Performance tuning method, device and equipment for disk array card

Abstract

The invention discloses a performance tuning method, device and equipment for a disk array card, and relates to the technical field of performance test.The method comprises the steps that model training data are obtained; according to the model training data, training the multiple linear regression model to obtain a trained target model; wherein the multiple linear regression model is used for reflecting a corresponding relationship between the system performance data of each system performance type and the performance dotting data of the corresponding dotting performance type, and the corresponding relationship comprises a performance index coefficient of the dotting performance type corresponding to each system performance type; generating performance tuning guidance information according to the performance index coefficient in the target model; according to the method, the index criticality can be evaluated by utilizing the coefficient trained by the multiple linear regression model, so that a worker is assisted in performing performance analysis and optimization, the efficiency and stability of RAID card performance optimization are improved, and the workload of related personnel is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of performance testing, and particularly to a method, device and equipment for performance tuning of a disk array card. Background Art

[0002] With the progress of integrated circuit technology, SOC (System on a Chip) has become the core of modern electronic devices. To ensure the efficient operation of SOC, monitoring its performance has become an essential part. The RAID (Redundant Array of Independent Disks) card is an important application example of SOC and plays a key role in enterprise and data center environments. It improves the performance, reliability and data security of the storage system; for the RAID card, performance is of utmost importance. How to reasonably monitor the performance of the RIAD card, select monitoring metrics, and then perform performance tuning on the RAID card is a problem worthy of research.

[0003] In related technologies, the process of monitoring and tuning the performance of a RAID card is as Figure 1 shown: ① The tester issues an fio (a test tool) command on the host side according to the test business scenario; ② The tester screens out the subset of metrics to be monitored from the full set of monitoring metrics and sends the subset of metrics to the RAID card chip (RIAD chip). The RAID card chip performs performance logging in each sub-module (such as Figure 1 Module1, 2 and 3 in it), and finally outputs the performance logging data; ③ After obtaining the performance logging data from the tester, the domain expert analyzes the key performance indicators, gives the direction of key performance tuning, and gives the RAID card chip configuration adjustment plan; ④ The tester adjusts the RAID card chip configuration according to the suggestions given by the domain expert and conducts a new round of tests, repeating steps ① to ③ until the performance meets the expectations.

[0004] However, due to the large number of performance metrics that need to be monitored in complex RAID card scenarios, a large amount of performance logging data will make performance analysis and tuning extremely difficult, requiring a large amount of time and labor costs; moreover, the work threshold for manual performance analysis and tuning is high, and the manual analysis and tuning process will bring many unstable factors to performance analysis. Therefore, how to improve the efficiency and stability of RAID card performance tuning and reduce the workload of relevant personnel is an urgent problem to be solved today. Summary of the Invention

[0005] The object of the present invention is to provide a method, device and equipment for performance tuning of a disk array card to improve the efficiency and stability of RAID card performance tuning and reduce the workload of relevant personnel.

[0006] In order to solve the above technical problems, the present invention provides a performance tuning method for a disk array card, comprising:

[0007] Acquire model training data; wherein the model training data includes system performance data and performance management data of the disk array card;

[0008] According to the model training data, the multivariate linear regression model is trained to obtain a target model after the training is completed; wherein the multivariate linear regression model is used to reflect the corresponding relationship between the system performance data of each system performance type and the performance dotting data of the corresponding dotting performance type, and the corresponding relationship includes the performance index coefficient of the dotting performance type corresponding to each of the system performance types;

[0009] Generate performance tuning guidance information according to the performance indicator coefficients in the target model; wherein the performance tuning guidance information includes at least one of the dotting performance type identifier corresponding to the target performance indicator coefficient, the performance dotting data and the system performance type identifier.

[0010] On the other hand, the multivariate linear regression model is trained according to the model training data to obtain a target model after training, including:

[0011] Performing data preprocessing on the performance scoring data to obtain preprocessing results; wherein the data preprocessing includes normalization processing;

[0012] The multivariate linear regression model is trained according to the system performance data and the preprocessing result to obtain a trained target model.

[0013] On the other hand, performing data preprocessing on the performance rating data to obtain preprocessing results includes:

[0014] Normalizing the performance scoring data to obtain a normalized result;

[0015] The normalized result is amplified to obtain the preprocessing result.

[0016] On the other hand, the normalizing the performance rating data to obtain a normalized result includes:

[0017] According to the preset maximum value and the preset minimum value corresponding to the current dotting performance type, the current performance dotting data is normalized to obtain a normalized result corresponding to the current performance dotting data; wherein, the current dotting performance type is any of the dotting performance types, and the current performance dotting data is any performance dotting data corresponding to the current dotting performance type.

[0018] On the other hand, the amplification process of the normalization result to obtain the preprocessing result includes:

[0019] By , calculate the preprocessing result corresponding to the current performance measurement data; where x cur is the normalization result corresponding to the current performance measurement data, x n_max and x n_min are the normalization results corresponding to the maximum value and the minimum value in the target performance measurement data respectively. The target performance measurement data is the performance measurement data within a preset time period corresponding to the current measurement performance type, and the target performance measurement data includes the current performance measurement data; β is a preset amplification factor.

