Root cause server positioning method and root cause server positioning device

By collecting and analyzing the sampling data of the outgoing traffic of the resource pool and the outgoing traffic of each server, and calculating the mutation detection threshold and difference value, the root cause of server positioning problems when the outgoing traffic of the resource pool is muted, improving accuracy and reducing the risk of misjudgment , suitable for large-scale network monitoring.

CN113810207BActive Publication Date: 2024-11-12CHINA TELECOM CLOUD TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010535812.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-06-12
Publication Date
2024-11-12
Estimated Expiration
2040-06-12

AI Technical Summary

Technical Problem

The prior art is difficult to quickly and accurately locate the root server when the outgoing traffic of the resource pool changes suddenly, especially when the traffic sampling time is not synchronized and the risk of misjudgment is high, there is a lack of effective automated inspection and positioning methods.

Method used

By collecting sampling data of the resource pool exit traffic and outflow traffic of each server, the mutation detection threshold and difference value are calculated, the interquartile range and extreme difference value are used to detect traffic mutations, the threshold is dynamically updated to deal with the problem of time out of synchronization, and the resulting exit is determined. The root cause of traffic mutations is the server.

Benefits of technology

It improves the accuracy of root cause server positioning and reduces the misjudgment rate. It is suitable for situations where traffic sampling time is inconsistent in large-scale network monitoring, shortens the device inspection path and reduces the computational complexity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113810207B_ABST
    Figure CN113810207B_ABST
Patent Text Reader

Abstract

The present invention provides a root cause server positioning method and a root cause server positioning device. A root cause server positioning method is used to locate the root cause server that causes the sudden change of the export traffic of a resource pool including multiple servers, and the method comprises: collecting sampling data of the export traffic of the resource pool and sampling data of the outflow traffic of each server in the resource pool; using the sampling data of the export traffic of the resource pool to calculate the mutation detection threshold of the resource pool, and using the mutation detection threshold of the resource pool to perform mutation detection on the export traffic of the resource pool; when a sudden change in the export traffic of the resource pool is detected, the mutation detection threshold of each server is calculated according to the sampling data of the outflow traffic of each server in the resource pool, and the outflow traffic of each server is respectively subjected to mutation detection using the mutation detection threshold of each server to obtain the traffic change amount of each server; and determining the root cause server that causes the sudden change in the export traffic of the resource pool according to the traffic change amount of each server.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a resource pool network operation and maintenance monitoring technology, and in particular to a method and device for locating a root cause server that causes a sudden change in resource pool export traffic. Background Art

[0002] The resource pool monitoring system is an important guarantee for the smooth operation of the cloud platform, and network traffic monitoring is an important component of the resource pool monitoring system. It is also an important means for operation and maintenance personnel to promptly detect network anomalies and troubleshoot network failures.

[0003] The distributed network monitoring system can collect traffic data in two modes: active mode and passive mode. Active and passive refer to the client that collects traffic data. Active mode means that the client periodically collects traffic data and returns it to the server without the need for intervention from the server; passive mode means that the server periodically issues instructions to the client, and the client passively receives the request and collects traffic data, and then returns it to the server. If the active mode is adopted, the pressure on the server can be reduced, but since each client collects data independently and the collection time is not synchronized, the timestamps of the traffic monitoring data may not be the same. If the passive mode is adopted, data is collected according to the instructions of the server, which can alleviate the problem of time asynchrony, but when monitoring hundreds of servers and network devices, the server is under too much pressure and may not receive the latest data in time.

[0004] In large-scale resource pool monitoring, due to the large number of monitoring objects, in order to reduce server pressure, the active mode is generally used to monitor network traffic changes, and the sampling interval is set at the minute level rather than the second level. As a result, the actual collected traffic data has the following characteristics:

[0005] 1) In the same time interval, the timestamps of the traffic sampling points of each port may be inconsistent, and the number of sampling points may also be different;

[0006] 2) Due to the influence of network and device performance, the sampling time interval of each port may be inconsistent with the set value, resulting in fluctuating errors;

[0007] 3) Since the sampling time interval is at the minute level, the monitored traffic is the average value within this interval, which cannot reflect the actual peak change of the traffic;

[0008] 4) Due to the asynchronous sampling time and the error in the sampling interval, the traffic data of each port cannot establish a one-to-one correspondence in time.

[0009] At present, the monitoring of network traffic is mainly used to detect abnormal traffic values. By analyzing the size of network traffic online, abnormal traffic values ​​can be discovered in time and alarm information can be generated. Compared with abnormal traffic values, sudden traffic changes refer to a large increase or decrease in traffic in a short period of time, which reflects the abnormality of the traffic change range. Even if the traffic value at a certain moment does not exceed the alarm threshold, sudden changes may occur.

[0010] There is currently a lack of effective means to automatically troubleshoot and locate the root cause device that causes a sudden change in the resource pool export traffic. Patent CN106209404B provides a method and system for analyzing network abnormal traffic. When a traffic anomaly is detected on a designated port, the similarity with the traffic changes on each upstream port of the lower-level device is calculated; when the similarity is greater than a preset threshold, it is selected as the current device, and then continues to go down step by step until the root device is determined, and this root device is determined as the root cause device that generates the traffic anomaly. However, in the implementation of patent CN106209404B, it was found that this method has two prominent problems:

[0011] 1) The time consistency requirement for sampling data of each port is high. According to the similarity calculation formula of this method, the number of traffic sampling points of each port needs to be the same within the preset time period. In practice, there are deviations in the sampling time and sampling interval of each client traffic, so each port does not necessarily have the same number of traffic sampling points within the same time period.

[0012] 2) There is a high risk of misjudging the root cause device. Traffic from lower-level devices will converge and overlap in upper-level network devices; abnormal traffic at the upper-level device outlet may be the result of abnormal convergence of uplink traffic from multiple different lower-level devices. When using a top-down, step-by-step method to check the root device, due to the presence of traffic sampling errors and traffic superposition factors, once the intermediate calculation deviates and the wrong device branch is selected, it is likely to lead to misjudgment of the root cause device.

