A push service node load balancing management method based on big data analysis

Abstract

The application relates to the field of push service and discloses a push service node load balancing management method based on big data analysis, which comprises the following steps: S1, collecting performance data of each push service node, including response time, error rate, connection number, request rate, response rate, data transmission speed, memory usage rate and CPU usage rate; S2, analyzing the collected performance data of each push service node, calculating a performance state coefficient of each push service node, and comparing and analyzing the performance state coefficient with a preset performance state threshold interval; S3, generating a load balancing strategy according to the comparison and analysis result in S2; S4, monitoring the performance data of each push service node in real time, detecting the performance change of each push service node, and dynamically adjusting the load balancing strategy according to the change. The application analyzes the performance state change of each recommendation service node in a working period and predicts the performance state in the current working period, so that the weight of each recommendation service node can be dynamically changed, and the purpose of adjusting the load balancing strategy is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of push service, in particular to a push service node load balancing management method based on big data analysis. BACKGROUND

[0002] Push service refers to the technology of pushing data or information from the server to the client. Push service can be used in various scenarios, such as: 1. Real-time message push: push real-time messages to the client, such as chat software, social media; 2. Application update push: push application updates to the client, such as software updates, game updates; 3. Notification push: push notifications to the client, such as push notifications, warning notifications; 4. Data push: push data to the client, such as stock prices, weather information.

[0003] The above push service needs to use multiple push service nodes, and whether the multiple push service nodes match the push task will cause waste or insufficient resources, so big data is introduced. The big data analysis technology can help administrators better understand the performance of the nodes, such as which node has the best performance and which node has the worst performance; but in the process of implementing load balancing management, there are still some problems, for example, the weight cannot be dynamically adjusted according to the performance changes of the push service nodes, so as to achieve the purpose of dynamically adjusting the load balancing management strategy, which causes insufficient availability of resources and also affects user experience and satisfaction. SUMMARY

[0004] The purpose of the present application is to provide a push service node load balancing management method based on big data analysis, which solves the above technical problems.

[0005] The purpose of the present application can be achieved by the following technical solutions:

[0006] A push service node load balancing management method based on big data analysis, comprising:

[0007] S1, collecting performance data of each push service node, including response time, error rate, connection number, request rate, response rate, data transmission speed, memory usage and CPU usage;

[0008] S2, analyzing the performance data of each push service node collected, calculating the performance state coefficient of each push service node, and comparing and analyzing the performance state coefficient with the preset performance state threshold interval;

[0009] S3, generating a load balancing strategy according to the comparison and analysis result in S2;

[0010] S4, real-time monitoring of performance data of each push service node, detecting performance changes of each push service node, and dynamically adjusting load balancing strategy according to the changes.

[0011] As a further technical solution, the process of obtaining the performance state coefficient of each push service node is:

[0012] Obtain the response time A of each push service node j , error rate B j , request rate C j , response rate D j , data transmission speed E j , connection number F j , memory usage rate G j and CPU usage rate H j ;

[0013] Through the formula:

[0014]

[0015] Calculate the performance state coefficient x of each push service node j ;

[0016] Wherein, α1, α2, α3, α4, α5, α6, α7 are preset weight coefficients, K i (f j , g j , h j ) is a judgment function about the remaining connection number f j , the remaining memory capacity g j , and the remaining CPU resource h j .

[0017] As a further technical solution, the process of obtaining the K i (f j , g j , h j ) is:

[0018] Obtain the connection number f ji , memory capacity g ji and CPU resource h ji required by the current push task;

[0019] According to the connection number f ji , memory capacity g ji and CPU resource h ji , establish:

[0020]

[0021] Wherein, is a preset weight factor.

[0022] As a further technical solution, the performance state coefficient x j is compared with the preset performance state threshold interval [x jth , x jch ] for analysis, and the process is as follows:

[0023] If x j ∈ [x jth , x jch ], it is determined that the performance state of the current push service node for the current push task is good.

[0024] If x j , it is determined that the performance state of the current push service node for the current push task is poor.

[0025] As a further technical solution, the process of generating a load balancing strategy according to the comparison and analysis result in S2 is as follows:

[0026] S31, the current push task is compared with each push service node;

[0027] S32, the push service node with poor performance state is removed;

[0028] S33, the remaining push service nodes are arranged in descending order of performance state coefficient, and the push service nodes are set with weights x j from high to low.

[0029] As a further technical solution, the process of dynamically adjusting the load balancing strategy according to the change is as follows:

[0030] S41, the performance state coefficient of each push service node changes with time x j (t) in a statistical period is obtained;

[0031] S42, the performance state coefficient of each push service node changes with time x j0 (t) is obtained based on big data;

[0032] S43, the time point corresponding to x is obtained, and the adjacent two time points are taken as a time period;

[0033] S44, the deviation coefficient of each push service node is obtained according to the comparison and analysis of x j (t) and x j0 (t);

[0034] S45, according to the comparison result of the deviation coefficient and the deviation threshold, it is determined whether to adjust the load balancing strategy of the current push service node.

