AI-assisted data center equipment performance system compatibility automatic acceptance system

Abstract

The invention discloses an AI-assisted data center equipment performance system compatibility automatic acceptance system, relates to the field of computers, and solves the problem that an existing data center equipment performance system compatibility automatic acceptance system is poor in acceptance effect. Setting a real-time performance monitoring period for the target acceptance equipment group, performing performance index analysis on the target acceptance equipment group in the real-time performance monitoring period, and acquiring equipment performance index acquisition data according to an analysis result, the compatibility acceptance module carries out performance index compatibility analysis on the target acceptance equipment group according to the equipment performance index acquisition data and obtains performance compatibility judgment data according to an analysis result, and the acceptance feedback module carries out performance index acceptance result feedback on the target acceptance equipment group according to the performance compatibility judgment data. The system compatibility automatic acceptance efficiency of the data center equipment performance can be improved.

Description

Technical Field

[0001] This invention belongs to the field of computer science and involves artificial intelligence technology, specifically an AI-assisted automatic acceptance system for the performance and compatibility of data center equipment. Background Technology

[0002] Existing automated performance compatibility acceptance systems for data center equipment have the following shortcomings when performing performance compatibility acceptance on equipment groups:

[0003] 1. Existing automatic acceptance systems for the performance compatibility of data center equipment often adopt a method of parallel acceptance of multiple performance indicators when conducting performance compatibility acceptance of equipment groups. Since it is impossible to assess the performance compatibility status of the equipment group in advance, if the equipment group cannot withstand the high load brought by multiple parallel processes, the overall system response speed will drop significantly, making it difficult to carry out the acceptance process effectively.

[0004] 2. When performing performance compatibility acceptance tests on equipment groups, existing data center equipment performance system compatibility automatic acceptance systems typically conduct synchronous performance index deviation analysis on multiple virtual devices that can run in the equipment group. When multiple virtual devices are running and analyzed synchronously, the mutual influence between devices may cause crosstalk to the performance index, resulting in deviations in the performance index and making the analysis results unable to accurately reflect the performance compatibility of the equipment group itself.

[0005] To address this, we propose an AI-assisted automatic acceptance system for the performance and compatibility of data center equipment. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this invention is to provide an AI-assisted automatic acceptance system for the performance and system compatibility of data center equipment. This invention aims to improve the acceptance efficiency of the AI-assisted automatic acceptance system for the performance and system compatibility of data center equipment.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: an AI-assisted automatic acceptance system for the performance and system compatibility of data center equipment, characterized in that it includes:

[0008] The data acquisition module acquires the target acceptance equipment group, sets a real-time performance monitoring cycle for the target acceptance equipment group, performs performance index analysis on the target acceptance equipment group in the real-time performance monitoring cycle, and acquires equipment performance index data based on the analysis results.

[0009] The compatibility acceptance module performs a performance index compatibility analysis on the target acceptance equipment group based on the collected equipment performance index data, and obtains performance compatibility judgment data based on the analysis results.

[0010] The acceptance feedback module provides feedback on the performance index acceptance results of the target acceptance equipment group based on the performance compatibility judgment data.

[0011] Furthermore, the equipment performance data is collected as follows:

[0012] The data center equipment groups deployed by the data center equipment performance system are obtained, resulting in multiple data center equipment groups. Then, one target acceptance equipment group is randomly selected from the multiple data center equipment groups obtained.

[0013] During the performance compatibility acceptance of the target acceptance equipment group, the start time of the current computing task of the target acceptance equipment group is obtained to obtain the first characteristic time point. The time point corresponding to the current moment is used as the second characteristic time point. The time period between the first characteristic time point and the second characteristic time point is set as the historical analysis period of performance indicators.

[0014] Mark the physical devices in the target acceptance device group as target host devices, and acquire the virtual devices in the target acceptance device group to obtain multiple target virtual devices;

[0015] During the performance compatibility acceptance of the target acceptance equipment group, several different types of equipment performance indicators are set, and one sample performance indicator is randomly selected from them.