[0013] Therefore, in order to solve the problem of sudden changes in resource pool egress traffic, a method and system for quickly locating the root cause server is needed, which can cope with the asynchronization of traffic sampling time and reduce the risk of misjudgment of the root cause device. Summary of the invention

[0014] In view of this, an object of the present invention is to provide a method and device for locating a root cause server that causes a sudden change in resource pool egress traffic.

[0015] According to one aspect of the present invention, a root cause server location method is provided, which is used to locate the root cause server that causes the sudden change of the egress traffic of a resource pool including multiple servers, characterized in that the method comprises:

[0016] Collecting sampled data of the egress traffic of the resource pool and sampled data of the outflow traffic of each server in the resource pool;

[0017] Calculating a mutation detection threshold of the resource pool using sampling data of the egress flow of the resource pool, and performing mutation detection on the egress flow of the resource pool using the mutation detection threshold of the resource pool;

[0018] When a sudden change is detected in the egress flow of the resource pool, a sudden change detection threshold of each server is calculated according to the sampling data of the outflow flow of each server in the resource pool, and a sudden change detection is performed on the outflow flow of each server using the sudden change detection threshold of each server to obtain the flow change amount of each server;

[0019] The root cause server causing the sudden change in the egress traffic of the resource pool is determined according to the traffic change of each server.

[0020] Preferably, when collecting the sampling data of the egress traffic of the resource pool, the egress traffic of the resource pool is sampled at a first sampling frequency to obtain the sampling data of the egress traffic of the resource pool; when collecting the sampling data of the outflow traffic of each server in the resource pool, the outflow traffic of each server is sampled at a second sampling frequency that is the same as or different from the first sampling frequency to obtain the sampling data of the outflow traffic of each server.

[0021] The sampling data is the average value of flow rate changes within the sampling time interval.

[0022] Preferably, when calculating the mutation detection threshold, the interquartile range of the sampling data is used to calculate the mutation detection threshold.

[0023] Preferably, when performing mutation detection on the flow, the mutation detection threshold and the one-step difference and two-step difference of the sampling data are used to detect whether the flow has a mutation.

[0024] Preferably, the step of calculating the mutation detection threshold of the resource pool includes:

[0025] Get the export flow of the resource pool and the sampling time [t1, t2, ..., t n ] corresponding to the sampling data [x1,x2,…,x n ], where n is a natural number greater than 1;

[0026] Calculate [x1,x2,…,x n ]’s first quartile Q1 and third quartile Q3;

[0027] Calculate the difference between the third quartile and the first quartile to get the interquartile range IQR, that is

[0028] IQR=Q3-Q1

[0029] Get [x1,x2,…,x n The maximum value x in ] max and the minimum value x min , calculate the maximum value x max and the minimum value x min The difference between the two is the range RG, that is,

[0030] RG=x max -x min

[0031] The calculation formula for the mutation detection threshold h is:

[0032]

[0033] Wherein, w is a constant greater than 0.

[0034] Preferably, the step of performing mutation detection on the egress flow of the resource pool comprises:

[0035] Get the export flow of the resource pool and the sampling time t n+1 The corresponding sample data x n+1 , calculate the 1-step difference d1 = |x n+1 -x n |, and perform mutation detection on the export traffic of the resource pool by comparing the 1-step difference d1 with the mutation detection threshold h of the resource pool.

[0036] Preferably, when no sudden change in the egress flow of the resource pool is detected by comparing the 1-step difference d1 with the sudden change detection threshold h of the resource pool, the 2-step difference d2 is calculated as |x n+1 -x n-1 |, and perform mutation detection on the export flow of the resource pool by comparing the 2-step difference d2 and the mutation detection threshold h of the resource pool.

[0037] Preferably, when no sudden change in the egress flow of the resource pool is detected by comparing the two-step difference d2 with the sudden change detection threshold h of the resource pool, the egress flow of the resource pool and the sampling time [t2,…,t n ,t n+1 ] corresponds to the sampling data [x2,…,x n ,x n+1 ], and the sampling time [t2,…,t n ,t n+1 ] as [t1,t2,…,t n ], the sampled data [x2,…,x n ,x n+1 ] as [x1,x2,…,xn ], recalculate the mutation detection threshold of the resource pool and use the recalculated mutation detection threshold of the resource pool to perform mutation detection on the export traffic of the resource pool.

[0038] Preferably, assuming that there are m servers in the resource pool, h j represents the server mutation detection threshold of the jth server in the resource pool, j = 1, 2, ..., m, calculate the server mutation detection threshold h of the jth server j The steps include:

[0039] Get the outflow traffic and sampling time of the jth server Corresponding outflow traffic sampling data Wherein, k is a natural number greater than 1;

[0040] calculate The first quartile Q1 j and the third quartile Q3 j ;

[0041] Calculate the difference between the third quartile and the first quartile to get the interquartile range IQR j ,Right now

[0042] IQR j =Q3 j -Q1 j

[0043] Get The maximum value in and minimum value Calculate the maximum value and minimum value The difference between j ,Right now

[0044]

[0045] Mutation detection threshold h j The calculation formula is

[0046]

[0047] Among them, w j is a constant greater than 0.

[0048] Preferably, the step of performing mutation detection on the outflow traffic of the j-th server includes:

[0049] Get the export traffic and sampling time of the jth server The corresponding sampling data Calculate the 1-step difference of the j-th server And by comparing the 1 step difference and the mutation detection threshold h of the jth server j To detect mutations in the outflow traffic of the jth server.

[0050] Preferably, when comparing the 1-step difference of the j-th server and the mutation detection threshold h of the jth server j When no sudden change in the export traffic of the jth server is detected, the 2-step difference of the jth server is calculated And by comparing the 2-step difference and the mutation detection threshold h of the jth server j To detect mutations in the export traffic of the jth server.

