the bandwidth peaks of each customer domain name during the historical holidays, and uses the bandwidth peak of the last day before the historical holiday as the holiday benchmark peak. The ratio of the bandwidth peak on the first day of the holiday to the holiday benchmark peak is used as the first-day feature ratio of the historical holiday.

[0125] The average value of the bandwidth peak value of each day in the middle period relative to the holiday benchmark peak value is taken as the middle characteristic ratio of the historical holiday. The ratio of the bandwidth peak value on the last day of the holiday relative to the holiday benchmark peak value is taken as the last day characteristic ratio of the historical holiday.

[0126] The characteristic proportions of the first day, the middle section and the last day in multiple historical holidays are averaged respectively to obtain the averaged characteristic proportions of the first day, the middle section and the last day.

[0127] Step 412: Generate a first predicted bandwidth using the SARIMAX model according to the historical bandwidth corresponding to the core domain name.

[0128] In this step, for each core domain name, the CDN management system inputs the historical bandwidth of the core domain name at a granularity of five minutes in the past seven days into the SARIMAX model to obtain the first predicted bandwidth at a granularity of five minutes in the next day output by the SARIMAX model.

[0129] Step 414: Use a weighted moving average method to generate a second predicted bandwidth according to the historical bandwidth corresponding to the long-tail domain name.

[0130] In this step, for each long-tail domain name, the CDN management system determines the period of the weighted moving average, and generates a second predicted bandwidth of five-minute granularity within the next day based on the historical bandwidth of the long-tail domain name at five-minute granularity in the past seven days and the corresponding weight.

[0131] Step 416: Determine the overall predicted bandwidth by combining the predicted average bandwidth within the next day and the averaged daily feature ratio, the first day feature ratio, the middle day feature ratio, and the last day feature ratio.

[0132] In this step, the CDN management system calculates the scaling factor for each customer domain name by dividing the average bandwidth peak value by the maximum value of the average bandwidth at the five-minute granularity in the next day. The average bandwidth at the five-minute granularity corresponding to the customer domain name is scaled proportionally, and the average bandwidth value of each five-minute granularity is multiplied by the scaling factor to obtain the scaled predicted average bandwidth.

[0133] Taking the next billing cycle as an example, the average bandwidth of each day in the next billing cycle is calculated by combining the predicted average bandwidth in the next day and the average daily characteristic ratio, first day characteristic ratio, middle day characteristic ratio and last day characteristic ratio, with reference to weekly seasonal characteristics and holiday characteristics. The average bandwidth of each day in the next billing cycle is combined to form the overall predicted bandwidth.

[0134] Step 418: Amplify the average bandwidth in the specified time period based on the bandwidth growth multiple indicated by the burst expected requirement.

[0135] In this step, the CDN management system obtains the expected burst requirement corresponding to each customer domain name, and amplifies the overall predicted bandwidth within the specified time period based on the bandwidth growth multiple indicated by the expected burst requirement to obtain the overall predicted bandwidth that meets the expected burst requirement.

[0136] Step 420: Scale the overall predicted bandwidth based on the specified 95 peak values ​​indicated by the checkout requirements.

[0137] In this step, the CDN management system obtains the checkout requirements corresponding to each customer domain name, and divides the specified 95 peak value indicated by the checkout requirements by the predicted 95 peak value of the overall predicted bandwidth to obtain a calculated scaling factor. Using the calculated scaling factor, the average bandwidth of the five-minute granularity in the overall predicted bandwidth is scaled proportionally to obtain the overall predicted bandwidth that meets the checkout requirements.

[0138] Step 422: Allocate service areas to the overall predicted bandwidth based on the area share data of each summary period in the historical time period.

[0139] In this step, the CDN management system allocates business areas to the average bandwidth at a five-minute granularity in the overall predicted bandwidth for the next billing period based on the regional share data at a one-hour granularity of the previous billing period, thereby obtaining complete, accurate and future-proof “customer domain name-business area-bandwidth data” at a five-minute granularity for the next billing period. This can more accurately predict future traffic trends, provide accurate data support for subsequent CDN resource planning and business bandwidth allocation, and provide a reference for the subsequent addition or retirement of CDN nodes, thereby reducing the overall operating cost of CDN services.