[0016] If the sample performance metric is CPU load, perform CPU load consistency analysis on the target virtual device, and obtain the performance consistency of the metric corresponding to the sample performance metric based on the analysis results.

[0017] Each device's performance index is acquired and analyzed. Based on the analysis results, the performance consistency of each device's performance index is obtained, thus acquiring the device performance index data.

[0018] Furthermore, a consistency analysis was conducted on the sample performance metrics, as detailed below:

[0019] Based on the time slice rotation order of the target host device, the acquired multiple target virtual devices are sequentially marked as virtual device V1, virtual device V2, virtual device V3... virtual device Va;

[0020] Collect historical CPU load data for the V1 virtual device during the historical performance analysis period, and plot CPU load change curves based on the historical CPU load data.

[0021] The CPU load change curve is divided into several curve division points, and any two consecutive curve division points are set as adjacent curve points to obtain multiple adjacent curve point combinations.

[0022] The existence of load change points is determined for adjacent points of the curve, and the load change rate at the time corresponding to the load change characteristic point is obtained based on the determination result.

[0023] The load change feature points of the V1 virtual device during the historical analysis period of performance indicators are obtained, a time axis is created for the historical analysis period of performance indicators, and the load change feature points are marked in the created time axis to obtain the V1 load change time axis.

[0024] Repeat the process of creating the V1 load change time axis to obtain the load change time axis corresponding to different target virtual devices, and obtain the V2 load change time axis to the Va load change time axis. Then, name the multiple load change feature points in the load change time axis in order of increasing acquisition time value as the first priority load change feature point, the second priority load change feature point, the third priority load change feature point, ... the bth priority load change feature point.

[0025] The first load change feature point in the V1 load change time axis is sampled to obtain the V1 sampling time value Tv1. The first load change feature point in the V2 load change time axis is sampled to obtain the V2 sampling time value Tv2. The duration of the time slice allocated to the target host device at the V1 sampling time value is obtained to obtain the V1 time slice duration Pv1. The load change time deviation ratio is calculated by |Tv1-Tv2| / Pv1. A preset value for the load change time deviation ratio is set. If the load change time deviation ratio is greater than the preset value, the first load change feature point is directly classified as an abnormal load feature point.

[0026] Furthermore, a consistency analysis was conducted on the sample performance metrics, as detailed below:

[0027] If the load change time deviation ratio is less than or equal to the preset value of the load change time deviation ratio, then the load change rate at the time corresponding to the first load change feature point in the V1 load change time axis is obtained to obtain the load change rate at time V1. The load change rate at the time corresponding to the first load change feature point in the V2 load change time axis is obtained to obtain the load change rate at time V2. The difference between the load change rate at time V1 and the load change rate at time V2 is calculated, and the absolute value of the obtained difference is taken to obtain the load change rate deviation. The preset deviation of the load change rate is set. If the load change rate deviation is greater than the preset deviation of the load change rate, then the first load change feature point is classified as an abnormal load feature point. If the load change rate deviation is less than or equal to the preset deviation of the load change rate, then the first load change feature point is classified as an abnormal load feature point. Then the V1 load change time axis is replaced with the V3 load change time axis, and the above process is repeated. If no abnormal load feature point appears after replacing with the Va load change time axis, then the first load change feature point is marked as a normal load feature point.

[0028] If the first priority load change feature point is an abnormal load feature point, then the second priority load change feature point is used to replace the first priority load change feature point, and the second priority load change feature point is divided, and so on, until the type division of the b-th priority load change feature point is completed.

[0029] The number of normal load characteristic points is statistically analyzed to obtain the number of normal load characteristic points. The ratio of the number of normal load characteristic points to b is calculated to obtain the consistency of the performance of the corresponding index.