[0051] Preferably, the following formula (1) is used to obtain the mutation detection result of each server [r 1 ,…,r j ,…,r m ],

[0052]

[0053] When j When it is 1, it means that the outflow traffic of the jth server has a sudden change. The traffic change p of the jth server is calculated using the following formula (2): j for,

[0054]

[0055] When j When it is 0, it means that the outflow traffic of the j-th server has not changed suddenly, and the traffic change of the j-th server is p j is 0,

[0056] The traffic change of each server is obtained [p 1 ,…,p j ,…,p m ].

[0057] Preferably, the step of determining the root cause server includes:

[0058] Statistics of the mutation detection results of the outflow traffic of each server [r 1 ,…,r j ,…,r m ] j The number of times c is equal to 1,

[0059] When c=1, the server with a mutation detection result of 1 is determined as the first root cause server;

[0060] When c=2, the traffic changes of the two servers with mutation detection results of 1 are compared, and the server with a relatively large traffic change is determined as the first root cause server, and the server with a relatively small traffic change is determined as the second root cause server.

[0061] When c≥3, obtain the traffic changes of the number of servers whose mutation detection results are 1 and the corresponding number of servers with the value of c, and sort the traffic changes from high to low, and select the servers with the top three traffic changes as the first root cause server, the second root cause server, and the third root cause server.

[0062] Preferably, the mutation detection results of the outflow traffic of each server are counted. 1 ,…,r j ,…,r m ] j The number of times c is equal to 1, when c = 0, obtain the sampling time of each server The corresponding sampling data And the sampling time As The sampled data As Recalculate the mutation detection threshold of each server and use the recalculated mutation detection threshold of each server to perform mutation detection on the outflow traffic of each server, re-obtain the traffic change of each server, and determine the root cause server based on the re-obtained traffic change of each server.

[0063] According to another aspect of the present invention, a root cause server locating device is provided, which is used to locate a root cause server that causes a sudden change in egress traffic of a resource pool including multiple servers. The root cause server locating device includes:

[0064] A data collection module, used to collect sampled data of the egress traffic of the resource pool and sampled data of the outflow traffic of each server in the resource pool;

[0065] A threshold calculation module, used to calculate the mutation detection threshold of the resource pool using the sampled data of the egress traffic of the resource pool collected by the data collection module, and used to calculate the mutation detection threshold of each server using the sampled data of the outflow traffic of each server;

[0066] a mutation detection module, configured to perform mutation detection on the egress flow of the resource pool according to the mutation detection threshold of the resource pool calculated by the threshold calculation module, and to perform mutation detection on the outflow flow of each server according to the mutation detection threshold of each server calculated by the threshold calculation module;

[0067] The root cause server location module determines the root cause server causing the sudden change in the egress traffic of the resource pool according to the sudden change detection result of the outflow traffic of each server by the sudden change detection module.

[0068] According to another aspect of the present invention, there is provided a storage medium having computer program instructions stored thereon, and the program implements the above-mentioned root cause server location method when executed by a processor.

[0069] According to the present invention, time-asynchronous traffic sampling data is used to detect changes in the export traffic of the resource pool. When a sudden change in the export traffic of the resource pool is found, the root cause server causing the mutation can be quickly checked and located, thereby improving the accuracy of locating the root cause server and reducing the misjudgment rate.

[0070] The present invention solves the problems of asynchronous flow sampling time, inaccurate sampling interval, and inconsistent number of sampling points in the same time interval through methods such as independent mutation detection, dynamic threshold update, 1-step difference, 2-step difference, and delaying one sampling point. When the flow at the resource pool outlet mutates, the root cause server causing the flow mutation at the resource pool outlet is determined by checking the servers with flow mutations in adjacent time domains. Compared with methods with higher requirements for flow sampling time consistency, the present invention is more suitable for large-scale network monitoring of resource pools using active mode.

[0071] When the resource pool outlet traffic suddenly changes, the outflow traffic changes of each server are directly detected. Compared with the method of detecting the similarity of traffic changes step by step, the method provided by the present invention not only shortens the device inspection path and reduces the computational complexity, but also reduces the misjudgment rate of the root cause server. In particular, when the abnormal traffic of multiple servers causes a sudden change in the resource pool outlet traffic, the method of the present invention can reduce the missed judgment rate of the server and improve the accuracy of locating the root cause server. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a network structure diagram of the resource pool involved in the implementation mode of the present invention.

[0073] Figure 2 It is a structural diagram of a server locating device for locating the root cause of a sudden change in flow at an outlet of a resource pool involved in an implementation manner of the present invention.

[0074] Figure 3 This is a flow chart of a method for locating the server that is the root cause of a sudden change in the resource pool egress traffic.

[0075] Figure 4 is a flow chart showing the sudden change detection of the egress flow data of a resource pool.

[0076] Figure 5 It is a flowchart showing the detection results and traffic change amounts of each server. DETAILED DESCRIPTION

[0077] In order to make those skilled in the art better understand the technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0078] Figure 1 is a network structure diagram of a resource pool involved in an implementation mode of the present invention. Figure 1 As shown, the resource pool includes multiple servers 102, 4 access switches 103, 2 aggregation switches 104, 2 core switches 105, and 2 egress routers 101. The server 102 is connected to the access switch 103, the access switch 103 is connected to the core switch 105 through the aggregation switch 104, the core switch 105 is connected to the egress router 101, and the egress router 101 is connected to the upper network. The network structure of the resource pool can select different levels of switch connection methods according to the number of servers and business requirements, and is not limited to Figure 1 That is, the number of servers 102, access switches 103, aggregation switches 104, core switches 105, and egress routers 101 is not limited to Figure 1 In addition, the level, connection mode and number of switches between the server 102 and the egress router 101 are not limited to Figure 1 The structure shown.

[0079] The embodiments of the present invention provide a server capable of quickly locating the server that causes the sudden change in the egress traffic of the resource pool. In the embodiments of the present invention, the server that causes the sudden change in the egress traffic of the resource pool is called the root cause server. Figure 2 FIG. 1 is a structural diagram of a server locating device for locating the root cause of a sudden change in the flow rate at the outlet of a resource pool according to an embodiment of the present invention. Figure 2 As shown, the root cause server location device for resource pool export traffic mutation includes a data collection module 201, a data storage module 202, a data reading module 203, a threshold calculation module 204, a mutation detection module 205, a root cause server location module 206 and a result display module 207.