[0020] On the other hand, the multiple linear regression model is specifically ;

[0021] where y j is the system performance data at the current moment of the jth system performance type, k ji is the performance index coefficient of the ith measurement performance type corresponding to the jth system performance type, x i is the preprocessing result corresponding to the performance measurement data at the current moment of the ith measurement performance type, m is the number of all the system performance types, and n is the number of all the measurement performance types.

[0022] On the other hand, the performance tuning guidance information further includes the coefficient type corresponding to the target performance index coefficient, and the coefficient type includes positive coefficient, negative coefficient, and zero coefficient;

[0023] The generation of the performance tuning guidance information according to the performance index coefficients in the target model includes:

[0024] Obtain a positive coefficient set, a negative coefficient set, and a zero coefficient set according to the performance index coefficients in the target model;

[0025] Determine the target performance index coefficient according to the positive coefficient set, the negative coefficient set, and the zero coefficient set;

[0026] Generate the performance tuning guidance information according to the target performance index coefficient.

[0027] On the other hand, the determination of the target performance index coefficient according to the positive coefficient set, the negative coefficient set, and the zero coefficient set includes:

[0028] Determine the target performance index coefficient according to the system performance impact weights corresponding to each performance index coefficient in the positive coefficient set and the negative coefficient set; wherein, when the current performance index coefficient belongs to the positive coefficient set, the system performance impact weight corresponding to the current performance index coefficient is the quotient of the current performance index coefficient and the sum of the positive coefficients of the target system performance type; when the current performance index coefficient belongs to the negative coefficient set, the system performance impact weight corresponding to the current performance index coefficient is the quotient of the current performance index coefficient and the sum of the negative coefficients of the target system performance type; the target system performance type is the system performance type corresponding to the current performance index coefficient, the sum of positive coefficients is the sum of all performance index coefficients corresponding to the target system performance type in the positive coefficient set, and the sum of negative coefficients is the sum of all performance index coefficients corresponding to the target system performance type in the negative coefficient set.

[0029] The present invention also provides a performance tuning device for a disk array card, including:

[0030] An acquisition module, configured to acquire model training data; wherein, the model training data includes system performance data and performance marking data of the disk array card;

[0031] A training module, configured to train a multiple linear regression model according to the model training data to obtain a target model after training; wherein, the multiple linear regression model is used to reflect the corresponding relationship between the system performance data of each system performance type and the performance marking data of the corresponding marking performance type, and the corresponding relationship includes the performance index coefficients of the marking performance types corresponding to each system performance type;

[0032] A generation module, configured to generate performance tuning guidance information according to the performance index coefficients in the target model; wherein, the performance tuning guidance information includes at least one of the marking performance type identifier corresponding to the target performance index coefficient, the performance marking data, and the system performance type identifier.

[0033] In addition, the present invention also provides a performance tuning device for a disk array card, including:

[0034] A memory, configured to store a computer program;

[0035] A processor, configured to implement the steps of the performance tuning method for the disk array card as described above when executing the computer program.

[0036] A method for performance tuning of a disk array card provided by the present invention includes: obtaining model training data; wherein the model training data includes system performance data and performance marking data of the disk array card; training a multiple linear regression model according to the model training data to obtain a target model after training; wherein the multiple linear regression model is used to reflect the corresponding relationship between the system performance data of each system performance type and the performance marking data of the corresponding marking performance type, and the corresponding relationship includes the performance index coefficients of the marking performance types corresponding to each system performance type; generating performance tuning guidance information according to the performance index coefficients in the target model; wherein the performance tuning guidance information includes at least one of the marking performance type identifier corresponding to the target performance index coefficient, the performance marking data, and the system performance type identifier.

[0037] It can be seen that by generating performance tuning guidance information according to the performance index coefficients in the target model, the present invention can use the coefficients trained by the multiple linear regression model to evaluate the key performance indicators, thereby assisting the staff in performance analysis and tuning, improving the efficiency and stability of RAID card performance tuning, and reducing the workload of relevant personnel. In addition, the present invention also provides a performance tuning device and equipment for a disk array card, which also have the above beneficial effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0039] Figure 1 It is a schematic flow chart of performance monitoring and tuning of a RAID card chip in the related art;

[0040] Figure 2 It is a flow chart of a method for performance tuning of a disk array card provided by an embodiment of the present invention;

[0041] Figure 3 It is a schematic diagram of a model training process provided by an embodiment of the present invention;

[0042] Figure 4 It is a range display diagram of data normalization provided by an embodiment of the present invention;

[0043] Figure 5 It is a schematic diagram of a process for screening performance index coefficients provided by an embodiment of the present invention;

[0044] Figure 6A flowchart showing the process of performance monitoring and optimization of a RAID card chip provided by an embodiment of the present invention;

[0045] Figure 7 A structural block diagram of a performance optimization device for a disk array card provided by an embodiment of the present invention;

[0046] Figure 8 A structural diagram of a performance optimization device for a disk array card provided by an embodiment of the present invention. Detailed implementation manners

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0048] Please refer to Figure 2 , Figure 2 A flowchart of a performance optimization method for a disk array card provided by an embodiment of the present invention. The method may include:

[0049] Step 101: Obtain model training data; wherein, the model training data includes system performance data and performance marking data of the disk array card.