[0030] Furthermore, the load change rate at the corresponding time point of the load change characteristic is obtained, as follows:

[0031] In any adjacent combination of curves, the curve segmentation point whose acquisition time value is relatively close to the start time of the historical analysis period of the performance index is marked as the first combination feature segmentation point, and the curve segmentation point whose acquisition time value is relatively close to the end time of the historical analysis period of the performance index is marked as the second combination feature segmentation point.

[0032] The historical CPU load of the V1 virtual device at the first combination feature segmentation point is collected to obtain the CPU load value at the first segmentation point. The historical CPU load of the V1 virtual device at the second combination feature segmentation point is collected to obtain the CPU load value at the second segmentation point. If the CPU load value at the first segmentation point is equal to the CPU load value at the second segmentation point, it is determined that there is no load change point in the adjacent combination of curves. If the CPU load value at the first segmentation point is not equal to the CPU load value at the second segmentation point, the second combination feature segmentation point is set as the load change feature point. The difference between the CPU load values ​​at the first segmentation point and the CPU load values ​​at the second segmentation point is calculated, and the ratio of the absolute value of the difference to the CPU load value at the first segmentation point is calculated to obtain the load change rate at the time corresponding to the load change feature point.

[0033] Furthermore, the performance compatibility assessment data is obtained as follows:

[0034] Acquire equipment performance index data, obtain the performance consistency of each equipment performance index based on the collected data, and name the equipment performance indexes as Y1 compatibility acceptance index, Y2 compatibility acceptance index, Y3 compatibility acceptance index... Yc compatibility acceptance index according to the performance consistency value from small to large, thus obtaining the compatibility acceptance sorting queue;

[0035] If the Y1 compatibility acceptance criterion is a memory performance criterion, then a memory performance compatibility analysis is performed on the target acceptance equipment group, and the Y1 compatibility acceptance criterion is judged to be abnormal in terms of performance compatibility based on the analysis results.

[0036] According to the compatibility acceptance sorting queue, compatibility anomaly judgments are performed on the Y2 compatibility acceptance indicators to the Yc compatibility acceptance indicators respectively to obtain performance compatibility judgment data.

[0037] Furthermore, a memory performance compatibility analysis was conducted on the target acceptance equipment group, as detailed below:

[0038] The target host devices and multiple target virtual devices included in the target acceptance device group are acquired, and an initial virtual device is selected from the acquired multiple target virtual devices. A memory compatibility acceptance program is created for the initial virtual device, and the time period during which the target host device only enables the initial virtual device to run the memory compatibility acceptance program is set as the first memory performance acceptance period.

[0039] The memory access operations performed by the initial virtual device during the first memory performance acceptance period are acquired, and the acquired memory access operations are named N1 memory access operation, N2 memory access operation, N3 memory access operation, ... Nd memory access operation in chronological order of operation time.

[0040] During the first memory performance acceptance period, the operation time value corresponding to each memory access operation is obtained to obtain the N1 memory operation time value to the Nd memory operation time value. The A1 performance analysis tool is used to map the memory access latency corresponding to the N1 memory operation time value to the Nd memory operation time value of the initial virtual device to the N1 memory access operation to the Nd memory access operation to obtain the N1 initial performance latency to the Nd initial performance latency.

[0041] Furthermore, a memory performance compatibility analysis was conducted on the target acceptance equipment group, as detailed below:

[0042] Using the A1 spreadsheet tool, a Cartesian coordinate system is created with the operation time value as the x-axis and the performance latency as the y-axis to obtain the first time period acceptance time coordinate system. In the first time period acceptance time coordinate system, the memory operation time value is used as the x-axis and the initial performance latency is used as the y-axis to create coordinate points, resulting in the N1 initial performance coordinate point to the Nd initial performance coordinate point.

[0043] After the first memory performance acceptance period ends, the target host device needs to maintain the running state of the initial virtual device so that it continues to repeat the memory compatibility acceptance procedure. At the same time, an A1 target virtual device is dynamically added to the target host device, and it is ensured that the new device runs the memory compatibility acceptance procedure synchronously with the initial virtual device. The period in which the memory compatibility acceptance procedure is run is marked as the second memory performance acceptance period.