[0080] Among them, the data collection module 201 is responsible for collecting the egress flow data of the resource pool and the outflow flow data of each server 102. Among them, the egress flow data of the resource pool is collected by collecting the uplink egress flow data of the egress router 101. In addition, the egress flow data of the resource pool can be the sum of the uplink egress flows of all egress routers 101 contained in the resource pool. The data collection module 201 can collect the egress flow data of the egress router 101 and the outflow flow data of each server 102 at a preset sampling frequency, and then send the sampled flow data to the data storage module 202, and send the sampled flow data to the data reading module 203 at the same time. The data storage module 202 saves the received flow data into the historical database, and the data reading module 203 caches the received current flow data. In addition, in the present invention, the data storage module 202 is not necessary, and the data storage module 202 may not be set, and the data collection module 201 may also collect the flow data and save it.

[0081] As an implementation method of the data collection module 201, the data collection module 201 can be implemented using Zabbix open source software. Here, the egress router 101 can enable the Simple Network Management Protocol (SNMP), and then use the Template SNMP Interface on the Zabbix server to monitor the traffic of the uplink ports of the two egress routers 101, and aggregate the uplink traffic of the two egress routers 101 of the resource pool through the summation (grpsum) function in the Zabbix aggregate checks function, so as to obtain the egress traffic data of the resource pool. Deploy and configure the Zabbix agent (Zabbix Agent) on each server 102, enable the active mode, and use net.if.out and grpsum to realize the data collection of the outflow traffic of each server 102.

[0082] The data reading module 203 obtains the required historical sampling data and the current sampling data in the cache from the data storage module 202 according to the requirements of the threshold calculation module 204 and the mutation detection module 205. In the present invention, the data reading module 203 is not necessary, and the data reading module 203 may not be set. The mutation detection module 205 and the threshold calculation module 204 may also obtain the sampling data collected by the data collection module 201 according to the requirements. In other words, the storage and reading of the sampling data can be adopted in any way, not limited to Figure 2 The method shown.

[0083] The threshold calculation module 204 uses the data reading module 203 to obtain the required traffic data, and then calculates the corresponding mutation detection threshold, and sends the mutation detection threshold to the mutation detection module 205. The method of calculating the mutation detection threshold in the threshold calculation module 204 will be described later. Figure 4 Step 402 and Figure 5 The step 502 is described in detail.

[0084] The mutation detection module 205 uses the data reading module 203 to obtain the required flow data, and obtains the mutation detection threshold from the threshold calculation module 204, detects whether the flow data has a mutation, and calculates the detection result and the flow change. The method of detecting whether the flow data has a mutation in the mutation detection module 205 will be described later. Figure 4 Steps 403 to 407 and Figure 5 The steps 503 to 507 are described in detail.

[0085] The mutation detection module 205 can send the detection result and the traffic change amount to the root cause server location module 206 and the result display module 207 as needed.

[0086] The root cause server location module 206 determines the root cause server that causes the sudden change in the resource pool outlet traffic by analyzing and calculating the resource pool outlet traffic mutation detection result, the resource pool outlet traffic change, the server sudden change detection result, and the server outflow traffic change of the sudden change, and sends the result to the result display module 207. The method for determining the root cause server in the root cause server location module 206 will be described later. Figure 3 The details are described in step 309.

[0087] The result display module 207 is used to output mutation detection results, traffic change, and root cause server information.

[0088] Figure 3 This is a flow chart of the method for locating the server that is the root cause of the sudden change in the resource pool export traffic. Figure 3 As shown, the following steps are included:

[0089] Step 301 : collecting sampling data of the egress traffic of the resource pool and sampling data of the outflow traffic of each server in the resource pool.

[0090] As an example, when collecting sampling data of the outlet flow of the resource pool, the outlet flow of the resource pool may be sampled at a first sampling frequency to obtain sampling data of the outlet flow of the resource pool. When collecting sampling data of the outflow flow of each server in the resource pool, the outflow flow of each server may be sampled at a second sampling frequency that is the same as or different from the first sampling frequency to obtain sampling data of the outflow flow of each server. When collecting sampling data using active mode, the collection time for the outlet flow data of the resource pool and the outflow flow data of each server may not be synchronized.

[0091] The sampled data of the egress traffic of the resource pool and the sampled data of the outflow traffic of each server in the resource pool may be an average value of traffic changes within a sampling time interval.

[0092] In addition, step 301 can be performed by Figure 2 The data collection module 201 in the example can be implemented as described above. For example, the Zabbix API can be used to obtain the sampled data of the outflow flow data of the resource pool and the outflow flow data of each server through the flow monitoring of the Zabbix Server and the ZabbixAgent. In addition, the collection method of the outflow flow data of the resource pool and the outflow flow data of each server is not limited to the method given in the embodiment of the present invention.

[0093] Step 302: Calculate a mutation detection threshold of the resource pool using the sampled data of the egress traffic of the resource pool, and perform mutation detection on the egress traffic of the resource pool using the mutation detection threshold of the resource pool.

[0094] Figure 4 3 is a flow chart showing the sudden change detection of the export flow data of the resource pool in step 302. Figure 4 As shown, the following steps are included.

[0095] Step 401: Obtain historical sampling data [x1, x2, ..., x n ] and the corresponding sampling time data [t1, t2,…, t n ].

[0096] This step can be done, for example, Figure 2 The data reading module 203 shown is implemented.

[0097] Step 402, using the sampled data of the export flow [x1, x2, ..., x n ], calculate the mutation detection threshold h, and the method for calculating the mutation detection threshold h is:

[0098] Calculate [x1,x2,…,xn ]’s first quartile Q1 and third quartile Q3;

[0099] Calculate the difference between the third quartile and the first quartile to get the interquartile range IQR;

[0100] The interquartile range IQR is

[0101] IQR=Q3-Q1

[0102] Calculate [x1,x2,…,x n ] the maximum value x max and the minimum value x min The difference between them is the range RG;

[0103] The range RG is

[0104] RG=x max -x min

[0105] The calculation formula for the mutation detection threshold h is:

[0106]

[0107] Herein, w is a weight coefficient, which can be a number greater than 0 and can be set as needed, for example, can be set to 4.

