A server fault diagnosis rule screening method based on GRA

Abstract

The present invention provides a server fault diagnosis rule screening method based on GRA, and the method steps are as follows: traverse each fault and each rule, calculate the rule fault matching rate, and obtain a fault matching rate sequence; remove the number of faults with low fault levels by ignoring the minimum value algorithm between partitions, calculate the severity of the rule matching fault according to the preset weight of each fault level and the corresponding number of faults, and obtain a fault severity sequence; calculate the association degree of any two rules by the GRA algorithm, calculate the association degree of each rule with the rule database set, obtain the association degree sequence, and then turn to obtain the turn association degree sequence; calculate the weight value of the fault matching rate sequence, the fault severity sequence and the turn similarity sequence by the inverse variance algorithm, and obtain the good and bad value sequence by weighted summation, and set the priority for each rule according to the good and bad value. The present invention gives priority to using rules with high good and bad values ​​for diagnosis, reduces the redundancy of repeated diagnosis, and improves the diagnosis speed.

Description

Technical Field

[0001] The invention belongs to the technical field of server fault diagnosis, and in particular relates to a server fault diagnosis rule screening method based on GRA. Background Art

[0002] GRA is the abbreviation of Gray Relation Analysis, which is gray relational analysis.

[0003] ISCDS is a server log diagnostic tool.

[0004] As the demand for big data processing grows, there is an urgent need for faster and more accurate data processing methods, and the requirements for the stability and security of data centers are getting higher and higher. The stability and security of data centers are related to each server. The current way to ensure the safety and stability of servers is to apply fault diagnosis rules to various performance indicators of the server. In a large number of practices, we have summarized a large number of diagnostic rule bases, but in actual applications, we have found that due to the large amount of data in the rule base and the certain similarity of the rules, the diagnostic process is cumbersome and time-consuming. Therefore, how to obtain fast and accurate diagnostic rules for diagnosing various performance indicators of the server is an urgent problem to be solved.

[0005] This is a shortcoming of the prior art. Therefore, in view of the above defects in the prior art, it is very necessary to provide a server fault diagnosis rule screening method based on GRA. Summary of the invention

[0006] In view of the defects of the existing server fault diagnosis rule bases in the prior art, such as the large number of rule bases and high similarity, which causes redundancy in the diagnosis process and a long time consumption, the present invention provides a server fault diagnosis rule screening method based on GRA to solve the above technical problems.

[0007] The present invention provides a server fault diagnosis rule screening method based on GRA, comprising the following steps:

[0008] S1. Traverse each fault in the server fault record and each rule in the rule database, calculate the fault matching rate of each rule, and obtain the fault matching rate sequence of each rule in the rule database;

[0009] S2. Count the faults that each rule successfully matches, and count the number of faults corresponding to each fault level, and then remove the number of faults corresponding to low fault levels through the interval-ignoring minimum value algorithm, and then calculate the severity of each rule matching fault based on the preset weight of each fault level and the number of faults of the fault level, and obtain the fault severity sequence of each rule in the rule database set;

[0010] S3. Calculate the correlation between any two rules in the rule database set by the GRA algorithm, and then calculate the correlation between each rule and the rule database set, and obtain the correlation sequence of each rule in the rule database set, and then perform a redirection process on the correlation sequence to obtain a redirection correlation sequence;

[0011] S4. Calculate the weight values ​​of the fault matching rate sequence, fault severity sequence and steering similarity sequence through the inverse variance algorithm, and perform weighted summation on the fault matching rate sequence, fault severity sequence and steering similarity sequence to obtain a sequence of good and bad values, and then set the priority for each rule according to the good and bad values.

[0012] Furthermore, the specific steps of step S1 are as follows:

[0013] S11. Obtaining a fault status database of server fault records and a rule base data set for server fault diagnosis; each fault in the fault status database includes several fault states, each rule in the rule database includes several indicators, and the number of fault states of the fault is equal to the number of indicators of the rule;

[0014] S12. A rule in the positioning rule database is used to calculate the deviation rate between each fault state data of each fault and the corresponding indicator data of the positioning rule in turn. If each deviation rate is less than the set ratio, it is determined that the corresponding fault matches the positioning rule successfully. The matching situation of each fault and the positioning rule is counted to obtain the fault matching rate of the positioning rule. The next rule in the rule database is located in turn until the fault matching rate of all rules in the positioning rule database is calculated. The fault state and the indicator of the rule are in a one-to-one correspondence. When the deviation rate between the fault state and the indicator corresponding to a rule and a fault is less than the set ratio, the task is that the rule matches the fault successfully.

[0015] Furthermore, the indicators of the rule include the fault level, the underlying data read error rate, the remapped sector count, the read channel margin, the spindle spin-up retry count, and the head calibration retry count. The indicators of the rule include but are not limited to the above, and any performance indicator used to determine the server failure can be added, and the fault status is the same as the rule indicator item.