[0044] A memory performance compatibility assessment was conducted on the target acceptance equipment group during the second memory performance acceptance period. Based on the assessment results, the second memory performance acceptance period was divided into an effective compatibility period and a failure compatibility period.

[0045] Furthermore, a memory performance compatibility analysis was conducted on the target acceptance equipment group, as detailed below:

[0046] If the second memory performance acceptance period is classified as a failure compatibility period, then the Y1 compatibility acceptance index is determined to have compatibility anomalies. If the second memory performance acceptance period is classified as a valid compatibility period, then after the second memory performance acceptance period ends, the target host device maintains the running state of the initial virtual device and the A1 target virtual device, allowing it to continue to repeatedly execute the memory compatibility acceptance program. At the same time, an A2 target virtual device is dynamically added to the target host device, and it is ensured that the new device runs the memory compatibility acceptance program synchronously with the initial virtual device. The period in which the memory compatibility acceptance program runs is marked as the third memory performance acceptance period, and performance compatibility analysis is performed on the third memory performance acceptance period. If the third memory performance acceptance period is classified as a valid compatibility period, then the third memory performance acceptance period is obtained for performance compatibility analysis, and so on, until all target virtual devices are added.

[0047] If any memory performance acceptance period is a failure compatibility period, then the Y1 compatibility acceptance indicator is determined to be an abnormal compatibility indicator. If any memory performance acceptance period is a failure compatibility period during the loop, then the Y1 compatibility acceptance indicator is determined to be an abnormal compatibility indicator. If none of these periods exist, then the Y1 compatibility acceptance indicator is determined to be a normal compatibility indicator.

[0048] Furthermore, a memory performance compatibility assessment was conducted on the target acceptance equipment group during the second memory performance acceptance phase, as detailed below:

[0049] In the first time period, the performance coordinate points of the new target virtual device are created in the time period coordinate system, resulting in the newly added performance coordinate points A1 to Ad.

[0050] In the first period of the acceptance period coordinate system, the distance between the newly added performance coordinate point A1 and the initial performance coordinate point N1 is obtained to obtain the initial performance deviation of A1. The distance between the newly added performance coordinate point A2 and the initial performance coordinate point N2 is obtained to obtain the initial performance deviation of A2. Similarly, the distance between the newly added performance coordinate point Ad and the initial performance coordinate point Nd is obtained to obtain the initial performance deviation of Ad.

[0051] The number of newly added performance coordinate points during the second memory performance acceptance period with an initial performance deviation of 0 is counted to obtain the new performance overlap. The average of the initial performance deviation from A1 to Ad is calculated to obtain the first parallel performance deviation characteristic value. The standard deviation of the initial performance deviation from A1 to Ad is calculated to obtain the second parallel performance deviation characteristic value.

[0052] The virtual device memory conflict degree corresponding to the second memory performance acceptance period is obtained by calculating the newly added performance overlap, the first parallel performance deviation feature value, and the second parallel performance deviation feature value.

[0053] The virtual device memory conflict level corresponding to the second memory performance acceptance period is calculated using the following formula: ;

[0054] Where Jrd is the virtual device memory conflict degree corresponding to the second memory performance acceptance period, Chd is the new performance overlap degree, Tzz1 is the first parallel performance deviation feature value, and Tzz2 is the second parallel performance deviation feature value.

[0055] Set a baseline range for virtual device memory conflict. If the virtual device memory conflict is within the baseline range, the second memory performance acceptance period is determined to be a valid compatibility period. If the virtual device memory conflict is not within the baseline range, the second memory performance acceptance period is determined to be a invalid compatibility period.

[0056] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0057] 1. This invention conducts periodic performance index load consistency analysis on different performance indicators of multiple virtual devices in the target acceptance equipment group. Based on the analysis results, the acceptance priority of these device performance indicators is sorted, and then compatibility acceptance is carried out according to the sorting results. This can prioritize the allocation of computing resources to the acceptance of device performance indicators with large fluctuations in load consistency, discover performance compatibility issues more quickly, and thus improve the overall compatibility acceptance efficiency.