[0050] It can be understood that the model training data in this embodiment may be data for training a multiple linear regression model; the multiple linear regression model in this embodiment is used to reflect the corresponding relationships between the system performance data of each system performance type and the performance marking data of the corresponding marking performance type, that is, abstract the relationships between the system performance data of each system performance type and the performance marking data of the corresponding marking performance type into a multiple linear regression model to train the performance index coefficients of the marking performance types corresponding to each system performance type.

[0051] Correspondingly, the model training data in this embodiment may include performance marking data of each marking performance type (i.e., monitoring indicators) of the RAID (Redundant Array of Independent Disks) card. For example, when performing performance monitoring on the RAID card, the performance marking data obtained and output by the RAID card chip according to each monitoring indicator in the issued indicator subset in the corresponding internal sub-module; it may also include system performance data of each system performance type of the RAID card, that is, data of each system performance (such as throughput, latency, and bandwidth, etc.) presented externally during performance marking.

[0052] Correspondingly, for the specific number and type of the dotting performance type and the system performance type in this embodiment, it can be set by the designer according to the practical scenario and user requirements. For example, the system performance type can include any one or more of IOPS (Input / Output Per Second), throughput, latency, and bandwidth. The dotting performance type can include all monitoring metrics in the metric subset sent to the RAID card chip. This embodiment does not impose any restrictions on this.

[0053] Similarly, for the specific method of obtaining the model training data in this step, it can be set by the designer according to the practical scenario and user requirements. For example, it can receive the model training data sent by the RAID card. For example, the host device can send the metric subset to the RAID card chip, so that the RAID card chip performs performance dotting based on each monitoring metric in the sent metric subset in the corresponding internal sub-module to obtain the performance dotting data and the corresponding system performance data, and then receive the performance dotting data and the corresponding system performance data returned by the RAID card chip. In this step, it can also directly receive the model training data transmitted by the user, or read the pre-stored model training data. This embodiment does not impose any restrictions on this.

[0054] Step 102: Train the multiple linear regression model according to the model training data to obtain the trained target model. Among them, the multiple linear regression model is used to reflect the corresponding relationship between the system performance data of each system performance type and the performance dotting data of the corresponding dotting performance type. The corresponding relationship includes the performance metric coefficients of the dotting performance type corresponding to each system performance type.

[0055] It should be noted that in this embodiment, the performance metric coefficients of the dotting performance type corresponding to each system performance type in the multiple linear regression model are used to reflect the influence degree of the performance dotting data on the system performance data. The purpose of training the multiple linear regression model in this embodiment is to obtain the performance metric coefficients in the trained multiple linear regression model (i.e., the target model), rather than using the target model to predict data.

[0056] Correspondingly, for the specific process of training the multiple linear regression model according to the model training data in this embodiment to obtain the trained target model, it can be set by the designer according to the practical scenario and user requirements. For example, it can directly use the system performance data and performance dotting data in the model training data to train the multiple linear regression model to obtain the trained target model. Due to problems such as unit confusion and different value ranges in the performance dotting data, the performance metric coefficients of the trained multiple linear regression model will be affected by the values of the performance dotting data itself. Therefore, Figure 3As shown, in this step, data preprocessing (such as normalization) can be performed on the performance measurement data (performance monitoring data) to eliminate the influence of dimensions on the data, so as to use the preprocessing results obtained from the data preprocessing to train the model and obtain the target model that best fits the real model.

[0057] That is to say, in this step, data preprocessing can be performed on the performance measurement data to obtain preprocessing results; among them, data preprocessing includes normalization; according to the system performance data and the preprocessing results, the multiple linear regression model is trained to obtain the target model after training.

[0058] Correspondingly, for the specific method of performing data preprocessing on the performance measurement data and obtaining preprocessing results above, it can be set by the designer himself. For example, only normalization processing can be performed on the performance measurement data, that is, the above data preprocessing can only include normalization processing. For example, in some embodiments, the above data preprocessing process can perform data normalization on the current performance measurement data according to the preset maximum value and preset minimum value corresponding to the current measurement performance type, to obtain the preprocessing result corresponding to the current performance measurement data; among them, the current measurement performance type is any measurement performance type, and the current performance measurement data is any performance measurement data corresponding to the current measurement performance type.

[0059] In other embodiments, the above process can also perform data normalization on the current performance measurement data according to the maximum value and minimum value in the performance measurement data within the preset time period corresponding to the current measurement performance type, to obtain the preprocessing result corresponding to the current performance measurement data.