[0140] Figure 5 1 is a schematic diagram of the structure of an electronic device according to an exemplary embodiment of the present application. The electronic device may be, for example, a mobile phone, a computer, a digital broadcast terminal, a message transceiver, a game console, a tablet device, a personal digital assistant, a server, a smart home appliance, a car computer, etc. Figure 5 At the hardware level, the electronic device includes a processor 502, an internal bus 504, a network interface 506, a memory 508, and a non-volatile memory 510, and may also include hardware required for other services. The processor 502 reads the corresponding computer program from the non-volatile memory 510 into the memory 508 and then runs it, forming a CDN bandwidth prediction device at the logical level. Of course, in addition to the software implementation, this application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0141] Figure 6 is a block diagram of a CDN bandwidth prediction device according to an exemplary embodiment of the present application. Figure 6 The device may include: a domain name division module 602, a first prediction module 604, a second prediction module 606 and an overall prediction module 608, wherein:

[0142] The domain name division module 602 is used to divide each customer domain name into a core domain name and a long-tail domain name based on the traffic data of each customer domain name corresponding to the CDN node, and the bandwidth requirement of the core domain name is higher than the bandwidth requirement of the long-tail domain name;

[0143] The first prediction module 604 is used to generate a first predicted bandwidth according to the historical bandwidth corresponding to the core domain name by using a time series prediction model;

[0144] The second prediction module 606 is used to generate a second predicted bandwidth according to the historical bandwidth corresponding to the long-tail domain name by using a time series smoothing method, wherein the first predicted bandwidth and the second predicted bandwidth are used to reflect the predicted average bandwidth of each customer domain name in the next day;

[0145] The overall prediction module 608 is used to determine the overall predicted bandwidth within the prediction time period according to the predicted average bandwidth within the future day.

[0146] In an example, the first prediction module 604, before being used to generate the first predicted bandwidth using the time series prediction model according to the historical bandwidth corresponding to the core domain name, further includes: obtaining the daily bandwidth peak value of each customer domain name every day in the historical week; taking the average of the daily bandwidth peak values ​​of multiple days in the historical week as the weekly benchmark peak value; respectively calculating the ratio of each daily bandwidth peak value to the weekly benchmark peak value to obtain the daily feature ratio in the historical week; respectively averaging the daily feature ratios in multiple historical weeks to obtain the averaged daily feature ratio; the overall prediction module 608, when used to determine the overall predicted bandwidth in the prediction time period according to the predicted average bandwidth in the future day, includes: determining the overall predicted bandwidth in combination with the predicted average bandwidth and the averaged daily feature ratio.

[0147] In one example, the first prediction module 604, before being used to generate the first predicted bandwidth using the time series prediction model according to the historical bandwidth corresponding to the core domain name, further includes: obtaining the multi-day bandwidth peak value of each customer domain name during the historical holiday, the multi-day bandwidth peak value during the historical holiday including: the bandwidth peak value on the first day of the holiday, the bandwidth peak value in the middle period, and the bandwidth peak value on the last day of the holiday; taking the ratio of the bandwidth peak value on the first day of the holiday to the holiday benchmark peak value as the first-day feature ratio of the historical holiday, the holiday benchmark peak value is the bandwidth peak value on the last day before the historical holiday; taking the average ratio of the bandwidth peak value in the middle period to the holiday benchmark peak value The average is used as the middle characteristic ratio within the historical holiday; the ratio of the bandwidth peak value of the last day of the holiday to the holiday benchmark peak value is used as the last day characteristic ratio within the historical holiday; the first day characteristic ratio, the middle characteristic ratio and the last day characteristic ratio of multiple historical holidays are averaged respectively to obtain the averaged first day characteristic ratio, the middle characteristic ratio and the last day characteristic ratio; the overall prediction module 608, when used to determine the overall predicted bandwidth within the prediction time period according to the predicted average bandwidth within the future day, includes: combining the predicted average bandwidth and the averaged first day characteristic ratio, the middle characteristic ratio and the last day characteristic ratio to determine the overall predicted bandwidth.

[0148] In one example, the overall prediction module 608 is also used to obtain the burst expected requirement corresponding to each customer domain name, and the burst expected requirement is used to indicate the bandwidth growth multiple within a specified time period, and the specified time period is within the prediction time period; based on the bandwidth growth multiple, the average bandwidth within the specified time period is amplified to obtain the overall predicted bandwidth that meets the burst expected requirement.

[0149] In one example, the overall prediction module 608 is further used to obtain the checkout requirement corresponding to each customer domain name, where the checkout requirement is used to indicate the expected specified 95 peak value; the overall predicted bandwidth is proportionally scaled to obtain the overall predicted bandwidth that meets the checkout requirement, and the predicted 95 peak value of the overall predicted bandwidth that meets the checkout requirement is the same as the specified 95 peak value.

[0150] In an example, the domain name division module 602, when used to divide each customer domain name into a core domain name and a long-tail domain name based on the traffic data of each customer domain name corresponding to the CDN node, includes: determining the bandwidth peak average value of each customer domain name in a historical time period based on the traffic data of each customer domain name; sorting the customer domain names based on the bandwidth peak average value to obtain a sorting result; according to the sorting result, taking the customer domain name that meets the preset sorting condition as the core domain name, and taking the customer domain name that does not meet the preset sorting condition as the long-tail domain name.