[0035] As a further technical solution, the process of obtaining the deviation coefficient of each push service node is:

[0036] Through the formula:

[0037]

[0038] The deviation coefficient θ of each push service node is calculated j ;

[0039] Wherein, β y is the preset proportion coefficient corresponding to y sub-periods, t i , t i+1 are the starting time point and the ending time point of the sub-period respectively, and N is the total number of sub-periods.

[0040] As a further technical solution, the working process of the S45 is:

[0041] The deviation coefficient θ j is compared with the deviation threshold θ j0 ;

[0042] If θ j ≥ θ j0 , it is judged that the performance state fluctuation of the recommended service node is large, and the load balancing strategy of the recommended service node needs to be adjusted, and the adjustment process is:

[0043] Through the formula:

[0044]

[0045] The weight of the recommended service node after adjustment is calculated

[0046] If θ j < θ j0 , it is judged that the performance state fluctuation of the service node is small, and no load balancing strategy adjustment is performed.

[0047] The beneficial effects of the present application are:

[0048] (1) The present application analyzes the performance state change in the last working period of each recommended service node, and predicts the performance state in the current working period, so as to dynamically change the weight of each recommended service node, and achieve the purpose of adjusting the load balancing strategy;

[0049] (2) The present application compares the performance state coefficient with the preset performance state threshold interval, and the matching degree of the push service node obviously located in the preset performance state threshold interval with the current push task is moderate, and the push service node not belonging to the preset performance state threshold interval indicates that there is a situation of not meeting the requirements or exceeding the requirements too much, which will cause waste of resources, and therefore the push service node not belonging to the preset performance state threshold interval is not recommended. BRIEF DESCRIPTION OF DRAWINGS

[0050] The present application will be further described below in combination with the drawings.

[0051] Figure 1 The method steps of the present application are shown in the figure. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0053] Please refer to Figure 1 The present application is a push service node load balancing management method based on big data analysis, which comprises:

[0054] S1, collect the performance data of each push service node, including response time, error rate, connection number, request rate, response rate, data transmission speed, memory usage rate and CPU usage rate;

[0055] S2, analyze the collected performance data of each push service node, calculate the performance state coefficient of each push service node, and compare the performance state coefficient with the preset performance state threshold interval;

[0056] S3, generate a load balancing strategy according to the comparison and analysis result in S2;

[0057] S4, real-time monitor the performance data of each push service node, detect the performance change of each push service node, and dynamically adjust the load balancing strategy according to the change.

[0058] In the embodiment, the performance data of each push service node is collected first, and then the collected performance data of each push service node is analyzed to calculate the performance state coefficient of each push service node. The performance state coefficient is compared and analyzed with the preset performance state threshold interval to generate a load balancing strategy according to the comparison and analysis result. Then, the performance data of each push service node is monitored in real time, the performance change of each push service node is detected, and the load balancing strategy is dynamically adjusted according to the change. Through the above technical solution, the performance state of each push service node can be monitored in a weighted round robin manner to achieve different load balancing strategies according to push tasks and achieve the optimal push mode. It should be noted that the process of adjusting the load balancing strategy in the present application is an analysis based on the performance state change of each push service node in a working cycle, and the performance state in the current working cycle is predicted to dynamically change the weight of each push service node to achieve the purpose of adjusting the load balancing strategy.

[0059] The process of obtaining the performance state coefficient of each push service node is:

[0060] Obtain the response time A of each push service node j , error rate B j , request rate C j , response rate D j , data transmission speed E j , connection number F j , memory usage rate G j , and CPU usage rate H j .

[0061] The performance state coefficient x of each push service node is calculated by the formula:

[0062]

[0063] j .

[0064] Wherein, α1, α2, α3, α4, α5, α6, α7 are preset weight coefficients, K i (f j , g j , h j ) is a judgment function about the remaining connection number F j , the remaining memory capacity g j , and the remaining CPU resource h j .

[0065] In the embodiment, a method for obtaining the performance state coefficient of each push service node is provided. Specifically, the data of each parameter is obtained by using existing detection tools, and then substituted into ​The performance state coefficient x of each push service node is calculated j , by It can be seen that when the response time A j , error rate B j , connection number F j , memory usage rate G j and CPU usage rate H j is smaller, the performance of the current recommended service node is better, and the request rate C j , response rate D j , data transmission speed E j is larger, which indicates that the performance of the current recommended service node is better; by K i (f j , g j , h j ), each push task is evaluated, and the performance state coefficient is obtained by comprehensively considering the demand of each different push task and the performance of each push service node, so as to truly reflect the real performance of each recommended service node.

