CN114006833B

CN114006833B - Detection evasion method for ENUM (Enterprise management Module) query fault

Info

Publication number: CN114006833B
Application number: CN202111273801.1A
Authority: CN
Inventors: 黄书涵; 陈淼生; 郑仲嵩
Original assignee: China Telecom Fufu Information Technology Co Ltd
Current assignee: China Telecom Fufu Information Technology Co Ltd
Priority date: 2021-10-29
Filing date: 2021-10-29
Publication date: 2023-04-11
Anticipated expiration: 2041-10-29
Also published as: CN114006833A

Abstract

The invention discloses a detection and evasion method for an ENUM (enterprise management um) query fault, which aims at fault expressions that response time delay of an ENUM server is continuously increased or response rate is continuously reduced. And fitting a function through the data to predict the occurrence of the phenomenon. And (4) by adjusting and reducing the load ratio sent to the corresponding server, observing whether the abnormity is relieved or not, and dynamically planning the next load adjustment and processing. The invention introduces a multi-order curve equation fitting process based on a matrix equation and a computer algorithm to realize the multi-order curve equation fitting process for predicting and calculating the response rate, and realizes the preliminary automation by using an intelligent algorithm. The invention constructs a set of strategy flow for dynamically adjusting and recovering the load ratio, and has higher flexibility and non-manual intervention capability compared with the one-time load transfer.

Description

Detection evasion method for ENUM (Enterprise management Module) query fault

Technical Field

The invention relates to the technical field of 5G communication, in particular to a detection and avoidance method for an ENUM (enterprise management um) query fault.

Background

Under the background of VoLTE and 5G, with the popularization of number portability services, the requirements of service systems represented by short message and message platforms on the performance and reliability of a process for inquiring a large-area ENUM system to acquire user information are sharply improved.

The ENUM query procedure builds on the UDP based DNS protocol. Unlike TCP based on connection, UDP without connection does not guarantee data transmission reliability, and cannot accurately and timely acquire a data transmission result under the condition that the network itself is reachable. In some key processes such as home inquiry of message delivery called party, an ENUM inquiry result cannot be obtained quickly and correctly, which is equal to that message issuing failure is declared directly, thus causing communication failure.

For unreliable fault detection or reliability guarantee of UDP, the current main research schemes at home and abroad are divided into several categories. For example, the method is realized by adding a control identifier or a check bit to a data segment of an IP message of a UDP protocol based on a transport layer; or simulating a data guarantee mechanism of the TCP in the UDP; or sharing the network structure by adopting the safeguard measures such as dual-network cascade and the like; or a check algorithm at the receiving end is used for ensuring the data to be correct, and the like. Because the ENUM/DNS protocol is encapsulated, the reliability improvement achieved by the above modifications to the transport protocol or the network architecture is less feasible.

Under the condition of large-scale mass concurrent query, if the method for judging whether the response returns or not by synchronously waiting for the response causes serial congestion, the concurrency capability is severely restricted in speed.

If the method of sending the state detection packet at regular time for response detection is used, a large number of queries are lost in the process of waiting for the detection result when a fault occurs. And aiming at the conditions of partial loss and overlarge time delay of the response packet, the detection of a single detection packet cannot accurately detect the fault.

Disclosure of Invention

The invention aims to provide a detection and avoidance method for an ENUM query fault.

The technical scheme adopted by the invention is as follows:

a detection evasion method for ENUM inquiry fault comprises the following steps:

step 1, after receiving the response, storing the IDs and the time sent by different servers according to ID inquiry reservation for matching,

step 2, accumulating the sending number and the answering number in the same time slice, and calculating the inquiry unresponsive rate in the time slice;

step 3, judging whether the current time interval is zero load or not; if yes, performing inquiry test and executing step 6; otherwise, executing step 4;

step 4, obtaining the unresponsive rate data of several continuous time slices to form a group of fitting data sequences;

step 5, fitting the fitting data sequence by using a programming language based on the uniqueness theorem existing in the solution of the least square method to obtain a coefficient matrix of a normal equation set,

step 6, substituting the fitted curve function into the self-variation quantity to calculate the predicted value of the unresponsive rate of the next time slice;

step 7, judging whether the non-response rate of a certain server reaches a set threshold value in the following time slice; if yes, executing step 8; otherwise, waiting and executing the step 1;

step 8, adding 1 to the number of times of reaching the threshold value and judging whether the number of times of reaching the threshold value reaches the set number of times; if yes, the load returns to zero and the step 1 is executed; otherwise, the load sharing rate of the server is reduced according to the set rate, and then the step 1 is executed.