[0058] 2. When accepting individual performance indicators, this invention adopts a cyclical incremental approach to gradually increase the number of verification individuals and conducts verification work based on the degree of performance indicator conflict generated when different devices are connected. This avoids the problem of excessive resource consumption caused by connecting a large number of devices for acceptance at once. In the initial stage of acceptance, only a small number of devices are needed for verification. Then, the number of devices is gradually increased according to the actual situation of performance indicator conflict, thereby achieving a more reasonable allocation and utilization of computing, storage and other resources, and effectively improving the efficiency of resource utilization. Attached Figure Description

[0059] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0060] Figure 1 This is an overall system block diagram of the present invention;

[0061] Figure 2 This is the V1 load change time axis of the present invention;

[0062] Figure 3 This is the coordinate system for the first time period of acceptance in this invention. Detailed Implementation

[0063] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0064] Example 1

[0065] Please see Figure 1 This invention provides a technical solution: an AI-assisted automatic acceptance system for the performance and compatibility of data center equipment. The specific working process of each module is as follows:

[0066] The data acquisition module acquires the target acceptance equipment group, sets a real-time performance monitoring cycle for the target acceptance equipment group, performs performance index analysis on the target acceptance equipment group in the real-time performance monitoring cycle, and acquires equipment performance index data based on the analysis results.

[0067] Specifically as follows:

[0068] The data center equipment groups deployed by the data center equipment performance system are obtained, resulting in multiple data center equipment groups. Then, one target acceptance equipment group is randomly selected from the multiple data center equipment groups obtained.

[0069] It should be noted here that:

[0070] In this application, the data center equipment group referred to herein is specifically a data center equipment group equipped with an automated compatibility acceptance system.

[0071] In this application, the data center equipment group referred to herein specifically includes physical equipment and virtual equipment.

[0072] During the performance compatibility acceptance of the target acceptance equipment group, the start time of the current computing task of the target acceptance equipment group is obtained to obtain the first characteristic time point. The time point corresponding to the current moment is used as the second characteristic time point. The time period between the first characteristic time point and the second characteristic time point is set as the historical analysis period of performance indicators.

[0073] It should be noted here that:

[0074] In this application, the current computing task executed by the target acceptance equipment group specifically refers to the computing task executed by the virtual machine in the target acceptance equipment group. The computing tasks involved here include, but are not limited to, molecular dynamics simulation, user clustering analysis, and transaction data mining.

[0075] Perform performance resource analysis on the target acceptance equipment group that is in the historical performance index analysis cycle, and obtain performance resource ranking data based on the analysis results;

[0076] Specifically as follows:

[0077] Mark the physical devices in the target acceptance device group as target host devices, and acquire the virtual devices in the target acceptance device group to obtain multiple target virtual devices;

[0078] It should be noted here that:

[0079] In this application, the target acceptance device group involved herein includes a target host device and multiple virtual devices residing on the target host device. The target host device involved herein is a physical host, and the virtual devices involved herein are specifically VMware virtual machines.

[0080] Throughout the historical performance metric analysis period, the tasks executed by different VMware virtual machines and the accuracy of those tasks remained consistent.

[0081] During the performance compatibility acceptance of the target acceptance equipment group, several different types of equipment performance indicators are set, and one sample performance indicator is randomly selected from them.

[0082] It should be noted here that:

[0083] In this application, the device performance metrics referred to herein include, but are not limited to, CPU load, memory load, and network bandwidth load.

[0084] If the sample performance metric is CPU load, perform CPU load consistency analysis on the target virtual device, and obtain the performance consistency of the metric corresponding to the sample performance metric based on the analysis results.