[0060] Among them, the above data preprocessing process can not only perform normalization processing on the performance measurement data, but also perform amplification processing on the normalization result to avoid the problem that the data is too small and not friendly enough to the calculation, thereby accelerating the training speed, that is, the above data preprocessing can include normalization processing and amplification processing. For example, in some embodiments, the above data preprocessing process can include: performing normalization processing on the performance measurement data to obtain a normalization result; performing amplification processing on the normalization result to obtain a preprocessing result.

[0061] For example, the process of performing normalization processing on the performance measurement data to obtain a normalization result can include: performing data normalization on the current performance measurement data according to the preset maximum value and preset minimum value corresponding to the current measurement performance type, to obtain the normalization result corresponding to the current performance measurement data; among them, the current measurement performance type is any measurement performance type, and the current performance measurement data is any performance measurement data corresponding to the current measurement performance type.

[0062] Correspondingly, the preset maximum and minimum values corresponding to different RBI performance types may be different. For example, for a period of time (such as per second), the number of IOs processed (such as Figure 4 Indicator 1 in the table) and the maximum data bandwidth of a submodule (such as Figure 4 Indicator 2 or indicator 3 in the above table are two completely different types of performance management. The fluctuation of the number of IOs processed will be relatively large, while the fluctuation of bandwidth changes will be relatively small. Therefore, Figure 4 As shown, the range between the preset minimum value and the preset maximum value corresponding to the number of IOs to be processed (such as Figure 4 The range between the preset minimum value and the preset maximum value corresponding to the maximum data bandwidth (such as 0-1000 in Figure 4 0-100 in .

[0063] The above process of amplifying the normalized result to obtain the preprocessing result may include: , calculate the preprocessing results corresponding to the current performance dot data; where x cur is the normalized result corresponding to the current performance dot data, x n_max and x n_min They are the normalized results corresponding to the maximum and minimum values in the target performance dotting data respectively. The target performance dotting data is the performance dotting data within a preset time period corresponding to the current dotting performance type. The target performance dotting data includes the current performance dotting data. β is the preset amplification factor.

[0064] It can be understood that the specific content of the multiple linear regression model in this embodiment can be set by the designer according to practical scenarios and user needs. For example, when the multiple linear regression model reflects the corresponding relationship between the system performance data of each system performance type and the performance dot data of all dot performance types, the multiple linear regression model can be specifically ; Among them, y j is the system performance data of the jth system performance type at the current moment, that is, y j It can represent the system performance index of the j-th system performance type, and is used to substitute the system performance data of the j-th system performance type at the current moment; k ji is the performance index coefficient of the i-th RBI performance type corresponding to the j-th system performance type; x i is the preprocessing result corresponding to the performance dotting data of the i-th dotting performance type at the current moment, that is, x i It can represent the RBI performance index of the i-th RBI performance type, and is used to substitute the performance RBI data of the i-th RBI performance type at the current moment or the corresponding preprocessing result; m is the number of all system performance types, and n is the number of all RBI performance types.

[0065] Among them, for the specific model training method of training the multiple linear regression model in this embodiment, it can be set in a manner similar to the training method of the multiple linear regression model in the related art. For example, the mean square error function is used as the loss function, and the multiple linear regression model is trained through supervised learning using the model training data to obtain the target model after training. That is, when the error reaches the allowable deviation range, it is determined that the training of the multiple linear regression model is completed, and the desired linear model is obtained.

[0066] Step 103: Generate performance tuning guidance information according to the performance index coefficients in the target model; among them, the performance tuning guidance information includes at least one of the dotting performance type identifier corresponding to the target performance index coefficient, the performance dotting data, and the system performance type identifier.

[0067] It can be understood that the target performance index coefficients in this embodiment can be partial performance index coefficients in the target model, that is, the performance index coefficients selected from the performance index coefficients in the target model for guiding subsequent performance analysis and tuning. That is to say, this step may include determining the target performance index coefficients from the performance index coefficients in the target model; generating performance tuning guidance information according to the target performance index coefficients.

[0068] Correspondingly, the performance tuning guidance information in this embodiment can be information corresponding to the target performance index coefficients for guiding subsequent performance analysis and tuning, to assist relevant personnel (such as testers or domain experts) in performing performance analysis and tuning, so that relevant personnel (such as testers or domain experts) can view the performance tuning guidance information, adjust the RAID card configuration, optimize the RAID card performance, thereby improving the efficiency and stability of RAID card performance tuning and reducing the workload of relevant personnel.

[0069] Among them, for the specific content of the performance tuning guidance information in this embodiment, it can be set by the designer according to the practical scenario and user requirements. For example, the performance tuning guidance information may include the dotting performance type identifier corresponding to the target performance index coefficient to indicate the key dotting performance type (i.e., the monitoring index); the performance tuning guidance information may also include the performance dotting data corresponding to the target performance index coefficient to display the data obtained from the actual performance monitoring of the key dotting performance type; the performance tuning guidance information may further include the system performance type identifier corresponding to the target performance index coefficient to indicate the system performance type that is greatly affected by the key dotting performance type. The performance tuning guidance information may also include other contents, such as the system performance impact weight and coefficient type (such as positive coefficient and negative coefficient) corresponding to the target performance index coefficient, and this embodiment does not make any restrictions on this.