Step 9, judging whether the test is normal or not; if yes, after the current load is increased, waiting for the end of the sleep period and executing the step 1; otherwise, waiting for the end of the sleep cycle and executing the step 1;

further, the set number of times in step 8 is 3.

Further, the set ratio in step 8 is: the decreasing load will share half the server load by the ratio.

Further, in step 8, if the response rate is still not increased after the rates of the three consecutive time slices are decreased, the Enum does not share the load and generates an alarm, and notifies the human to perform intervention and troubleshoot the problem.

Further, in step 9, for the server with the load returned to zero, 5 detection queries are fixedly sent per time slice, and once the total number of the detection queries of the consecutive K time slices obtains a response, the load ratio is gradually increased until the original load sharing ratio is restored.

The invention adopts the technical scheme and aims at the fault expression that the response time delay of the ENUM server is continuously increased or the response rate is continuously reduced. And fitting a function through the data to predict the occurrence of the phenomenon. And (4) by adjusting and reducing the load ratio sent to the corresponding server, observing whether the abnormity is relieved or not, and dynamically planning the next load adjustment and processing. The invention introduces a multi-order curve equation fitting process based on a matrix equation and the realization of a computer algorithm for the predictive calculation of the response rate, and realizes the preliminary automation by using an intelligent algorithm. The invention constructs a set of strategy flow for dynamically adjusting and recovering the load ratio, and has higher flexibility and non-manual intervention capability compared with the load transfer of one cutting.

The invention is applied to various service systems which need to inquire the ENUM in real time and have higher requirements on inquiry concurrency and response rate and various mass UDP interactive scenes based on load sharing architectures and one-to-one messages. The invention can avoid the inquiry fault in the service in time, effectively reduce the service call loss rate and improve the system reliability; meanwhile, a large amount of query waiting retransmission can be effectively eliminated, service time delay is reduced, and user perception is improved.

Drawings

The invention is described in further detail below with reference to the accompanying drawings and the detailed description;

FIG. 1 is a flow diagram illustrating a method for detecting and evading ENUM query failures according to the present invention;

FIG. 2 is a graph illustrating test data curves for turn-on 0 response detection;

FIG. 3 is a graph illustrating test data for unopened 0 response detection.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. The invention designs a set of massive ENUM concurrent query error detection and avoidance methods which are different from the traditional processing mode in a load sharing query system based on least square curve fitting of response rate data change trend aiming at the phenomena of response delay and low response rate of an ENUM server by depending on the characteristic that DNS message query and response appear in pairs.

Analyzing various abnormal phenomena of ENUM query and summarizing four situations. 1. And when the network is abnormal, the operating system layer can detect the network abnormality, and the inquiring party can know the fault in real time. 2. The server is hung up, the response rate is instantly returned to zero, and the response rate can be quickly obtained through response statistics of a small time slice. 3. The response speed becomes slower and the query delay is increased. 4. The response rate is continuously reduced, which is expressed by that the number of responses per unit time is less and less. The first 2 phenomena are easy to detect; the latter 2 phenomena cannot be detected in real time and even give wrong pre-judgment results.

Therefore, for the latter 2 cases, it is necessary to adjust and reduce the load ratio sent to the corresponding server, and observe whether the anomaly is relieved, so as to dynamically plan the next load adjustment and processing. In the process, the trend of the processing delay variation and the variation trend of the ratio of the unanswered queries are obtained through calculation, and the next numerical value is predicted.

According to the protocol specification, the ID matching information reserved before query and transmission is matched with the response, and the number of transmission and the number of response are accumulated in a short time slice by different servers to which the ID matching information is transmitted. The information such as the response rate or the response delay in the past time slice can be quickly obtained.