[0085] Specifically as follows:

[0086] Based on the time slice rotation order of the target host device, the acquired multiple target virtual devices are sequentially marked as virtual device V1, virtual device V2, virtual device V3... virtual device Va;

[0087] It should be noted here that:

[0088] In this application, V1, V2, V3...Va in virtual device V1, virtual device V2, virtual device V3...Va are the numbers corresponding to the target virtual device, and a is an integer greater than 0.

[0089] In this application, the scheduler of the target host device (such as Linux's CFS or Xen's credit scheduler) will determine which virtual device should occupy the CPU at different times according to the time-slice round-robin rule. The scheduling order here is virtual device V1 → virtual device V2 → virtual device V3 → ... → virtual device Va → virtual device V1 → virtual device V2 → virtual device V3 → ... → virtual device Va → ... virtual device Vi. Here, virtual device Vi can be any virtual device from virtual device to virtual device Va.

[0090] Collect historical CPU load data for the V1 virtual device during the historical performance analysis period, and plot CPU load change curves based on the historical CPU load data.

[0091] The CPU load change curve is divided into several curve division points, and any two consecutive curve division points are set as adjacent curve points to obtain multiple adjacent curve point combinations.

[0092] It should be noted here that:

[0093] In this application, the time interval between any two consecutive curve dividing points is 10 milliseconds.

[0094] In any adjacent combination of curves, the curve segmentation point whose acquisition time value is relatively close to the start time of the historical analysis period of the performance index is marked as the first combination feature segmentation point, and the curve segmentation point whose acquisition time value is relatively close to the end time of the historical analysis period of the performance index is marked as the second combination feature segmentation point.

[0095] It should be noted here that:

[0096] In the specific implementation, it is known that the historical analysis period of the performance index is 6:10:00-9:10:00, and the acquisition time values ​​corresponding to the combination of adjacent points of the curve are 7:10:11:11-7:10:11:21 respectively. Then, the curve segmentation point corresponding to the acquisition time value of 7:10:11:11 is the first combination feature segmentation point, and the curve segmentation point corresponding to the acquisition time value of 7:10:11:21 is the second combination feature segmentation point.

[0097] The historical CPU load of the V1 virtual device at the first combination feature segmentation point is collected to obtain the CPU load value at the first segmentation point. The historical CPU load of the V1 virtual device at the second combination feature segmentation point is collected to obtain the CPU load value at the second segmentation point. If the CPU load value at the first segmentation point is equal to the CPU load value at the second segmentation point, it is determined that there is no load change point in the adjacent combination of curves. If the CPU load value at the first segmentation point is not equal to the CPU load value at the second segmentation point, the second combination feature segmentation point is set as the load change feature point. The difference between the CPU load value at the first segmentation point and the CPU load value at the second segmentation point is calculated, and the ratio of the absolute value of the difference to the CPU load value at the first segmentation point is calculated to obtain the load change rate at the time corresponding to the load change feature point.

[0098] Please see Figure 2 The load change feature points of the V1 virtual device during the historical analysis period of performance indicators are obtained, a time axis is created for the historical analysis period of performance indicators, and the load change feature points are marked in the created time axis to obtain the V1 load change time axis.

[0099] Repeat the process of creating the V1 load change time axis to obtain the load change time axis corresponding to different target virtual devices, and obtain the V2 load change time axis to the Va load change time axis. Then, name the multiple load change feature points in the load change time axis in order of increasing acquisition time value as the first priority load change feature point, the second priority load change feature point, the third priority load change feature point, ... the bth priority load change feature point.

[0100] It should be noted here that:

[0101] In this application, the first load change feature point, the second load change feature point, the third load change feature point, ... the bth load change feature point are the marker symbols corresponding to the load change feature points, and b is an integer greater than 0.

[0102] The first load change feature point in the V1 load change time axis is sampled to obtain the V1 sampling time value Tv1. The first load change feature point in the V2 load change time axis is sampled to obtain the V2 sampling time value Tv2. The duration of the time slice allocated to the target host device in the V1 sampling time value is obtained to obtain the V1 time slice duration Pv1. The load change time deviation ratio is calculated by |Tv1-Tv2| / Pv1. A preset value for the load change time deviation ratio is set. If the load change time deviation ratio is greater than the preset value, the first load change feature point is directly classified as an abnormal load feature point.