[0108] This step 402 can be performed, for example, Figure 2 The threshold value calculation module 204 shown in FIG. 1 is implemented by inputting the sampling data [x1, x2, …, x n ], the threshold calculation module 204 outputs the calculated mutation detection threshold h.

[0109] Step 403: Obtain the export flow data of the resource pool and the next sampling time t n+1 The corresponding sample data x n+1 , and calculate the 1-step difference d1.

[0110] Among them, the calculation formula of the 1-step difference d1 is:

[0111] d1=|x n+1 -x n |

[0112] Step 404, perform mutation detection on the export flow of the resource pool by comparing the 1-step difference d1 and the mutation detection threshold h. When the 1-step difference d1 is less than the mutation detection threshold h, it is determined that no mutation of the export flow of the resource pool is detected, and the process goes to step 405. When the 1-step difference d1 is greater than or equal to the mutation detection threshold h, it is determined that a mutation of the export flow of the resource pool is detected, and the process goes to step 408.

[0113] Step 405, calculate the 2-step difference d2.

[0114] The calculation formula of the 2-step difference d2 is

[0115] d2=|x n+1 -x n-1 |

[0116] Step 406, performs mutation detection on the export flow of the resource pool by comparing the 2-step difference d2 and the mutation detection threshold h. When the 2-step difference d2 is less than the mutation detection threshold h, it is determined that no mutation of the export flow of the resource pool is detected, and the process goes to step 407. When the 2-step difference d2 is greater than or equal to the mutation detection threshold h, it is determined that a mutation of the export flow of the resource pool is detected, and the process goes to step 407.

[0117] Step 407, calculate the detection result r and the flow change p, and output the detection result r and the flow change p.

[0118] When a sudden change in the egress flow of the resource pool is detected in step 404 or step 406, the detection result r is marked as 1, for example; when a sudden change in the egress flow of the resource pool is not detected in step 404 or step 406, the detection result r is marked as 0, for example.

[0119] The calculation formula of the test result r can be expressed as follows:

[0120]

[0121] When r is 1, it indicates that the egress flow of the resource pool has a sudden change, and when r is 0, it indicates that the egress flow of the resource pool has no sudden change.

[0122] When a sudden change in the egress flow of the resource pool is detected in step 404, the flow change p can be represented by a 1-step difference d1, and when a sudden change in the egress flow of the resource pool is detected in step 406, the flow change p can be represented by a 2-step difference d2. That is, the flow change p can be represented as follows:

[0123]

[0124] In addition, it should be noted that, as required, the flow change p in step 407 may not be calculated. In addition, the detection result may not necessarily be expressed in the form of 0 and 1, and any expression may be used as long as it can identify whether a mutation is detected.

[0125] In addition, steps 403 to 407 can be performed by Figure 2 The mutation detection module 205 shown in FIG. 2 is implemented by inputting the sample data [x1, …, x n,x n+1 ], the mutation detection module 205 outputs the detection result r and the flow change p.

[0126] Then return to Figure 3 , continue to explain the method of locating the server that is the root cause of the sudden change in resource pool export traffic.

[0127] Step 303, determine whether a sudden change in the egress flow of the resource pool is detected, if yes, proceed to step 305, otherwise, proceed to step 304. For example, if the detection result r is 1, proceed to step 305, if the detection result is 0, proceed to step 304.

[0128] Step 304, updates the sampling data and sampling time data of the export flow of the resource pool, and returns to step 302, recalculates the mutation detection threshold of the resource pool based on the updated sampling data, and uses the recalculated mutation detection threshold to perform mutation detection on the export flow of the resource pool until a mutation in the export flow of the resource pool is detected.

[0129] In step 304, the time window is extended to a sampling time, that is, the sampling data sequence [x1, x2, ..., x n ] is updated to [x2,…,x n ,x n+1 ], the sampling time data [t1, t2,…, t n ] is updated to [t2,…,t n ,t n+1 ] and receives the next sampling time t n+2 The corresponding sample data x n+2 , then returns to step 302, recalculates the corresponding mutation detection threshold, and performs the next round of mutation detection.

[0130] After step 304, when entering step 302, in step 302, the updated sampling data [x2, ..., x n ,x n+1 ] is taken as [x1,x2,…,x n ], the updated sampling time data [t2,…,t n ,t n+1 ] is taken as [t1,t2,…,t n ], calculate the mutation detection threshold h. In addition, the newly acquired sampling data x n+2 As x n+1 , calculate the 1-step difference d1 and the 2-step difference d2.

[0131] Step 305, based on the sampling data of the outflow traffic of each server in the resource pool, calculate the mutation detection threshold of each server, and use the mutation detection threshold of each server to perform mutation detection on the outflow traffic of each server to obtain the detection result and traffic change of each server.

[0132] Assume there are m servers in the resource pool. Figure 5 The steps of obtaining the detection result and traffic change amount of the j-th server are shown, where j=1, 2, ..., m.

[0133] Step 501: Obtain the outflow traffic and sampling time of the jth server according to the preset time window size. The corresponding sampling data in, is the sampling data x of the export traffic of the resource pool n+1 Therefore, the sampling data It means from time t n+1 Start looking back at n+1 historical data.

[0134] This step 501 can be performed, for example, Figure 2 The data reading module 203 shown is implemented.

[0135] Step 502: using the sampled data of the outflow traffic of the jth server Calculate the mutation detection threshold h of the jth server j , calculate the mutation detection threshold h j The method is:

[0136] calculate The first quartile Q1 j and the third quartile Q3 j ;

[0137] Calculate the difference between the third quartile and the first quartile to get the interquartile range IQR j ;

[0138] The interquartile range (IQR) j for

[0139] IQR j =Q3 j -Q1 j

[0140] calculate The maximum value in and minimum value Calculate the maximum value and minimum value The difference between j ;

[0141] The range RG j for

[0142]

[0143] Mutation detection threshold h j The calculation formula is

[0144]

[0145] Among them, w j It can be a constant greater than 0 and can be set as needed, for example, it can be set to 4.