[0066] The process of obtaining K i (f j , g j , h j ) is as follows:

[0067] The connection number f ji , memory capacity g ji and CPU resource h ji required by the current push task are obtained;

[0068] According to the connection number f ji , memory capacity g ji and CPU resource h ji , the following is established:

[0069]

[0070] Wherein, is a preset weight factor.

[0071] In this embodiment, a method for obtaining K i (f j , g j , h j ) is provided, the connection number F ji , memory capacity g ji and CPU resource h ji required by the current push task are obtained first, and then K is obtained; obviously, as long as f j ≤f ji or g j ≤g jior h j ≤ h ji If any of the above conditions is true, it means that the current push service node does not meet the requirements of the current push task, so the value is 0, and the current push service node also does not meet the performance requirements for the current push task, when f j > f ji and g j > g ji and h j > h ji , then By weighted sum of the ratio of the actual state of the three parameters and the actual state required by the current push task, the actual state of the current push service node meeting the current push task is obtained, and a more accurate performance state coefficient x j is obtained to improve the effective implementation of the subsequent load balancing strategy.

[0072] The performance state coefficient x j is compared and analyzed with the preset performance state threshold interval [x jth , x jch ], and the process is as follows:

[0073] If x j ∈ [x jth , x jch ], it is judged that the performance state of the current push service node for the current push task is good.

[0074] If , it is judged that the performance state of the current push service node for the current push task is poor.

[0075] In this embodiment, by comparing the performance state coefficient x j with the preset performance state threshold interval [x jth , x jch ], it is obvious that the matching degree of the push service node located in the preset performance state threshold interval and the current push task is in a moderate state, and the push service node not belonging to the preset performance state threshold interval indicates that there is a situation that does not meet the requirements or exceeds the requirements too much, which will cause waste of resources. Therefore, those not belonging to the preset performance state threshold interval are not recommended.

[0076] The process of generating a load balancing strategy according to the comparison and analysis result in S2 is as follows:

[0077] S31, analyze and compare the current push task with each push service node;

[0078] S32, eliminate the push service node with poor performance state;

[0079] S33, the rest of the push service nodes are arranged in descending order according to the performance state coefficient, and the push service nodes are set with weights X from high to low j .

[0080] The process of dynamically adjusting the load balancing strategy according to the changes is:

[0081] S41, obtain the curve x of the performance state coefficient of each push service node changing with time in a statistical period j (t);

[0082] S42, obtain the parameter curve x of the performance state coefficient of each push service node changing with time based on big data j0 (t);

[0083] S43, obtain The corresponding time points, and the adjacent two time points are taken as a time period;

[0084] S44, according to the comparison of x j (t) and x j0 (t), the deviation coefficient of each push service node is obtained;

[0085] S45, according to the comparison result of the deviation coefficient and the deviation threshold, it is judged whether to adjust the load balancing strategy of the current push service node.

[0086] The process of obtaining the deviation coefficient of each push service node is:

[0087] Through the formula:

[0088]

[0089] The deviation coefficient θ j of each push service node is calculated;

[0090] Where, β y is a preset proportion coefficient corresponding to y time periods, t i , t i+1 are the starting time point and the ending time point of the time period respectively, and N is the total number of time periods.

[0091] The working process of S45 is:

[0092] Compare the deviation coefficient θ j with the deviation threshold θ j0 ;

[0093] If θ j ≥ θ j0 , it is judged that the performance state of the recommended service node fluctuates greatly, and the load balancing strategy of the recommended service node needs to be adjusted, and the adjustment process is:

[0094] Through the formula:

[0095]

[0096] The adjusted weight of the recommendation service node is calculated. γ j The conversion coefficient was selected and determined based on historical experimental data.

[0097] If θ j <θ j0 If the performance status of the service node fluctuates little, then no load balancing strategy adjustment will be performed.

[0098] In this embodiment, the current push task is first analyzed and compared with each push service node, and then push service nodes with poor performance are eliminated. Subsequently, the remaining push service nodes are sorted in descending order of performance coefficient, and a weight X is assigned to each push service node from high to low. j This process is designed to generate a load balancing strategy. However, in actual operation, the performance status coefficient of each push service node is in a dynamic state. Therefore, it is necessary to adjust the weights in combination with the dynamic changes in order to achieve the purpose of dynamically adjusting the load balancing strategy.