After a group of numerical sequences of response rate or time delay is obtained, an approximate function is solved to fit the group of data, so that the function can well reflect the basic variation trend of the data. And then, obtaining the expected function values of the next N time units according to the trend, if the function is found to have the trend of approaching the load threshold, adopting a load sharing ratio adjustment method to reduce the sending of requests to the server so as to alleviate the data forward curve, avoid the processing peak and increase the query success rate.

According to the delay characteristics under the third and fourth abnormal conditions, the response time and the response loss rate both show a curve-like rise when approaching the threshold, and the derivative (tangent slope) of each monitoring point on the curve is continuously increased until approaching infinity. It can be understood that the cache consumption of the server itself results in a rush. Then a shape such as f (x) = a can be fitted ₀ +a ₁ x+a ₂ x ² +a ₃ x ³ +...+a _n x ⁿ The polynomial curve equation of (a) describes this phenomenon.

According to the equation, the predicted value of each time slice in the current situation can be obtained. By setting a load adjustment strategy, by reducing the sending load of a certain server and matching with real-time data of the next time slice, whether the load is continuously reduced until the load reaches 0 or the load is improved after the load is stable is dynamically planned. Meanwhile, for the server in the 0 load period, the current state is judged by using a mode of sending detection packets in batch, and for the server with normal response, the load sharing proportion can be gradually started and increased. For automatically detecting unrecoverable faults, manual intervention may be alerted.

As shown in fig. 1, the invention discloses a method for detecting and evading ENUM query fault, which comprises the following steps:

step 3, judging whether the current time interval is zero load; if yes, performing inquiry test and executing step 6; otherwise, executing step 4;

step 6, substituting the fitted curve function into the self-variation quantity to calculate the predicted value of the unanswered rate of the next time slice;

step 8, adding 1 to the number of times of reaching the threshold value and judging whether the number of times of reaching the threshold value reaches the set number of times; if yes, the load returns to zero and the step 1 is executed; otherwise, the load is reduced according to the set ratio of the server load sharing ratio and then the step 1 is executed.

further, the set number of times in step 8 is 3.

Further, the set ratio in step 8 is: the decreasing load will load the server by half the ratio.

The working principle of the present invention is explained in detail as follows:

and (3) response rate statistics: by analyzing the message structure defined by the DNS standard protocol RFC1035, it is known that the length of the ID field in the DNS query message packet is 2 bytes and 16 bits, so the ID value range is 0 to 65535, and the ID values of the query message and the corresponding response message are the same and correspondingly matched. A wait structure of size 65536 may be maintained with the sequence number of the free node as the ID and the send time and the number of the server to send are marked in the structure. And after receiving the response, finding the corresponding node according to the response ID. Accumulating the number of transmissions in the current time slice while transmitting; and accumulating the receiving number in the current time slice while receiving the message packet. Therefore, the information such as the response rate or the response time delay sent to the same server in the past time slice can be quickly obtained. Stored as a sequence of numbers.

And (3) fitting a least square method curve equation: the least squares method is a common optimization in fitting calculations by minimizing the sum of the squares of the errors and finding the best functional match of the data. The unknown data is obtained through prediction according to the fitted curve function, and the sum of squares of errors delta between the predicted data and the actual data is minimized.

According to the theorem of uniqueness of existence of least-squares solutions, e.g. taking

Then an algebraic polynomial is obtained:

according to the variance formula, require

The variance of (c) is minimal:

the sum of the squared variances from each point to the curve is:

the right side of the equation is biased for i =0,1,2.. K, yielding:

after the Vandermonde matrix is simplified, the following matrix can be obtained:

the corresponding system of normal equations can be described as: x a = Y, a = (X '× X) -1 × X' = Y can be obtained. The coefficient matrix A is solved, and a polynomial function is obtained

And (3) implementing polynomial fitting by using a program language. The final output result of the program is the coefficient sequence of the polynomial. And substituting the coefficient into the polynomial model to obtain a fitting function. In the process of describing and fitting by using a program language, the coefficient A is subjected to partial derivation by an objective function, and an optimal value condition can be obtained to form a polynomial equation set. And solving the equation set by adopting determinant transformation, converting the original determinant into a trigonometric determinant and then converting the original determinant into a diagonal determinant, and thus solving the result. Because the fitting needs to be carried out by high-order operation, the precision problem of the floating point number needs to be paid attention to all the time in the implementation of the C language, and the number of the floating point number is processed within the range of the allowed error so as to prevent the program from crashing.