[0146] In addition, it should be noted that in this embodiment, when calculating the mutation detection threshold of each server, the same number of sampled data as that used to calculate the mutation detection threshold of the resource pool is used, but the present invention is not limited thereto. When calculating the mutation detection threshold of each server, a different number of sampled data as that used to calculate the mutation detection threshold of the resource pool may also be used. Even if different numbers of sampled data are used, the method for calculating the mutation detection threshold may be the same.

[0147] This step 502 can be performed, for example, Figure 2 The threshold value calculation module 204 shown in FIG. 2 is implemented by inputting sampling data into the threshold value calculation module 204. The threshold calculation module 204 outputs the calculated mutation detection threshold h j .

[0148] Step 503, using the sampling time of the jth server The corresponding sampling data And calculate 1 step difference

[0149] Among them, the calculation formula of the 1-step difference d1 is:

[0150]

[0151] Step 504, by comparing 1 step difference and mutation detection threshold h j To detect the mutation of the outflow traffic of the jth server, when the 1-step differential Less than the mutation detection threshold h j When , it is determined that no sudden change in the outflow traffic of the jth server is detected, and the process goes to step 505. Greater than or equal to the mutation detection threshold h j , it is determined that a sudden change in the outflow traffic of the j-th server is detected, and the process proceeds to step 507.

[0152] Step 505, calculate 2-step difference

[0153] The 2-step difference The calculation formula is

[0154]

[0155] Step 506, by comparing the 2-step difference and mutation detection threshold h j To detect the mutation of the outflow traffic of the jth server, when the two-step difference Less than the mutation detection threshold h j When , it is judged that no sudden change in the outflow traffic of the jth server is detected. Greater than or equal to the mutation detection threshold h j When , it is judged that a sudden change in the outflow traffic of the jth server is detected.

[0156] Step 507, calculate the detection result r j and flow rate change p j .

[0157] When a sudden change in the outflow traffic of the jth server is detected in step 504 or step 506, the detection result r j For example, if the mark is 1, when no sudden change in the outflow traffic of the jth server is detected in step 504 and step 506, the detection result r j For example, the mark is 0.

[0158] Test results j The calculation formula can be expressed as follows:

[0159]

[0160] Among them, when r j When it is 1, it means that the outflow traffic of the jth server has a sudden change. j When it is 0, it means that there is no sudden change in the outflow traffic of the jth server.

[0161] When a sudden change in the outflow traffic of the jth server is detected in step 504, the traffic change amount p j You can use 1 step difference It means that when a sudden change in the egress flow of the resource pool is detected in step 506, the flow change amount p j You can use 2-step difference In other words, the flow change p j It can be expressed as follows:

[0162]

[0163] By executing Figure 5 The steps shown in the figure are used to obtain the detection results of each server. 1 ,r 2 ,…r j ,…,r m ] and flow rate change [p 1 ,p 2 ,…p j ,…,p m ]. Among them, r j is the detection result of the jth server, p j is the traffic change of the jth server.

[0164] In addition, steps 503 to 507 can be performed, for example, Figure 2 The mutation detection module 205 shown in the figure is implemented by inputting the sampling data of the outflow traffic of the jth server into the mutation detection module 205. The mutation detection module 205 outputs the mutation detection result r about the jth server. j and flow rate change p j .

[0165] Return to Figure 3 Provide explanation.

[0166] Step 306: Determine whether there is a root cause server causing the sudden change in the egress traffic of the resource pool according to the traffic change of each server.

[0167] Specifically, the mutation detection results of the outflow traffic of each server are counted [r 1 ,…,r j ,…,r m The mutation detection results in j The number of times c is equal to 1, when the value of c is 0, it is determined that there is no root cause server, and the process goes to step 307.

[0168] When the value of c is greater than or equal to 1, it is determined that there is a server with a sudden change in outflow traffic, and the process goes to step 309 .

[0169] Step 307, update the sampling data of each server, calculate the mutation detection threshold of each server again based on the sampling data of the outflow traffic of each server in the resource pool, and use the mutation detection threshold of each server to perform mutation detection on the outflow traffic of each server to obtain the detection results and traffic changes of each server.

[0170] The method for updating the sampling data of each server is to obtain the next sampling time of each server. The corresponding sampling data And the sampling data of the outflow traffic of each server Updated to The sampling time Updated to

[0171] After updating the sampling data of each server, the updated sampling time Being Updated sampling data Being Newly acquired Being Recalculate the mutation detection threshold of each server and use the recalculated mutation detection threshold of each server to perform mutation detection on the outflow traffic of each server and re-obtain the detection result and traffic change of each server. These recalculation steps are the same as step 305, and can be adopted Figure 5 Steps shown.

[0172] Step 308 , judging whether there is a root cause server causing the sudden change in the egress traffic of the resource pool according to the traffic change of each server, and the judging method of step 308 is the same as that of step 306 .

[0173] When it is determined in step 308 that the root cause server exists, the process proceeds to step 309. When it is determined in step 308 that the root cause server does not exist, the process proceeds to step 304 to update the sampling data and sampling time of the resource pool export traffic, and then repeatedly executes steps 302 to 309.

[0174] Step 309: Based on the traffic change of each server [p 1 ,p 2 ,…,p m ], determine the root cause server, and output the root cause server.