[0016] Furthermore, the specific steps of step S12 are as follows:

[0017] S121. A rule in the positioning rule database;

[0018] S122. Locate a fault in the fault status database;

[0019] S123. Calculate the absolute value of the difference between each fault state data of the fault location and the corresponding indicator data of the location rule, and calculate the deviation rate of each absolute value and the corresponding indicator data;

[0020] S124. Determine whether each deviation rate is less than a set ratio;

[0021] If yes, the positioning fault matches the positioning rule successfully, the matching rate of the positioning rule is increased by one, and the process proceeds to step S125;

[0022] If not, the location fault fails to match the location rule;

[0023] S125. Determine whether all faults in the fault status database have been located;

[0024] If yes, count the matching situation between the positioning rule and each fault, obtain the matching rate of the positioning rule, and go to step S126;

[0025] If not, locate the next fault and return to step S123;

[0026] S126. Determine whether all the centralized rules in the positioning rule database have been positioned;

[0027] If so, the matching rates of each rule in the rule database set form a fault matching rate sequence;

[0028] If not, locate the next rule and return to step S122. Each rule is compared with all faults in the fault status database to obtain the fault matching rate of the rule. The fault matching rates of multiple rules form a sequence.

[0029] Furthermore, the specific steps of step S2 are as follows:

[0030] S21. Obtaining the pre-divided fault levels, the preset weights of each fault level, and the fault levels using the partition interval ignoring minimum value algorithm;

[0031] S22. A rule in the positioning rule database;

[0032] S23. Count the faults and corresponding fault levels that successfully match the positioning rules, and determine the faults and number within each fault classification;

[0033] S24. Determine whether there is a corresponding fault within the fault level using the partition interval ignoring minimum value algorithm;

[0034] If yes, the fault level of the algorithm using the minimum value between partitions is set to a low fault level, and the process goes to step S25;

[0035] If not, proceed to step S26;

[0036] S25. Determine whether the ratio of the number of faults within the low-level fault level to the total number of faults in the positioning rule is less than the ratio threshold;

[0037] If yes, set the number of faults in the low-level fault level to zero and proceed to step S26;

[0038] If not, proceed to step S26;

[0039] S26. Multiply the fault level value by the corresponding preset weight as a new weight, and use the new weight to weight the sum of the number of faults within the corresponding fault level to obtain the fault severity of the positioning rule;

[0040] S27. Determine whether all the centralized rules in the rule database have been located;

[0041] If so, the fault severity of each rule in the rule database is formed into a fault severity sequence;

[0042] If not, locate the next rule and return to step S23. The fault number of low-level fault levels is weakened by ignoring the minimum value algorithm between partitions, and the fault severity of the rule is calculated mainly by the fault number of high-level fault levels.

[0043] Further, in step S21, the pre-classified fault levels are 10 levels, wherein the fault level 1 has the lowest severity and the fault level 10 has the highest severity;

[0044] Level 1, 2, and 3 fault levels are set as minor faults, level 4, 5, 6, and 7 fault levels are set as moderate faults, and level 8, 9, and 10 fault levels are set as severe faults.

[0045] Further, in step 24, it is determined whether the ratio of the number of minor faults to the total number of faults of the positioning rule is less than the first ratio, or it is determined whether the ratio of the number of moderate faults to the total number of faults of the positioning rule is less than the second ratio, wherein the second ratio is less than the first ratio.

[0046] Furthermore, the specific steps of step S3 are as follows:

[0047] S31. A rule in the positioning rule database;

[0048] S32. Select a rule from the remaining rules in the rule database set;

[0049] S33. Obtain the indicator data of the positioning rules and the selection rules, and obtain the positioning rule indicator sequence and the selection rule indicator sequence;

[0050] S34. Selecting a maximum value from the positioning rule indicator sequence, and using the maximum value as a reference to standardize the positioning rule indicator sequence, to obtain a standardized positioning rule indicator sequence;

[0051] S35. Selecting a maximum value from the selection rule indicator sequence, and normalizing the selection rule indicator sequence with the maximum value of the sequence as a reference to obtain a normalized selection rule indicator sequence;

[0052] S36. Set the resolution coefficient, and calculate the correlation between the positioning rule and a certain indicator in the selection rule through the GRA algorithm, and then take the average of all indicators to obtain the correlation between the positioning rule and the selection rule;

[0053] S37. Determine whether the remaining rules of the rule database set have been selected;

[0054] If so, take the average of the correlation between the positioning rule and each selection rule to obtain the correlation between the positioning rule and the rule database set;

[0055] If not, select the next rule from the remaining rules in the rule database set and return to step S33;

[0056] S38. Determine whether all the centralized rules in the positioning rule database have been positioned;

[0057] If so, the relevance of each rule in the rule database set to the rule database set forms a relevance sequence;

[0058] If not, locate the next rule and return to step S32;

[0059] S39. Take the maximum value of the correlation sequence, subtract the maximum value from each element in the correlation sequence, and then take the absolute value to generate a new sequence, which is set as the steering correlation sequence. The rules with strong correlation with other rules in the rule database will have a large similarity in the output information, which will eventually lead to a high probability of redundant output information, affecting the diagnosis speed. Therefore, rules with low correlation with other rules should be retained as much as possible; steering the correlation sequence ensures that the gap between the correlation sequence values ​​remains unchanged, and unifies the change trend with the fault matching rate sequence and the fault severity sequence.