[0070] Correspondingly, for the specific method of generating performance tuning guidance information based on the performance metric coefficients in the target model in this step, it can be set by the designer according to the practical scenario and user requirements. For example, in the process of determining the target performance metric coefficient from the performance metric coefficients in the target model as described above, the target performance metric coefficient can be directly determined based on the absolute values of the performance metric coefficients. For example, the performance metric coefficients with absolute values greater than the threshold are determined as the target performance metric coefficients.

[0071] In some other embodiments, the process of determining the target performance metric coefficient may include: obtaining a positive coefficient set, a negative coefficient set, and a zero coefficient set according to the performance metric coefficients in the target model; and determining the target performance metric coefficient according to the positive coefficient set, the negative coefficient set, and the zero coefficient set. That is to say, according to the signs of the performance metric coefficients, all the performance metric coefficients in the target model can be classified into a positive coefficient set (i.e., the set of performance metric coefficients greater than 0), a negative coefficient set (i.e., the set of performance metric coefficients less than 0), and a zero coefficient set (i.e., the set of performance metric coefficients equal to 0); then the target performance metric coefficient is screened from the positive coefficient set, the negative coefficient set, and the zero coefficient set to improve the accuracy of the determined target performance metric coefficient.

[0072] Correspondingly, for the specific method of determining the target performance metric coefficient according to the positive coefficient set, the negative coefficient set, and the zero coefficient set as described above, it can be set by the designer according to the practical scenario and user requirements. For example, for the system performance metrics, the coefficients in the positive coefficient set have positive gains, the coefficients in the negative coefficient set have negative losses, and the coefficients in the zero coefficient set have no impact; therefore, the target performance metric coefficient can be determined from the positive coefficient set and the negative coefficient set. For example, the performance metric coefficients with absolute values greater than the first absolute value threshold are screened from the positive coefficient set as the target performance metric coefficients; the performance metric coefficients with absolute values greater than the second absolute value threshold are screened from the negative coefficient set as the target performance metric coefficients.

[0073] Further, although the value of the performance index coefficient itself can measure the impact of the corresponding dotting performance index on the system performance, in order to make the measurement more quantifiable, in this embodiment, the system performance impact weight obtained by normalization processing can be used to reflect the percentage of the impact on the system performance, thereby improving the accuracy of the determined target performance index coefficient. For example, for the process of determining the target performance index coefficient according to the positive coefficient set, negative coefficient set, and zero coefficient set as described above, it may include: determining the target performance index coefficient according to the system performance impact weights corresponding to the respective performance index coefficients in the positive coefficient set and the negative coefficient set; where when the current performance index coefficient belongs to the positive coefficient set, the system performance impact weight corresponding to the current performance index coefficient is the quotient of the current performance index coefficient and the sum of the positive coefficients of the target system performance type; when the current performance index coefficient belongs to the negative coefficient set, the system performance impact weight corresponding to the current performance index coefficient is the quotient of the current performance index coefficient and the sum of the negative coefficients of the target system performance type; the target system performance type is the system performance type corresponding to the current performance index coefficient, the sum of the positive coefficients is the sum of all the performance index coefficients corresponding to the target system performance type in the positive coefficient set, and the sum of the negative coefficients is the sum of all the performance index coefficients corresponding to the target system performance type in the negative coefficient set.

[0074] That is to say, the performance index coefficient k ji The corresponding system performance impact weight α ji can represent the impact weight of the i-th dotting performance type on the j-th system performance type, and the value range is from 0% to 100%; the performance index coefficient k ji When it belongs to the positive coefficient set, ; the performance index coefficient k ji When it belongs to the negative coefficient set, . can represent the sum of all the performance index coefficients corresponding to the j-th system performance type in the positive coefficient set, that is, the sum of the system performance impact weights of all the performance index coefficients corresponding to the j-th system performance type in the positive coefficient set is 100%; can represent the sum of all the performance index coefficients corresponding to the j-th system performance type in the negative coefficient set, that is, the sum of the system performance impact weights of all the performance index coefficients corresponding to the j-th system performance type in the negative coefficient set is 100%.

[0075] Correspondingly, in the process of determining the target performance index coefficients according to the system performance impact weights corresponding to the respective performance index coefficients in the positive coefficient set and the negative coefficient set, performance index coefficients with system performance impact weights greater than the first weight threshold can be screened out from the positive coefficient set as the target performance index coefficients; performance index coefficients with system performance impact weights greater than the second weight threshold can be screened out from the negative coefficient set as the target performance index coefficients. Alternatively, the first quantity of performance index coefficients with relatively large system performance impact weights can be screened out from the positive coefficient set as the target performance index coefficients; the second quantity of performance index coefficients with relatively large system performance impact weights can be screened out from the negative coefficient set as the target performance index coefficients.