[0175] The root cause server is determined as follows:

[0176] Statistics of the mutation detection results of the outflow traffic of each server [r 1 ,…,r j ,…,r m The mutation detection results in j The number of times c is equal to 1,

[0177] When c=1, it means that the detection result of one server is 1, that is, r j =1, and r j =1 is determined as the first root cause server;

[0178] When c=2, it means that the detection results of two servers are 1, that is, r i =1, r j =1, and get ri =1, the traffic change of the i-th server p i and r j =1, the traffic change of the jth server p j , and compare the traffic change p of the i-th server i and the traffic change p of the jth server j , determining a server with a relatively large traffic change as a first root cause server, and determining a server with a relatively small traffic change as a second root cause server;

[0179] When c≥3, it means that the detection result of the number of servers corresponding to the value of c is 1, and the traffic change of the servers with the detection result of 1 is obtained, and the traffic change is sorted from high to low, and the top 3 are selected, and the corresponding servers are respectively determined as the first root cause server, the second root cause server and the third root cause server.

[0180] In addition, in step 306, step 308 and step 309, the mutation detection results of the outflow traffic of each server are counted. 1 ,…,r j ,…,r m The mutation detection results in j The number of times c is equal to 1, but the mutation detection results r can also be ignored j The number of times c is equal to 1. For example, in step 306, step 308, and step 309, the flow change amount [p 1 ,p 2 ,…p j ,…,p m ] j The number of times c' is non-zero is used, and the root cause server is determined based on the value of c'. At this time, when detecting the sudden change of the outflow traffic of each server, it is not necessary to obtain the sudden change detection result r j , but only the flow change p is obtained j .

[0181] This step 309 can be performed, for example, Figure 2 The root cause server locating module 206 shown is implemented by inputting the detection results and traffic changes of each server into the root cause server locating module 206, and the root cause server locating module 206 can locate the root cause server according to the detection results and traffic changes of each server.

[0182] In an embodiment of the present invention, when it is determined in step 308 that the root cause server does not exist, it means that the root cause server has not been found according to the detection results of steps 305 and 307, which means that the sudden change in the export traffic of the resource pool may not be caused by the sudden change in the outflow traffic of the server. Therefore, steps 305 or 307 are no longer repeatedly executed separately, but steps 302 to 309 are repeatedly executed again.

[0183] According to an implementation manner of the present invention, time-asynchronous traffic sampling data is used to detect changes in the egress traffic of a resource pool. When a sudden change in the egress traffic of a resource pool is found, the root cause server causing the mutation can be quickly checked and located, thereby improving the accuracy of locating the root cause server and reducing the misjudgment rate.

[0184] In addition, in the specific implementation Figure 3 to Figure 5 In the steps shown, after the main process is started, the process pool can be initialized and the maximum number of processes can be set. For example, when implemented in Python, the process pool can be initialized through the process pool (Pool) of the multiprocessing module, and the maximum number of processes set is not less than 2.

[0185] In addition, the root cause server location task Task request described in steps 305 to 309 can be submitted to the process pool Pool in a non-blocking manner apply_async. When the task request is submitted to the Pool, if the number of processes in the process pool does not reach the specified maximum value, the Pool will enable an idle child process to execute the task request; otherwise, the task request waits for a process in the pool to end, and then the Pool enables this idle child process to execute the root cause server location task Task.

[0186] In addition, the resource pool outlet traffic mutation root cause server location device of the embodiment of the present invention can be implemented by a processing circuit, and each unit such as the data acquisition module 201, the data storage module 202, the data reading module 203, the threshold calculation module 204, the mutation detection module 205, the root cause server location module 206 and the result display module 207 is only a logical module divided according to the specific functions implemented, rather than being used to limit the specific implementation method. In actual implementation, the above-mentioned units can be implemented as independent physical entities, or can also be implemented by a single entity (for example, a processor (CPU or DSP, etc.), an integrated circuit, etc.).

[0187] In some other embodiments, a computer readable storage medium stores computer program instructions, which when executed by a processor implement Figures 3 to 5The steps of the method in the corresponding embodiment. Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, devices, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable non-transient storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0188] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terms used herein are selected to best explain the principles of the embodiments, practical applications, or technical improvements to market technologies, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A root cause server location method, used to locate the root cause server that causes a sudden change in the egress traffic of a resource pool including multiple servers, characterized in that: The method comprises: Collecting sampled data of the egress traffic of the resource pool and sampled data of the outflow traffic of each server in the resource pool; The sudden change detection threshold h of the resource pool is calculated by using the sampling data of the outlet flow of the resource pool, and the sudden change detection threshold h of the resource pool is performed on the outlet flow of the resource pool. The calculation formula of the mutation detection threshold h of the resource pool is: Wherein, w is a constant greater than 0, RG is the ratio of the export flow of the resource pool to the sampling time [t1, t2, ..., t n ] corresponds to the sampling data [x1, x2, ..., x n ], IQR is the difference between the maximum and minimum values ​​of the export flow of the resource pool and the sampling time [t1, t2, ..., t n ] corresponds to the sampling data [x1, x2, ..., x n ] is the difference between the third quartile and the first quartile; When a sudden change is detected in the egress flow of the resource pool, a sudden change detection threshold of each server is calculated according to the sampling data of the outflow flow of each server in the resource pool, and a sudden change detection is performed on the outflow flow of each server using the sudden change detection threshold of each server to obtain the flow change amount of each server; The root cause server causing the sudden change in the egress traffic of the resource pool is determined according to the traffic change of each server.

2. The root cause server location method according to claim 1, wherein: When collecting the sampled data of the outlet flow of the resource pool, the outlet flow of the resource pool is sampled according to a first sampling frequency to obtain the sampled data of the outlet flow of the resource pool, When collecting the sampling data of the outflow traffic of each server in the resource pool, the outflow traffic of each server is sampled according to a second sampling frequency that is the same as or different from the first sampling frequency to obtain the sampling data of the outflow traffic of each server, The sampling data is the average value of flow rate changes within the sampling time interval.

3. The root cause server location method according to claim 1, wherein: When performing mutation detection on traffic, the mutation detection threshold and the one-step difference and two-step difference of the sampling data are used to detect whether the traffic has a mutation.

4. The root cause server location method according to claim 1, wherein: The step of performing mutation detection on the egress flow of the resource pool includes: Get the export flow of the resource pool and the sampling time t n+1 The corresponding sample data x n+1 , calculate the 1-step difference d1 = |x n+1 -x n |, and perform mutation detection on the export flow of the resource pool by comparing the 1-step difference d1 with the mutation detection threshold h of the resource pool.