[0060] Furthermore, in step S36, the value of the resolution system is greater than or equal to 0 and less than or equal to 1.

[0061] Furthermore, the specific steps of step S4 are as follows:

[0062] S41. Calculate the variances of the fault matching rate sequence, the fault severity sequence and the steering correlation sequence respectively, and they are the first variance, the second variance and the third variance respectively;

[0063] S42. Using the inverse of the first variance as the first weight, the inverse of the second variance as the second weight, and the inverse of the third variance as the third weight;

[0064] S43. weighting the calculated fault matching rate sequence with the first weight, weighting the fault severity sequence with the second weight, and weighting the steering correlation sequence with the third weight, and then summing them to obtain a sequence of merit values;

[0065] S44. Sort the sequence of merits and demerits, set high priority for the rules with high merits and demerits, and provide them to the operation and maintenance personnel for server log analysis and diagnosis. The merits and demerits sequence covers the characteristics of the fault matching rate sequence, the fault severity sequence, and the steering correlation sequence. It is used as a standard for judging server rules and provided to server operation and maintenance personnel. The larger the merits and demerits, the higher the prediction accuracy of this rule, the more serious the matching fault, and the lower the degree of coverage in the rule database. In this case, the rule will be applied first in the diagnosis process to remove redundant diagnosis results, save diagnosis time, and improve accuracy.

[0066] The beneficial effects of the present invention are:

[0067] The server fault diagnosis rule screening method based on GRA provided by the present invention uses a large amount of existing fault status data and rule indicator data, and analyzes three factors of rule fault matching rate, matching fault severity and rule correlation through the interval ignored minimum value algorithm, GRA method and variance inverse algorithm. Under the premise of ensuring the accuracy of the rules, the correlation between the rules is added, and the rules with low coverage in the rule database are preferentially used for diagnosis, which reduces the redundancy of repeated diagnosis and improves the diagnosis speed. The merits and demerits of the rules can also provide a reference for server operation and maintenance personnel, helping them to make fault judgments quickly and accurately.

[0068] In addition, the invention has a reliable design principle, a simple structure and a very broad application prospect.

[0069] It can be seen that compared with the prior art, the present invention has outstanding substantive features and significant progress, and the beneficial effects of its implementation are also obvious. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0071] Figure 1 The present invention is a schematic diagram of the server fault diagnosis rule screening method based on GRA Figure 1 .

[0072] Figure 2 The present invention is a schematic diagram of the server fault diagnosis rule screening method based on GRA Figure 2.

[0073] Figure 3 The present invention is a schematic diagram of the server fault diagnosis rule screening method based on GRA Figure 3 .

[0074] Figure 4 The present invention is a schematic diagram of the server fault diagnosis rule screening method based on GRA Figure 4 .

[0075] Figure 5 The present invention is a schematic diagram of the server fault diagnosis rule screening method based on GRA Figure 5 . DETAILED DESCRIPTION

[0076] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0077] Embodiment 1:

[0078] like Figure 1 As shown, the present invention provides a server fault diagnosis rule screening method based on GRA, comprising the following steps:

[0079] S1. Traverse each fault in the server fault record and each rule in the rule database, calculate the fault matching rate of each rule, and obtain the fault matching rate sequence of each rule in the rule database;

[0080] S2. Count the faults that each rule successfully matches, and count the number of faults corresponding to each fault level, and then remove the number of faults corresponding to low fault levels by ignoring the minimum value method between partitions, and then calculate the severity of each rule matching fault based on the preset weight of each fault level and the number of faults of the fault level, and obtain the fault severity sequence of each rule in the rule database set;

[0081] S3. Calculate the correlation between any two rules in the rule database set by the GRA algorithm, and then calculate the correlation between each rule and the rule database set, and obtain the correlation sequence of each rule in the rule database set, and then perform a redirection process on the correlation sequence to obtain a redirection correlation sequence;

[0082] S4. Calculate the weight values ​​of the fault matching rate sequence, fault severity sequence and steering similarity sequence through the inverse variance algorithm, and perform weighted summation on the fault matching rate sequence, fault severity sequence and steering similarity sequence to obtain a sequence of good and bad values, and then set the priority for each rule according to the good and bad values.