[0099] Specifically, first obtain the curve x showing the change of the performance status coefficient of each push service node over time within a statistical period. j (t); and then, based on big data, obtain the parameter curve x of the performance status coefficient of each push service node changing over time. j0 (t); then by obtaining The corresponding time points are divided into two adjacent time periods to segment a statistical period; then, the formula is used... The deviation coefficient θ of each push service node is calculated. j ; the deviation coefficient θ j With deviation threshold θ j0 Comparison, if θ j ≥θ j0 If the performance of the recommendation service node is highly volatile, its load balancing strategy needs to be adjusted. The adjustment process is as follows: using the formula: The adjusted weight of the recommendation service node is calculated. If θ j <θ j0 If the performance status of the service node fluctuates little, then no load balancing strategy adjustment will be performed.

[0100] Through the technical scheme, the deviation between the change curve of the performance state coefficient of the recommended service node in the last statistical period and a reference change curve can be determined, and the weight of the recommended service node in the next statistical period is adjusted, so that the recommended service node after the change of the sexual state coefficient still occupies a large weight, the load balancing strategy cannot be effectively implemented, and finally the overall push quality is affected.

[0101] It should be noted that the calculation formula and the parameters participating in the operation in the present application are all pre-processed by dimensionless processing, and the process of dimensionless processing is known in the industry and will not be described here.

[0102] The above describes one embodiment of the present application in detail, but the content described is only a preferred embodiment of the present application and cannot be considered as limiting the scope of the implementation of the present application. Any equivalent changes and improvements made within the scope of the present application should still belong to the patent coverage of the present application.

Claims

1. A method for managing load balancing of push service nodes based on big data analysis, characterized in that, Comprise: S1, collect the performance data of each push service node, including response time, error rate, connection number, request rate, response rate, data transmission speed, memory usage and CPU usage; S2, analyze the performance data of each push service node collected, calculate the performance state coefficient of each push service node, and compare and analyze the performance state coefficient with the preset performance state threshold interval; S3, generate a load balancing strategy according to the comparison and analysis result in S2; S4, real-time monitoring the performance data of each push service node, detecting the performance change of each push service node, and dynamically adjusting the load balancing strategy according to the change; The process of obtaining the performance state coefficient of each push service node is: obtaining the response time Aj, error rate Bj, request rate Cj, response rate Dj, data transmission speed Ej, connection number Fj, memory usage Gj and CPU usage Hj of each push service node; Through the formula: The performance state coefficient x of each push service node is calculated j ; wherein, α1, α2, α3, α4, α5, α6, α7 are preset weight coefficients, K i (f j , g j , h j ) are judgment functions about the remaining connection number f j , the remaining memory capacity g j , and the remaining CPU resource hj; Obtain the K i (f j , g j , h j ) process is: obtain the required number of connections f ji , memory capacity g ji and CPU resources h ji of the current push task; according to the number of connections f ji , memory capacity g ji and CPU resources h ji establish: Wherein, , , Is a preset weight factor; the performance state coefficient x j Is compared with a preset performance state threshold interval , If , it is judged that the performance state of the current push service node for the current push task is good. If then the current push service node is determined to be in a poor performance state for the current push task.

2. The big data analytics based push service node load balancing management method of claim 1, wherein, The process of generating a load balancing strategy according to the comparison and analysis result in S2 in step S3 is: S31, analyze and compare the current push task with each push service node; S32, eliminate the push service node with poor performance state; S33, the rest of the push service node according to the performance state coefficient descending order, from high to low on the push service node set weight . 3.The big data analysis based push service node load balancing management method according to claim 1 or 2, characterized in that, The process of dynamically adjusting the load balancing strategy according to the change in step S4 is: S41, obtain the performance state coefficient curve of each push service node in a statistical period changing with time ; S42, acquire the parameter curve of the performance state coefficient of each push service node changing with time based on big data ; S43, obtaining The adjacent two time points are taken as a sub-period. S44、According to With The deviation coefficient of each push service node is obtained through comparison and analysis. S45, according to the comparison result of the deviation coefficient and the deviation threshold, judge whether to adjust the load balancing strategy of the current push service node.

4. The big data analytics based push service node load balancing management method of claim 3, wherein, The process of obtaining the deviation coefficient of each push service node is: By the formula: ; The deviation coefficient of each push service node is calculated ; wherein, is a preset proportion coefficient corresponding to the y sub-periods, , are a starting time point and an ending time point of the sub-period respectively, and N is a total number of the sub-periods.

5. The big data analytics based push service node load balancing management method of claim 4, wherein, The working process of step S45 includes: comparing the deviation coefficient with the deviation threshold ; if ≥ , it is judged that the performance state fluctuation of the recommended service node is large, and the load balancing strategy of the recommended service node needs to be adjusted, and the adjustment process is: By the formula: ; The weight of the recommended service node after adjustment is calculated ; is the conversion coefficient; If < , it is determined that the service node performance state fluctuation is small, and no load balancing strategy adjustment is performed.

Images