5. The root cause server location method according to claim 4, wherein: When no sudden change in the egress flow of the resource pool is detected by comparing the 1-step difference d1 with the sudden change detection threshold h of the resource pool, the 2-step difference d2 is calculated as |x n+1 -x n-1 |, and perform mutation detection on the export flow of the resource pool by comparing the 2-step difference d2 and the mutation detection threshold h of the resource pool.

6. The root cause server location method according to claim 5, wherein: When no sudden change in the egress flow of the resource pool is detected by comparing the two-step difference d2 with the sudden change detection threshold h of the resource pool, the egress flow of the resource pool and the sampling time [t2, ..., t n , t n+1 ] corresponds to the sampling data [x2,...,x n , x n+1 ], and the sampling time [t2, ..., t n , t n+1 ] as [t1, t2, ..., t n ], the sampled data [x2, ..., x n , x n+1 ] as [x1, x2, ..., x n ], recalculate the mutation detection threshold of the resource pool and use the recalculated mutation detection threshold of the resource pool to perform mutation detection on the export traffic of the resource pool.

7. The root cause server location method according to claim 1, wherein: Assume that there are m servers in the resource pool, h j represents the server mutation detection threshold of the jth server in the resource pool, j = 1, 2, ..., m, calculate the server mutation detection threshold h of the jth server j The steps include: Get the outflow traffic and sampling time of the jth server Corresponding outflow traffic sampling data Wherein, k is a natural number greater than 1; calculate The first quartile Q1 j and the third quartile Q3 j ; Calculate the difference between the third quartile and the first quartile to get the interquartile range IQR j ,Right now IQR j =Q3 j -Q1 j Get The maximum value in and minimum value Calculate the maximum value and minimum value The difference between j ,Right now Mutation detection threshold h j The calculation formula is Among them, w j is a constant greater than 0.

8. The root cause server location method according to claim 7, wherein: The steps of performing mutation detection on the outflow traffic of the jth server include: Get the export traffic and sampling time of the jth server The corresponding sampling data Calculate the 1-step difference of the j-th server And by comparing the 1 step difference and the mutation detection threshold h of the jth server j To detect mutations in the outflow traffic of the jth server.

9. The root cause server location method according to claim 8, wherein: When comparing the 1-step difference of the j-th server and the mutation detection threshold h of the jth server j When no sudden change in the export traffic of the jth server is detected, the 2-step difference of the jth server is calculated And by comparing the 2-step difference and the mutation detection threshold h of the jth server j To detect mutations in the export traffic of the jth server.

10. The root cause server location method according to claim 9, wherein: Using the following formula (1), we can get the mutation detection results of each server [r 1 , ..., r j , ..., r m ], When j When it is 1, it means that the outflow traffic of the jth server has a sudden change. The traffic change p of the jth server is calculated using the following formula (2): j for, When j When it is 0, it means that the outflow traffic of the j-th server has not changed suddenly, and the traffic change of the j-th server is p j is 0, The traffic change of each server is obtained [p 1 , ..., p j , ..., p m ].

11. The root cause server location method according to claim 10, wherein: The steps to identify the root cause server include: Statistics of the mutation detection results of the outflow traffic of each server [r 1 , ..., r j , ..., r m ] j The number of times c is equal to 1, When c=1, the server with a mutation detection result of 1 is determined as the first root cause server; When c=2, the traffic changes of the two servers with mutation detection results of 1 are compared, and the server with a relatively large traffic change is determined as the first root cause server, and the server with a relatively small traffic change is determined as the second root cause server. When c≥3, obtain the traffic changes of the number of servers whose mutation detection results are 1 and the corresponding number of servers with the value of c, and sort the traffic changes from high to low, and select the servers with the top three traffic changes as the first root cause server, the second root cause server, and the third root cause server.

12. The root cause server location method according to claim 10, wherein: Statistics of the mutation detection results of the outflow traffic of each server [r 1 , ..., r j , ..., r m ] j The number of times c is equal to 1, when C = 0, obtain the sampling time of each server The corresponding sampling data And the sampling time As The sampled data As Recalculate the mutation detection threshold of each server and use the recalculated mutation detection threshold of each server to perform mutation detection on the outflow traffic of each server, re-obtain the traffic change of each server, and determine the root cause server according to the re-obtained traffic change of each server.

13. A root cause server locating device, used to locate a root cause server causing a sudden change in egress traffic of a resource pool including multiple servers, the root cause server locating device comprising: A data collection module, used to collect sampled data of the egress traffic of the resource pool and sampled data of the outflow traffic of each server in the resource pool; A threshold calculation module is used to calculate the mutation detection threshold h of the resource pool using the sampled data of the egress flow of the resource pool collected by the data collection module, and is used to calculate the mutation detection threshold of each server using the sampled data of the outflow flow of each server, The calculation formula of the mutation detection threshold h of the resource pool is: Wherein, w is a constant greater than 0, RG is the ratio of the export flow of the resource pool to the sampling time [t1, t2, ..., t n ] corresponds to the sampling data [x1, x2, ..., x n ], IQR is the difference between the maximum and minimum values ​​of the export flow of the resource pool and the sampling time [t1, t2, ..., t n ] corresponds to the sampling data [x1, x2, ..., x n ] is the difference between the third quartile and the first quartile; a mutation detection module, configured to perform mutation detection on the egress flow of the resource pool according to the mutation detection threshold of the resource pool calculated by the threshold calculation module, and to perform mutation detection on the outflow flow of each server according to the mutation detection threshold of each server calculated by the threshold calculation module; The root cause server location module determines the root cause server causing the sudden change in the egress traffic of the resource pool according to the sudden change detection result of the outflow traffic of each server by the sudden change detection module.

14. A computer-readable storage medium having computer program instructions stored thereon, wherein when the program is executed by a processor, the root cause server location method according to any one of claims 1 to 12 is implemented.