[0083] Embodiment 2:

[0084] like Figure 1 , Figure 2 , Figure 3 , Figure 4 and Figure 5 As shown, the present invention provides a server fault diagnosis rule screening method based on GRA, comprising the following steps:

[0085] S1. Traverse each fault in the server fault record and each rule in the rule database, calculate the fault matching rate of each rule, and obtain the fault matching rate sequence of each rule in the rule database; the specific steps are as follows:

[0086] S11. Obtaining a fault status database of server fault records and a rule base data set for server fault diagnosis; each fault in the fault status database includes several fault states, each rule in the rule database includes several indicators, and the number of fault states of the fault is equal to the number of indicators of the rule;

[0087] S12. A rule in the positioning rule database is selected, and the deviation rate between each fault state data of each fault and the corresponding indicator data of the positioning rule is calculated in turn. If each deviation rate is less than the set ratio, it is determined that the corresponding fault matches the positioning rule successfully, and the matching situation of each fault and the positioning rule is counted to obtain the fault matching rate of the positioning rule, and the next rule in the rule database is located in turn until the fault matching rate of all rules in the positioning rule database is calculated; the specific steps are as follows:

[0088] S121. A rule in the positioning rule database;

[0089] S122. Locate a fault in the fault status database;

[0090] S123. Calculate the absolute value of the difference between each fault state data of the fault location and the corresponding indicator data of the location rule, and calculate the deviation rate of each absolute value and the corresponding indicator data;

[0091] S124. Determine whether each deviation rate is less than a set ratio;

[0092] If yes, the positioning fault matches the positioning rule successfully, the matching rate of the positioning rule is increased by one, and the process proceeds to step S125;

[0093] If not, the location fault fails to match the location rule;

[0094] S125. Determine whether all faults in the fault status database have been located;

[0095] If yes, count the matching situation between the positioning rule and each fault, obtain the matching rate of the positioning rule, and go to step S126;

[0096] If not, locate the next fault and return to step S123;

[0097] S126. Determine whether all the centralized rules in the positioning rule database have been positioned;

[0098] If so, the matching rates of each rule in the rule database set form a fault matching rate sequence;

[0099] If not, locate the next rule and return to step S122;

[0100] S2. Count the faults that each rule successfully matches, and count the number of faults corresponding to each fault level, and then remove the number of faults with low fault levels by ignoring the minimum value algorithm between partitions. Then, according to the preset weight of each fault level and the number of faults corresponding to the fault level, calculate the severity of the fault matched by each rule, and obtain the fault severity sequence of each rule in the rule database set; the specific steps are as follows:

[0101] S21. Obtaining the pre-divided fault levels, the preset weights of each fault level, and the fault levels using the partition interval ignoring minimum value algorithm;

[0102] S22. A rule in the positioning rule database;

[0103] S23. Count the faults and corresponding fault levels that successfully match the positioning rules, and determine the faults and number within each fault classification;

[0104] S24. Determine whether there is a corresponding fault within the fault level using the partition interval ignoring minimum value algorithm;

[0105] If yes, the fault level of the algorithm using the minimum value between partitions is set to a low fault level, and the process goes to step S25;

[0106] If not, proceed to step S26;

[0107] S25. Determine whether the ratio of the number of faults within the low-level fault level to the total number of faults in the positioning rule is less than the ratio threshold;

[0108] If yes, set the number of faults in the low-level fault level to zero and proceed to step S26;

[0109] If not, proceed to step S26;

[0110] S26. Multiply the fault level value by the corresponding preset weight as a new weight, and use the new weight to weight the sum of the number of faults within the corresponding fault level to obtain the fault severity of the positioning rule;

[0111] S27. Determine whether all the centralized rules in the rule database have been located;

[0112] If so, the fault severity of each rule in the rule database is formed into a fault severity sequence;

[0113] If not, locate the next rule and return to step S23;

[0114] S3. Calculate the correlation between any two rules in the rule database set by using the GRA algorithm, and then calculate the correlation between each rule and the rule database set, and obtain the correlation sequence of each rule in the rule database set, and then perform a redirection process on the correlation sequence to obtain a redirected correlation sequence; the specific steps are as follows:

[0115] S31. A rule in the positioning rule database;

[0116] S32. Select a rule from the remaining rules in the rule database set;

[0117] S33. Obtain the indicator data of the positioning rules and the selection rules, and obtain the positioning rule indicator sequence and the selection rule indicator sequence;

[0118] S34. Selecting a maximum value from the positioning rule indicator sequence, and using the maximum value as a reference to standardize the positioning rule indicator sequence, to obtain a standardized positioning rule indicator sequence;

[0119] S35. Selecting a maximum value from the selection rule indicator sequence, and normalizing the selection rule indicator sequence with the maximum value of the sequence as a reference to obtain a normalized selection rule indicator sequence;

[0120] S36. Set the resolution coefficient, and calculate the correlation between the positioning rule and a certain indicator in the selection rule through the GRA algorithm, and then take the average of all indicators to obtain the correlation between the positioning rule and the selection rule;

[0121] S37. Determine whether the remaining rules of the rule database set have been selected;

[0122] If so, take the average of the correlation between the positioning rule and each selection rule to obtain the correlation between the positioning rule and the rule database set;

[0123] If not, select the next rule from the remaining rules in the rule database set and return to step S33;

[0124] S38. Determine whether all the centralized rules in the positioning rule database have been positioned;

[0125] If so, the relevance of each rule in the rule database set to the rule database set forms a relevance sequence;

[0126] If not, locate the next rule and return to step S32;

[0127] S39. Take the maximum value of the association degree sequence, subtract the maximum value from each element in the association degree sequence and take the absolute value to generate a new sequence, and set the new sequence as the steering association degree sequence;