[0151] In one example, the predicted average bandwidth includes the average bandwidth of each preset period; the first prediction module 604, before being used to generate the first predicted bandwidth using the timing prediction model according to the historical bandwidth corresponding to the core domain name, also includes: proportionally scaling the average bandwidth of each preset period corresponding to each customer domain name to obtain the scaled predicted average bandwidth; the maximum value of the average bandwidth of each preset period in the scaled predicted average bandwidth is the same as the bandwidth peak average of each customer domain name.

[0152] In an example, the overall prediction module 608 is also used to obtain regional share data for each summary period within a historical time period, the regional share data is used to reflect the distribution of traffic data of each customer domain name in different business areas, and the time length of the historical time period is the same as the time length of the prediction time period; based on the regional share data for each summary period within the historical time period, the average bandwidth of each preset period in the overall predicted bandwidth is allocated to the business area, and the time length of the summary period is longer than the time length of the preset period.

[0153] In one example, the domain name division module 602, when used to determine the bandwidth peak average value of each customer domain name within a historical time period based on the traffic data of each customer domain name, includes: setting different weights for the bandwidth peak values ​​of multiple days within the historical time period based on the statistical dates of the bandwidth peak values ​​of multiple days, the closer the statistical date is to the current date, the higher the weight; and performing weighted averaging of the bandwidth peak values ​​of multiple days to obtain the bandwidth peak average value.

[0154] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0155] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying creative labor.

[0156] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory including instructions. The instructions can be executed by a processor of a CDN bandwidth prediction device to implement any of the methods described in the above embodiments.

[0157] The non-temporary computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc., and the present application is not limited thereto.

[0158] In an exemplary embodiment, a computer program product including a computer program / instruction is further provided. The computer program / instruction can be executed by a processor of a CDN bandwidth prediction device to implement any of the methods described in the above embodiments.

[0159] The above describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0160] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the inventions claimed herein. The present application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

[0161] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, 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 content distribution network CDN bandwidth prediction method, characterized in that: The method comprises: Based on the traffic data of each customer domain name corresponding to the CDN node, the customer domain names are divided into core domain names and long-tail domain names, the bandwidth demand of the core domain name is higher than the bandwidth demand of the long-tail domain name, and the core domain name and the long-tail domain name are divided based on the sorting result of the bandwidth peak average value of each customer domain name in the historical time period; Generate a first predicted bandwidth according to the historical bandwidth corresponding to the core domain name by using a time series prediction model; Using a time series smoothing method, generating a second predicted bandwidth according to the historical bandwidth corresponding to the long-tail domain name, the first predicted bandwidth and the second predicted bandwidth are used to reflect the predicted average bandwidth of each customer domain name in the next day, and the predicted average bandwidth includes the average bandwidth of each preset period; The average bandwidth of each preset period corresponding to each customer domain name is scaled proportionally to obtain a scaled predicted average bandwidth; the maximum value of the average bandwidth of each preset period in the scaled predicted average bandwidth is the same as the bandwidth peak value average of each customer domain name; The overall predicted bandwidth in the prediction time period is determined according to the predicted average bandwidth in the future day, and the overall predicted bandwidth is generated by combining the predicted average bandwidth, weekly seasonal characteristics and holiday characteristics of the prediction time period.

2. The prediction method according to claim 1, characterized in that: Before generating the first predicted bandwidth by using the time series prediction model according to the historical bandwidth corresponding to the core domain name, the prediction method further includes: Get the peak bandwidth of each customer domain name every day in the historical week; The average of the daily bandwidth peak values ​​of multiple days in the historical week is taken as the weekly base peak value; Calculate the ratio of each daily bandwidth peak value to the weekly base peak value to obtain the daily characteristic ratio in the historical week; The daily characteristic ratios in multiple historical weeks are averaged to obtain the average daily characteristic ratio; The determining, according to the predicted average bandwidth in the future day, the overall predicted bandwidth in the predicted time period includes: The overall predicted bandwidth is determined by combining the predicted average bandwidth and the averaged daily characteristic ratio.