[0076] For example, the process of determining the target performance index coefficients can be as Figure 5 shown, including coefficient classification, sorting, and key coefficient screening (feature selection); in the coefficient classification process, all performance index coefficients (coefficient set, or coefficient matrix) in the target model can be divided into three categories: positive system, negative coefficient, and zero coefficient according to the signs of the performance index coefficients; in the sorting process, the positive coefficients and negative coefficients can be sorted in descending order according to their absolute values respectively, and the zero coefficients do not need to be sorted; in the key coefficient screening process, the arranged positive coefficients and negative coefficients can be converted into the corresponding system performance impact weights; according to the preset weight threshold or screening quantity, the target performance index coefficients can be screened out from the positive coefficients and negative coefficients; alternatively, the tester can select the target performance index coefficients according to the system performance impact weights of the arranged positive coefficients and negative coefficients shown.

[0077] That is to say, in some embodiments, the process of determining the target performance index coefficients according to the system performance impact weights corresponding to the respective performance index coefficients in the positive coefficient set and the negative coefficient set can include presenting a first sorted queue and a second sorted queue; wherein, the first sorted queue includes the system performance impact weights corresponding to all performance index coefficients in the positive coefficient set arranged in descending order; the second sorted queue includes the system performance impact weights corresponding to all performance index coefficients in the negative coefficient set arranged in descending order; according to the obtained coefficient selection instruction, the performance index coefficients corresponding to the system performance impact weights selected by the coefficient selection instruction are determined as the target performance index coefficients.

[0078] Such as Figure 6As shown in the figure, the process of RAID card performance tuning may include: ① the tester sends the fio command on the host side according to the test business scenario; ② the tester collects the indicators to be monitored on the host side into the RAID card chip (RIAD chip), receives the performance monitoring performance data and system performance data returned by the RAID card chip, and obtains the model training data; ③ the performance data is preprocessed to obtain usable data, and the multivariate linear regression model is trained through supervised learning, so as to obtain the coefficient matrix of the trained target model, which is systematically classified and sorted and converted into the system performance impact weight; ④ according to the system performance impact weight and coefficient category, the key performance data and reference tuning direction can be selected; ⑤ the domain expert gives the configuration adjustment plan of the RAID card chip according to the analysis; ⑥ the tester adjusts the configuration according to the suggestions given by the domain expert and adjusts the configuration of the RAID card chip; ⑦ the tester conducts a new round of testing to check whether the performance reaches the expected range; if not, repeat steps ① to ⑥ until the performance reaches an expected value range.

[0079] In this embodiment, the embodiment of the present invention generates performance tuning guidance information according to the performance indicator coefficients in the target model, and can use the coefficients trained by the multivariate linear regression model to evaluate the criticality of the indicator, thereby assisting the staff in performance analysis and tuning, improving the efficiency and stability of RAID card performance tuning, and reducing the workload of related personnel.

[0080] Corresponding to the above method embodiment, the embodiment of the present invention further provides a performance tuning device for a disk array card. The performance tuning device for a disk array card described below and the performance tuning method for a disk array card described above can be referred to each other.

[0081] Please refer to Figure 7 , Figure 7 This is a structural block diagram of a performance tuning device for a disk array card provided by an embodiment of the present invention. The device may include:

[0082] The acquisition module 10 is used to acquire model training data; wherein the model training data includes system performance data and performance management data of the disk array card;

[0083] The training module 20 is used to train the multivariate linear regression model according to the model training data to obtain a target model after the training is completed; wherein the multivariate linear regression model is used to reflect the corresponding relationship between the system performance data of each system performance type and the performance dotting data of the corresponding dotting performance type, and the corresponding relationship includes the performance index coefficient of the dotting performance type corresponding to each system performance type;

[0084] A generation module 30, configured to generate performance tuning guidance information according to performance metric coefficients in a target model; wherein, the performance tuning guidance information includes at least one of a dotting performance type identifier corresponding to the target performance metric coefficient, performance dotting data, and a system performance type identifier.

[0085] On the other hand, the training module 20 may include:

[0086] A preprocessing sub-module, configured to perform data preprocessing on the performance dotting data to obtain a preprocessing result; wherein, the data preprocessing includes normalization processing;

[0087] A training sub-module, configured to train a multiple linear regression model according to the system performance data and the preprocessing result to obtain a target model after training is completed.

[0088] On the other hand, the preprocessing sub-module may include:

[0089] A normalization unit, configured to perform normalization processing on the performance dotting data to obtain a normalization result;

[0090] An amplification unit, configured to perform amplification processing on the normalization result to obtain a preprocessing result.

[0091] On the other hand, the normalization unit may be specifically configured to perform data normalization processing on the current performance dotting data according to a preset maximum value and a preset minimum value corresponding to the current dotting performance type to obtain a normalization result corresponding to the current performance dotting data; wherein, the current dotting performance type is any dotting performance type, and the current performance dotting data is any performance dotting data corresponding to the current dotting performance type.