Because the overflow risk involved in the high-order operation is large, the method generally takes 4 orders when fitting, and a better effect can be fitted.

Load adjustment and detection recovery: after the polynomial equation is obtained, the predicted value of each time slice under the current situation can be easily obtained according to the value of the independent variable. According to the actual test effect, we generally set the dynamic adjustment policy as follows: if the no-response rate sent to the Enum server by the next time slice reaches M, the Enum load sharing rate is reduced by half by the next time slice, and if the response rate is still not improved after the rates of three continuous time slices are reduced, the Enum does not share the load at all, an alarm is generated, and intervention is performed manually and problems are solved.

Meanwhile, the dispatching process is to the server with the load set to be 0 completely, 5 detection queries are sent in each time slice, once the total number of the detection queries of K time slices is continuously responded, the load ratio is gradually increased by taking each time slice as a unit until the original load sharing ratio is recovered.

As shown in fig. 2, the method according to the present invention uses one query client to perform 200 queries per second for two servers, each server being burdened with 50% of the query load, for several minutes. And setting one of the servers to gradually slow down the speed of sending the inquiry response at a certain time in the sending process. The response is then turned on after 10 seconds of no response at all. While turning on 0 answer detection. The following test data curves were obtained.

The situation of the sending response rate without starting the error detection is shown in fig. 3, and it can be observed that, in the case of starting the error detection, the response rate sudden drop and zero caused by the slow processing of the server at one time are successfully avoided. There are 1355 queries without any result without turning on error detection, and it is necessary to wait for re-query or declare query failure. And 1299 times of queries are loaded and transferred to another server under the condition of starting detection, so that a result without response is avoided, and the one-time transmission success rate is effectively improved. The detection method proves to be successful.

Compared with the prior art, the invention has the main advantages that: 1. compared with a query method based on a synchronous blocking mode, the method can greatly improve the overall message interaction speed by utilizing the advantages of asynchronous transmission, and does not cause serial congestion due to blocking. 2. Compared with the fault detection based on the state detection packet, the method reduces the time for waiting detection, avoids the fault packet loss in the period of time, and solves the faults of partial loss and delay of response, which cannot be detected by the state detection packet in full quantity. 3. Compared with a detection guarantee mode based on protocol modification and network system architecture modification, the method has extremely light weight in development difficulty and system complexity.

It is to be understood that the embodiments described are only a few embodiments of the present application and not all embodiments. The embodiments and features of the embodiments in the present application may be combined with each other without conflict. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the detailed description of the embodiments of the present application is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

Claims

1. A detection evasion method for ENUM inquiry fault is characterized in that: which comprises the following steps:

step 8, adding 1 to the number of times of reaching the threshold value and judging whether the number of times of reaching the threshold value reaches the set number of times; if yes, the load returns to zero and the step 1 is executed; otherwise, the load sharing rate of the server is reduced according to the set rate, and then the step 1 is executed;

step 9, judging whether the test is normal or not; if yes, after the current load is increased, waiting for the end of the sleep period and executing the step 1; otherwise, waiting for the end of the sleep period and executing the step 1.

2. The method as claimed in claim 1, wherein the method for detecting and evading ENUM query fault comprises: and 4, taking the order of 4 when the fitting is carried out in the step 5.

3. The method as claimed in claim 1, wherein the method for detecting and evading ENUM query fault comprises: the set number of times in step 8 is 3.

4. A method of detecting and circumventing ENUM query fault as claimed in claim 1, wherein: the set ratio in step 8 is: the decreasing load will load the server by half the ratio.

5. The method as claimed in claim 1, wherein the method for detecting and evading ENUM query fault comprises: and 8, if the response rate is still not improved after the rates of the three continuous time slices are reduced, the Enum does not share the load any more and generates an alarm, and informs the manual intervention and troubleshooting are performed.

6. The method as claimed in claim 1, wherein the method for detecting and evading ENUM query fault comprises: in step 9, 5 detection queries are fixedly sent to the server with the load returned to zero in each time slice, and once the total number of the detection queries of the continuous K time slices is responded, the load ratio is gradually increased until the original load sharing ratio is recovered.