3. The prediction method according to claim 1, characterized in that: Before generating the first predicted bandwidth by using the time series prediction model according to the historical bandwidth corresponding to the core domain name, the prediction method further includes: Obtain the bandwidth peak values ​​of each customer domain name for multiple days during historical holidays, where the bandwidth peak values ​​for multiple days during historical holidays include: the bandwidth peak value on the first day of the holiday, the bandwidth peak value in the middle period, and the bandwidth peak value on the last day of the holiday; The ratio of the bandwidth peak value on the first day of the holiday to the holiday base peak value is used as the first-day characteristic ratio of the historical holiday, and the holiday base peak value is the bandwidth peak value on the last day before the historical holiday; The average value of the ratio of the bandwidth peak value in the middle period to the holiday benchmark peak value is used as the middle-segment characteristic ratio in the historical holiday; The ratio of the bandwidth peak value on the last day of the holiday to the holiday benchmark peak value is used as the last day characteristic ratio of the historical holiday; The first-day characteristic ratio, the middle-day characteristic ratio and the last-day characteristic ratio of multiple historical holidays are averaged to obtain the averaged first-day characteristic ratio, the middle-day characteristic ratio and the last-day characteristic ratio; The determining, according to the predicted average bandwidth in the future day, the overall predicted bandwidth in the predicted time period includes: The overall predicted bandwidth is determined by combining the predicted average bandwidth and the averaged first-day characteristic ratio, middle-day characteristic ratio, and last-day characteristic ratio.

4. The prediction method according to claim 1, characterized in that: The prediction method further comprises: Obtaining a burst expected requirement corresponding to each customer domain name, the burst expected requirement being used to indicate a bandwidth growth multiple within a specified time period, the specified time period being within the predicted time period; Based on the bandwidth growth multiple, the average bandwidth in the specified time period is amplified to obtain an overall predicted bandwidth that meets the expected burst requirement.

5. The prediction method according to claim 1, characterized in that: The prediction method further comprises: Obtaining a checkout requirement corresponding to each customer domain name, where the checkout requirement is used to indicate an expected specified 95 peak value; The overall predicted bandwidth is scaled proportionally to obtain an overall predicted bandwidth that meets the checkout requirement, wherein the predicted 95% peak value of the overall predicted bandwidth that meets the checkout requirement is the same as the specified 95% peak value.

6. The prediction method according to claim 1, characterized in that: The traffic data of each customer domain name corresponding to the CDN node is used to divide each customer domain name into a core domain name and a long-tail domain name, including: Based on the traffic data of each customer domain name, determine the bandwidth peak average value of each customer domain name in a historical time period; Based on the bandwidth peak average value, sorting the customer domain names to obtain the sorting result; According to the sorting result, the customer domain names that meet the preset sorting conditions are used as the core domain names, and the customer domain names that do not meet the preset sorting conditions are used as the long-tail domain names.

7. The prediction method according to claim 1, characterized in that: The prediction method further comprises: Obtaining regional share data for each summary period within a historical time period, the regional share data being used to reflect the distribution of traffic data of each customer domain name in different business areas, wherein the length of the historical time period is the same as the length of the forecast time period; According to the area proportion data of each summary period in the historical time period, the service area is allocated to the average bandwidth of each preset period in the overall predicted bandwidth, and the time length of the summary period is longer than the time length of the preset period.

8. The prediction method according to claim 6, characterized in that: The determining, based on the traffic data of each customer domain name, the bandwidth peak average value of each customer domain name in a historical time period includes: Based on the statistical dates of the bandwidth peak values ​​of multiple days in the historical time period, different weights are set for the bandwidth peak values ​​of multiple days, and the closer the statistical date is to the current date, the higher the weight is; A weighted average is performed on the bandwidth peak values ​​for the multiple days to obtain the bandwidth peak average value.

9. A CDN bandwidth prediction device, characterized in that: The device comprises: A domain name division module is used to divide each customer domain name into a core domain name and a long-tail domain name based on the traffic data of each customer domain name corresponding to the CDN node, wherein the bandwidth requirement of the core domain name is higher than the bandwidth requirement of the long-tail domain name, and the core domain name and the long-tail domain name are divided based on the sorting result of the bandwidth peak average value of each customer domain name in a historical time period; A first prediction module, configured to generate a first predicted bandwidth according to a historical bandwidth corresponding to the core domain name by using a time series prediction model; A second prediction module is used to generate a second predicted bandwidth using a time series smoothing method according to the historical bandwidth corresponding to the long-tail domain name, wherein the first predicted bandwidth and the second predicted bandwidth are used to reflect the predicted average bandwidth of each customer domain name in the next day, and the predicted average bandwidth includes the average bandwidth of each preset period; The overall prediction module is used to proportionally scale the average bandwidth of each preset period corresponding to each customer domain name to obtain the scaled predicted average bandwidth; the maximum value of the average bandwidth of each preset period in the scaled predicted average bandwidth is the same as the bandwidth peak average of each customer domain name; based on the predicted average bandwidth in the future day, the overall predicted bandwidth in the prediction time period is determined, and the overall predicted bandwidth is generated by combining the predicted average bandwidth, the weekly seasonal characteristics of the prediction time period, and the holiday characteristics.

10. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor implements the method according to any one of claims 1 to 8 by running the executable instructions.

11. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

12. A computer program product having a computer program / instructions stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images