[0128] S4. Calculate the weight values ​​of the fault matching rate sequence, fault severity sequence and steering similarity sequence by using the inverse variance algorithm, and perform weighted summation on the fault matching rate sequence, fault severity sequence and steering similarity sequence to obtain a sequence of good and bad values, and then set the priority for each rule according to the good and bad values; the specific steps are as follows:

[0129] S41. Calculate the variances of the fault matching rate sequence, the fault severity sequence and the steering correlation sequence respectively, and set them as the first variance, the second variance and the third variance respectively;

[0130] S42. Using the inverse of the first variance as the first weight, the inverse of the second variance as the second weight, and the inverse of the third variance as the third weight;

[0131] S43. weighting the calculated fault matching rate sequence with the first weight, weighting the fault severity sequence with the second weight, weighting the steering correlation sequence with the third weight, and then summing them to obtain a sequence of merit values;

[0132] S44. Sort the sequence of good and bad values, set high priority for the rules with high good and bad values, and provide them to operation and maintenance personnel for server log analysis and diagnosis.

[0133] Embodiment 3:

[0134] like Figure 1 , Figure 2 , Figure 3 , Figure 4 and Figure 5 As shown, the present invention provides a server fault diagnosis rule screening method based on GRA, comprising the following steps:

[0135] S1. Traverse each fault in the server fault record and each rule in the rule database, calculate the fault matching rate of each rule, and obtain the fault matching rate sequence of each rule in the rule database; the specific steps are as follows:

[0136] S11. Obtain a fault status database of server fault records and a rule base data set for server fault diagnosis; each fault in the fault status database includes several fault states, each rule in the rule database includes several indicators, and the number of fault states of the fault is equal to the number of indicators of the rule; the rule base data set is derived from the real record when the server fails, ensuring the authenticity and reliability of the data source;

[0137] S12. A rule in the positioning rule database is selected, and the deviation rate between each fault state data of each fault and the corresponding indicator data of the positioning rule is calculated in turn. If each deviation rate is less than the set ratio, it is determined that the corresponding fault matches the positioning rule successfully, and the matching situation of each fault and the positioning rule is counted to obtain the fault matching rate of the positioning rule, and the next rule in the rule database is located in turn until the fault matching rate of all rules in the positioning rule database is calculated; the specific steps are as follows:

[0138] Assume there are n rules in total, each rule has m indicators, and the data structure of rule R is as follows:

[0139]

[0140] Assume there are k faults in total, and the number of states of each fault is consistent with the number of indicators of the rule, which is m. The data structure of fault F is as follows:

[0141]

[0142] S121. A rule in the positioning rule database;

[0143] S122. Locate a fault in the fault status database;

[0144] S123. Calculate the absolute value of the difference between each fault state data of the fault location and the corresponding indicator data of the location rule, and calculate the deviation rate of each absolute value and the corresponding indicator data;

[0145] S124. Determine whether each deviation rate is less than the set ratio of 10%;

[0146] If so, the positioning fault matches the positioning rule successfully, the matching rate of the positioning rule is increased by one, and the process proceeds to step S125; the initial value of the matching rate of each rule is zero, and if the absolute value of the corresponding difference between the fault status data and the rule index data does not exceed 10% of the rule index number, it is considered that the match is successful, and the matching rate of the rule is increased by one;

[0147] If not, the location fault fails to match the location rule;

[0148] S125. Determine whether all faults in the fault status database have been located;

[0149] If yes, count the matching situation between the positioning rule and each fault, obtain the matching rate of the positioning rule, and go to step S126;

[0150] If not, locate the next fault and return to step S123;

[0151] S126. Determine whether all the centralized rules in the positioning rule database have been positioned;

[0152] If so, the matching rates of each rule in the rule database set form a fault matching rate sequence;

[0153] If not, locate the next rule and return to step S122;

[0154] For example, the indicator data of the first rule is {R 11 …R 1m}, the status data of the first fault is: {F 11 …F 1m}, calculate the absolute value of the corresponding difference between the fault state and the rule index, and get m absolute values ​​of the difference {|R 11 -F 11 |…|R 1m -F 1m |}, calculate the absolute value of the difference as a percentage of the corresponding indicator data of the rule If each proportion is less than 10%, the error is considered to be within a reasonable range, and the matching rate of the first rule is increased by one; if there is a proportion greater than 10%, the fault fails to match the first rule. In this way, all faults are traversed for the first rule to obtain the fault matching rate of the first rule. Similarly, the fault matching rate of n rules in the rule base can be obtained, which is recorded as {C1…C n}.