[0092] On the other hand, the amplification unit may be specifically configured to calculate a preprocessing result corresponding to the current performance dotting data through , where x cur is the normalization result corresponding to the current performance dotting data, x n_max and x n_min are respectively the normalization results corresponding to the maximum value and the minimum value in the target performance dotting data, the target performance dotting data is the performance dotting data within a preset time period corresponding to the current dotting performance type, the target performance dotting data includes the current performance dotting data; β is a preset amplification coefficient.

[0093] On the other hand, the multiple linear regression model is specifically ;

[0094] wherein, y j is the system performance data at the current moment of the jth system performance type, k ji is the performance metric coefficient of the ith dotting performance type corresponding to the jth system performance type, xi is the preprocessing result corresponding to the performance dotting data of the i-th dotting performance type at the current moment, m is the number of all system performance types, and n is the number of all dotting performance types.

[0095] On the other hand, the performance tuning guidance information also includes a coefficient type corresponding to the target performance indicator coefficient, and the coefficient type includes a positive coefficient, a negative coefficient, and a zero coefficient;

[0096] The generation module 30 may include:

[0097] A classification submodule, used for obtaining a positive coefficient set, a negative coefficient set and a zero coefficient set according to the performance indicator coefficients in the target model;

[0098] A determination submodule, used to determine the target performance indicator coefficient according to the positive coefficient set, the negative coefficient set and the zero coefficient set;

[0099] The generation submodule is used to generate performance tuning guidance information according to the target performance indicator coefficient.

[0100] On the other hand, the determination submodule can be specifically used to determine the target performance indicator coefficient based on the system performance impact weights corresponding to each performance indicator coefficient in the positive coefficient set and the negative coefficient set; wherein, when the current performance indicator coefficient belongs to the positive coefficient set, the system performance impact weight corresponding to the current performance indicator coefficient is the quotient of the current performance indicator coefficient and the sum of the positive coefficients of the target system performance type; when the current performance indicator coefficient belongs to the negative coefficient set, the system performance impact weight corresponding to the current performance indicator coefficient is the quotient of the current performance indicator coefficient and the sum of the negative coefficients of the target system performance type; the target system performance type is the system performance type corresponding to the current performance indicator coefficient, the positive coefficient sum is the sum of all performance indicator coefficients corresponding to the target system performance type in the positive coefficient set, and the negative coefficient sum is the sum of all performance indicator coefficients corresponding to the target system performance type in the negative coefficient set.

[0101] In this embodiment, the embodiment of the present invention generates performance tuning guidance information according to the performance indicator coefficients in the target model through the generation module 30, and can use the coefficients trained by the multivariate linear regression model to evaluate the criticality of the indicators, thereby assisting the staff in performance analysis and tuning, improving the efficiency and stability of RAID card performance tuning, and reducing the workload of related personnel.

[0102] Corresponding to the above method embodiment, the embodiment of the present invention further provides a performance tuning device for a disk array card. The performance tuning device for a disk array card described below and the performance tuning method for a disk array card described above can refer to each other.

[0103] Please refer to Figure 8 , Figure 8The figure is a schematic structural diagram of a performance tuning device for a disk array card provided by an embodiment of the present invention. The device may include:

[0104] A memory D1 for storing a computer program;

[0105] A processor D2 for implementing the steps of the performance tuning method for the disk array card provided by the above method embodiment when executing the computer program.

[0106] Among them, the performance tuning device for the disk array card provided by this embodiment may specifically be a host device, such as a server host.

[0107] Corresponding to the above method embodiment, an embodiment of the present invention also provides a computer program product. The computer program product described below can be mutually corresponded and referred to the performance tuning method for a disk array card described above.

[0108] A computer program product includes computer programs / instructions. When the computer programs / instructions are executed by a processor, the steps of the performance tuning method for the disk array card provided by the above method embodiment are implemented.

[0109] Corresponding to the above method embodiment, an embodiment of the present invention also provides a computer-readable storage medium. The computer-readable storage medium described below can be mutually corresponded and referred to the performance tuning method for a disk array card described above.

[0110] A computer-readable storage medium has a computer program stored thereon. When the computer program is executed by a processor, the steps of the performance tuning method for the disk array card in the above method embodiment are implemented.

[0111] Specifically, the computer-readable storage medium may be various readable storage media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc that can store program codes.

[0112] The embodiments in the specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the devices, equipment, computer program products, and computer-readable storage media disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple. For the relevant parts, refer to the description in the method part.

[0113] The above has introduced in detail a method, device, and equipment for performance tuning of a disk array card provided by the present invention. Specific examples are used in this article to illustrate the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the protection scope of the present invention.

Claims

1. A method for performance tuning of a disk array card, characterized in that, Including: Obtain model training data; wherein, the model training data includes system performance data and performance marking data of a disk array card; Train a multiple linear regression model according to the model training data to obtain a target model after training; wherein, the multiple linear regression model is used to reflect the corresponding relationship between the system performance data of each system performance type and the performance marking data of the corresponding marking performance type, and the corresponding relationship includes the performance index coefficients of the marking performance type corresponding to each system performance type; Generate performance tuning guidance information according to the performance index coefficients in the target model; wherein, the performance tuning guidance information includes at least one of the marking performance type identifier corresponding to the target performance index coefficient, the performance marking data, and the system performance type identifier.