[0155] S2. Count the faults that each rule successfully matches, and count the number of faults corresponding to each fault level, and then remove the number of faults with low fault levels by ignoring the minimum value algorithm between partitions. Then, according to the preset weight of each fault level and the number of faults corresponding to the fault level, calculate the severity of the fault matched by each rule, and obtain the fault severity sequence of each rule in the rule database set; the specific steps are as follows:

[0156] S21. Obtaining the pre-divided fault levels, the preset weights of each fault level, and the fault levels using the partition interval ignoring minimum value algorithm;

[0157] For example, there are 10 fault severity levels. The larger the number, the more serious the fault. First, divide these 10 levels into three intervals. Levels 1 to 3 are considered minor faults, levels 4 to 7 are considered moderate faults, and levels 8 to 10 are considered severe faults. They are assigned weights of 1, 2, and 3 respectively. For n rules, count the faults that each rule successfully matches, and record the number of faults of each level as a ij , {i=1,…,n;j=1,…,10}, the total number of faults matched by the rule is recorded as a_sum i , {i=1,…,n}.

[0158] S22. A rule in the positioning rule database;

[0159] S23. Count the faults and corresponding fault levels that successfully match the positioning rules, and determine the faults and number within each fault classification;

[0160] S24. Determine whether there is a corresponding fault within the fault level using the partition interval ignoring minimum value algorithm;

[0161] If yes, the fault level of the algorithm using the minimum value between partitions is set to a low fault level, and the process goes to step S25;

[0162] If not, proceed to step S26;

[0163] S25. Determine whether the ratio of the number of faults within the low-level fault level to the total number of faults in the positioning rule is less than the ratio threshold;

[0164] If yes, set the number of faults in the low-level fault level to zero and proceed to step S26;

[0165] If not, proceed to step S26;

[0166] S26. Multiply the fault level value by the corresponding preset weight as a new weight, and use the new weight to weight the sum of the number of faults within the corresponding fault level to obtain the fault severity of the positioning rule;

[0167] S27. Determine whether all the centralized rules in the rule database have been located;

[0168] If so, the fault severity of each rule in the rule database is formed into a fault severity sequence;

[0169] If not, locate the next rule and return to step S23;

[0170] For the minor fault interval and the moderate fault interval, the algorithm of ignoring the minimum value of the interval is used to process the minimum number of faults in the two intervals. If the minimum number of faults in the minor interval accounts for less than 10% of the total number of faults or the minimum number of faults in the moderate interval accounts for less than 5% of the total number of faults, the number of faults corresponding to the minimum value of the interval is ignored, that is, the corresponding number of faults is set to zero;

[0171] Multiply the level value by the weight as the new weight, and calculate the severity of the fault matching the first rule in a weighted manner:

[0172]

[0173] Similarly, the fault severity of n rule matching in the rule base is calculated respectively, denoted as {S1…S n};

[0174] S3. Calculate the correlation between any two rules in the rule database set by using the GRA algorithm, and then calculate the correlation between each rule and the rule database set, and obtain the correlation sequence of each rule in the rule database set, and then perform a redirection process on the correlation sequence to obtain a redirected correlation sequence; the specific steps are as follows:

[0175] S31. A rule in the positioning rule database;

[0176] S32. Select a rule from the remaining rules in the rule database set;

[0177] S33. Obtain the indicator data of the positioning rules and the selection rules, and obtain the positioning rule indicator sequence and the selection rule indicator sequence;

[0178] S34. Selecting a maximum value from the positioning rule indicator sequence, and using the maximum value as a reference to standardize the positioning rule indicator sequence, to obtain a standardized positioning rule indicator sequence;

[0179] S35. Selecting a maximum value from the selection rule indicator sequence, and normalizing the selection rule indicator sequence with the maximum value of the sequence as a reference to obtain a normalized selection rule indicator sequence;

[0180] S36. Set the resolution coefficient, and calculate the correlation between the positioning rule and a certain indicator in the selection rule through the GRA algorithm, and then take the average of all indicators to obtain the correlation between the positioning rule and the selection rule;

[0181] S37. Determine whether the remaining rules of the rule database set have been selected;

[0182] If so, take the average of the correlation between the positioning rule and each selection rule to obtain the correlation between the positioning rule and the rule database set;

[0183] If not, select the next rule from the remaining rules in the rule database set and return to step S33;

[0184] S38. Determine whether all the centralized rules in the positioning rule database have been positioned;

[0185] If so, the relevance of each rule in the rule database set to the rule database set forms a relevance sequence;

[0186] If not, locate the next rule and return to step S32;

[0187] S39. Take the maximum value of the association degree sequence, subtract the maximum value from each element in the association degree sequence and take the absolute value to generate a new sequence, and set the new sequence as the steering association degree sequence;

[0188] Take the m index values ​​of the i-th rule and the j-th rule, R i = {R i1 …R im}, R j = {R j1 …R jm First, the maximum value of the sequence is used as a reference for standardization, and the rule after processing is recorded as r i and r j ,in

[0189]

[0190]

[0191] The resolution coefficient ρ in the GRA method is between 0 and 1, usually 0.5, and ρ=0.5 in the present invention;

[0192] The correlation between the yth indicator in the i-th rule and the j-th rule is denoted as D ijx , calculated as follows:

[0193]

[0194] Calculate the mean of m indicators to get the correlation between the i-th rule and the j-th rule There are n rules in the rule base. The correlation between the i-th rule and other rules is calculated to be n-1. The average value is taken to get the correlation between the i-th rule and the rule base. According to the above analysis method, the correlation degree of n rules is calculated as {D1…D n};

[0195] For the purpose of screening accurate rules with low similarity, the larger the values ​​of the two factors, the better, the rule matching rate and the matching fault severity, while the lower the similarity, the better. In order to facilitate the subsequent combination analysis, the similarity factor is first processed. Each value of the similarity sequence is subtracted from the maximum value and the absolute value is taken. This processing not only ensures that the gap between the similarity sequence values ​​remains unchanged, but also unifies the change trend with the other two factors. The processed similarity sequence is called the steering similarity, recorded as {D1'…D n '}.

[0196] S4. Calculate the weight values ​​of the fault matching rate sequence, fault severity sequence and steering similarity sequence by using the inverse variance algorithm, and perform weighted summation on the fault matching rate sequence, fault severity sequence and steering similarity sequence to obtain a sequence of good and bad values, and then set the priority for each rule according to the good and bad values; the specific steps are as follows:

[0197] S41. Calculate the variances of the fault matching rate sequence, the fault severity sequence and the steering correlation sequence respectively, and set them as the first variance, the second variance and the third variance respectively;

[0198] S42. Using the inverse of the first variance as the first weight, the inverse of the second variance as the second weight, and the inverse of the third variance as the third weight;

[0199] S43. weighting the calculated fault matching rate sequence with the first weight, weighting the fault severity sequence with the second weight, weighting the steering correlation sequence with the third weight, and then summing them to obtain a sequence of merit values;

[0200] S44. Sort the sequence of merit values, set high priority for rules with high merit values, and provide them to operation and maintenance personnel for server log analysis and diagnosis;

[0201] The weight values ​​of the three factor sequences of rule matching rate, matching fault severity and steering similarity are calculated by the inverse variance method, and the weighted sum is finally used to obtain the rule merit value {V1…V n};

[0202] First calculate the variance of the three sequences separately:

[0203]

[0204]

[0205]

[0206] The inverse of the variance of the three factor values ​​is used as the weight, that is,

[0207]

[0208]

[0209]

[0210] The merits of each rule in the rule database are:

[0211] V i =W C ×C i +W S ×S i +W D' ×D' i , {i=1,…,n}.

[0212] Although the present invention has been described in detail by referring to the accompanying drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, a person of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions shall be within the scope of the present invention. Any person of ordinary skill in the art may easily think of changes or substitutions within the technical scope disclosed by the present invention, and these shall be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A server fault diagnosis rule screening method based on GRA, characterized in that: The steps include: S1. Traverse each fault in the server fault record and each rule in the rule database, calculate the fault matching rate of each rule, and obtain the fault matching rate sequence of each rule in the rule database; S2. Count the faults that each rule successfully matches, and count the number of faults corresponding to each fault level, and then remove the number of faults with low fault levels by ignoring the minimum value algorithm between partitions. Then, according to the preset weight of each fault level and the number of faults corresponding to the fault level, calculate the severity of the fault matched by each rule, and obtain the fault severity sequence of each rule in the rule database set; S3. Calculate the correlation between any two rules in the rule database set by the GRA algorithm, and then calculate the correlation between each rule and the rule database set, and obtain the correlation sequence of each rule in the rule database set, and then perform a redirection process on the correlation sequence to obtain a redirection correlation sequence; S4. Calculate the weight values ​​of the fault matching rate sequence, the fault severity sequence and the steering correlation sequence by using the inverse variance algorithm, and perform weighted summation on the fault matching rate sequence, the fault severity sequence and the steering correlation sequence to obtain a sequence of good and bad values, and then set the priority for each rule according to the good and bad values; The specific steps of step S3 are as follows: S31. A rule in the positioning rule database; S32. Select a rule from the remaining rules in the rule database set; S33. Obtain the indicator data of the positioning rules and the selection rules, and obtain the positioning rule indicator sequence and the selection rule indicator sequence; S34. Selecting a maximum value from the positioning rule indicator sequence, and using the maximum value as a reference to standardize the positioning rule indicator sequence, to obtain a standardized positioning rule indicator sequence; S35. Selecting a maximum value from the selection rule indicator sequence, and normalizing the selection rule indicator sequence with the maximum value of the sequence as a reference to obtain a normalized selection rule indicator sequence; S36. Set the resolution coefficient, and calculate the correlation between the positioning rule and a certain indicator in the selection rule by using the GRA algorithm, and then take the average of all indicators to obtain the correlation between the positioning rule and the selection rule; S37. Determine whether the remaining rules of the rule database set have been selected; If so, take the average of the correlation between the positioning rule and each selection rule to obtain the correlation between the positioning rule and the rule database set; If not, select the next rule from the remaining rules in the rule database set and return to step S33; S38. Determine whether all the centralized rules in the positioning rule database have been positioned; If so, the relevance of each rule in the rule database set to the rule database set forms a relevance sequence; If not, locate the next rule and return to step S32; S39. Take the maximum value of the association degree sequence, subtract the maximum value from each element in the association degree sequence, and then take the absolute value to generate a new sequence, and the new sequence is set as the steering association degree sequence.