2. The performance tuning method of the disk array card according to claim 1, wherein The step of training the multiple linear regression model according to the model training data to obtain a target model after training includes: Perform data preprocessing on the performance marking data to obtain a preprocessing result; wherein, the data preprocessing includes normalization processing; Train the multiple linear regression model according to the system performance data and the preprocessing result to obtain a target model after training.

3. The performance tuning method of the disk array card according to claim 2, wherein The step of performing data preprocessing on the performance marking data to obtain a preprocessing result includes: Perform normalization processing on the performance marking data to obtain a normalization result; Perform amplification processing on the normalization result to obtain the preprocessing result.

4. The performance tuning method of the disk array card according to claim 3, wherein The step of performing normalization processing on the performance marking data to obtain a normalization result includes: Perform data normalization processing on the current performance marking data according to the preset maximum value and preset minimum value corresponding to the current marking performance type to obtain the normalization result corresponding to the current performance marking data; wherein, the current marking performance type is any one of the marking performance types, and the current performance marking data is any performance marking data corresponding to the current marking performance type.

5. The performance tuning method of the disk array card according to claim 4, characterized in that, The step of performing amplification processing on the normalization result to obtain the preprocessing result includes: Through , calculate the preprocessing result corresponding to the current performance data point; where x cur is the normalization result corresponding to the current performance data point, x n_max and x n_min are respectively the normalization results corresponding to the maximum value and the minimum value in the target performance data points. The target performance data points are the performance data points within a preset time period corresponding to the current data point performance type, and the target performance data points include the current performance data point; β is a preset amplification factor.

6. The performance tuning method of the disk array card according to claim 2, wherein The specific form of the multiple linear regression model is ; Among them, y j is the system performance data of the jth system performance type at the current moment, k ji is the performance index coefficient of the ith RBI performance type corresponding to the jth system performance type, x i is the preprocessing result corresponding to the performance dotting data of the i-th dotting performance type at the current moment, m is the number of all the system performance types, and n is the number of all the dotting performance types.

7. The performance tuning method of the disk array card according to any one of claims 1 to 6, characterized in that, The performance tuning guidance information further includes the coefficient type corresponding to the target performance index coefficient, and the coefficient type includes a positive coefficient, a negative coefficient, and a zero coefficient; The step of generating performance tuning guidance information according to the performance index coefficients in the target model includes: Obtain a positive coefficient set, a negative coefficient set, and a zero coefficient set according to the performance index coefficients in the target model; Determine the target performance index coefficient according to the positive coefficient set, the negative coefficient set, and the zero coefficient set; Generate the performance tuning guidance information according to the target performance index coefficient.

8. The performance tuning method of the disk array card according to claim 7, characterized in that, The step of determining the target performance index coefficient according to the positive coefficient set, the negative coefficient set, and the zero coefficient set includes: The target performance indicator coefficient is determined according to the system performance impact weight corresponding to each performance indicator coefficient in the positive coefficient set and the negative coefficient set; wherein, when the current performance indicator coefficient belongs to the positive coefficient set, the system performance impact weight corresponding to the current performance indicator coefficient is the quotient of the current performance indicator coefficient and the sum of the positive coefficients of the target system performance type; when the current performance indicator coefficient belongs to the negative coefficient set, the system performance impact weight corresponding to the current performance indicator coefficient is the quotient of the current performance indicator coefficient and the sum of the negative coefficients of the target system performance type; the target system performance type is the system performance type corresponding to the current performance indicator coefficient, the positive coefficient sum is the sum of all performance indicator coefficients corresponding to the target system performance type in the positive coefficient set, and the negative coefficient sum is the sum of all performance indicator coefficients corresponding to the target system performance type in the negative coefficient set.

9. A performance tuning device for a disk array card, characterized in that, include: An acquisition module is used to acquire model training data; wherein the model training data includes system performance data and performance management data of the disk array card; A training module is used to train the multivariate linear regression model according to the model training data to obtain a target model after the training is completed; wherein the multivariate linear regression model is used to reflect the corresponding relationship between the system performance data of each system performance type and the performance dotting data of the corresponding dotting performance type, and the corresponding relationship includes the performance index coefficient of the dotting performance type corresponding to each of the system performance types; A generation module is used to generate performance tuning guidance information according to the performance indicator coefficients in the target model; wherein the performance tuning guidance information includes at least one of the dot performance type identifier, performance dot data and system performance type identifier corresponding to the target performance indicator coefficient.

10. A performance tuning device for a disk array card, characterized in that, include: Memory for storing computer programs; A processor is used to implement the steps of the disk array card performance tuning method as claimed in any one of claims 1 to 8 when executing the computer program.

Images