2. The server fault diagnosis rule screening method based on GRA as claimed in claim 1, characterized in that: The specific steps of step S1 are as follows: S11. Obtaining a fault status database of server fault records and a rule base data set for server fault diagnosis; each fault in the fault status database includes several fault states, each rule in the rule database includes several indicators, and the number of fault states of the fault is equal to the number of indicators of the rule; S12. Locate a rule in the rule database set, and calculate the deviation rate between each fault status data of each fault and the corresponding indicator data of the location rule in turn. If each deviation rate is less than the set ratio, it is determined that the corresponding fault matches the location rule successfully, and the matching situation of each fault and the location rule is counted to obtain the fault matching rate of the location rule, and locate the next rule in the rule database set in turn until the fault matching rate of all rules in the location rule database set is calculated.

3. The server fault diagnosis rule screening method based on GRA as claimed in claim 2, characterized in that: Metrics for the rules include failure level severity, underlying data read error rate, remapped sector count, read channel margin, spin-up retry count, and head calibration retry count.

4. The server fault diagnosis rule screening method based on GRA as claimed in claim 2, characterized in that: The specific steps of step S12 are as follows: S121. A rule in the positioning rule database; S122. Locate a fault in the fault status database; S123. Calculate the absolute value of the difference between each fault state data of the fault location and the corresponding indicator data of the location rule, and calculate the deviation rate of each absolute value and the corresponding indicator data; S124. Determine whether each deviation rate is less than a set ratio; If yes, the positioning fault matches the positioning rule successfully, the matching rate of the positioning rule is increased by one, and the process proceeds to step S125; If not, the location fault fails to match the location rule; S125. Determine whether all faults in the fault status database have been located; If yes, count the matching situation between the positioning rule and each fault, obtain the matching rate of the positioning rule, and go to step S126; If not, locate the next fault and return to step S123; S126. Determine whether all the centralized rules in the positioning rule database have been positioned; If so, the matching rates of each rule in the rule database set form a fault matching rate sequence; If not, locate the next rule and return to step S122.

5. The server fault diagnosis rule screening method based on GRA as claimed in claim 1, characterized in that: The specific steps of step S2 are as follows: S21. Obtaining the pre-divided fault levels, the preset weights of each fault level, and the fault levels using the partition interval ignoring minimum value algorithm; S22. A rule in the positioning rule database; S23. Count the faults and corresponding fault levels that successfully match the positioning rules, and determine the faults and number within each fault classification; S24. Determine whether there is a corresponding fault within the fault level using the partition interval ignoring minimum value algorithm; If yes, the fault level of the algorithm using the minimum value between partitions is set to a low fault level, and the process goes to step S25; If not, proceed to step S26; S25. Determine whether the ratio of the number of faults within the low-level fault level to the total number of faults in the positioning rule is less than the ratio threshold; If yes, set the number of faults in the low-level fault level to zero and proceed to step S26; If not, proceed to step S26; S26. Multiply the fault level value by the corresponding preset weight as a new weight, and use the new weight to weight the sum of the number of faults within the corresponding fault level to obtain the fault severity of the positioning rule; S27. Determine whether all the centralized rules in the rule database have been located; If so, the fault severity of each rule in the rule database is formed into a fault severity sequence; If not, locate the next rule and return to step S23.

6. The server fault diagnosis rule screening method based on GRA as claimed in claim 5, characterized in that: In step S21, the fault levels pre-classified are 10 levels, where level 1 fault level has the lowest severity and level 10 fault level has the highest severity; Level 1, 2, and 3 fault levels are set as minor faults, level 4, 5, 6, and 7 fault levels are set as moderate faults, and level 8, 9, and 10 fault levels are set as severe faults.

7. The server fault diagnosis rule screening method based on GRA as claimed in claim 6, characterized in that: In step 24, it is determined whether the ratio of the number of minor faults to the total number of faults in the positioning rule is less than the first ratio, or it is determined whether the ratio of the number of moderate faults to the total number of faults in the positioning rule is less than the second ratio, wherein the second ratio is less than the first ratio.

8. The server fault diagnosis rule screening method based on GRA as claimed in claim 1, characterized in that: In step S36, the value of the resolution system is greater than or equal to 0 and less than or equal to 1.

9. The server fault diagnosis rule screening method based on GRA as claimed in claim 1, characterized in that: The specific steps of step S4 are as follows: S41. Calculate the variances of the fault matching rate sequence, the fault severity sequence and the steering correlation sequence respectively, and set them as the first variance, the second variance and the third variance respectively; S42. Using the inverse of the first variance as the first weight, the inverse of the second variance as the second weight, and the inverse of the third variance as the third weight; S43. weighting the calculated fault matching rate sequence with the first weight, weighting the fault severity sequence with the second weight, weighting the steering correlation sequence with the third weight, and then summing them to obtain a sequence of merit values; S44. Sort the sequence of good and bad values, set high priority for the rules with high good and bad values, and provide them to operation and maintenance personnel for server log analysis and diagnosis.

Images