[0103] If the load change time deviation ratio is less than or equal to the preset value of the load change time deviation ratio, then the load change rate at the time corresponding to the first load change feature point in the V1 load change time axis is obtained to obtain the load change rate at time V1. The load change rate at the time corresponding to the first load change feature point in the V2 load change time axis is obtained to obtain the load change rate at time V2. The difference between the load change rate at time V1 and the load change rate at time V2 is calculated, and the absolute value of the obtained difference is taken to obtain the load change rate deviation. The preset deviation of the load change rate is set. If the load change rate deviation is greater than the preset deviation of the load change rate, then the first load change feature point is classified as an abnormal load feature point. If the load change rate deviation is less than or equal to the preset deviation of the load change rate, then the first load change feature point is classified as an abnormal load feature point. Then the V1 load change time axis is replaced with the V3 load change time axis, and the above process is repeated. If no abnormal load feature point appears after replacing with the Va load change time axis, then the first load change feature point is marked as a normal load feature point.

[0104] If the first priority load change feature point is an abnormal load feature point, then the second priority load change feature point is used to replace the first priority load change feature point, and the second priority load change feature point is divided, and so on, until the type division of the b-th priority load change feature point is completed.

[0105] The number of normal load characteristic points is statistically analyzed to obtain the number of normal load characteristic points. The ratio of the number of normal load characteristic points to b is calculated to obtain the consistency of the performance of the corresponding index.

[0106] Repeat the process of obtaining the performance consistency of the corresponding indicators for the sample performance indicators, obtain the performance analysis of each equipment performance indicator, and obtain the performance consistency of each equipment performance indicator based on the analysis results to obtain the equipment performance indicator collection data.

[0107] The compatibility acceptance module performs a performance index compatibility analysis on the target acceptance equipment group based on the collected equipment performance index data, and obtains performance compatibility judgment data based on the analysis results.

[0108] Specifically as follows:

[0109] Acquire equipment performance index data, obtain the performance consistency of each equipment performance index based on the collected data, and name the equipment performance indexes as Y1 compatibility acceptance index, Y2 compatibility acceptance index, Y3 compatibility acceptance index... Yc compatibility acceptance index according to the performance consistency value from small to large, thus obtaining the compatibility acceptance sorting queue;

[0110] It should be noted here that:

[0111] In this application, Y1, Y2, Y3, ..., Yc are the compatibility acceptance ranking numbers of the equipment performance indicators, respectively. Yc is the last-ranked equipment performance indicator, and c is the quantity value corresponding to the equipment performance indicator, which is an integer greater than 0.

[0112] If the Y1 compatibility acceptance criterion is a memory performance criterion, then a memory performance compatibility analysis is performed on the target acceptance equipment group, and the Y1 compatibility acceptance criterion is judged to be abnormal in terms of performance compatibility based on the analysis results.

[0113] Specifically as follows:

[0114] The target host devices and multiple target virtual devices included in the target acceptance device group are acquired, and an initial virtual device is selected from the acquired multiple target virtual devices. A memory performance compatibility acceptance program is created for the initial virtual device, and the time period during which the target host device only enables the initial virtual device to run the memory performance compatibility acceptance program is set as the first memory performance acceptance time period.

[0115] It should be noted here that:

[0116] In this application, during the first memory performance acceptance period, the target host device has only one running instance, that is, the initial virtual device runs the memory performance compatibility acceptance program;

[0117] In this application, the memory performance compatibility acceptance procedure referred to herein is specifically the acceptance procedure executed by the initial virtual device during the first memory performance acceptance period. The acceptance procedure can call multiple memory access operations, including but not limited to single read / write operations, memory mapping operations, and memory copy operations.

[0118] The memory access operations performed by the initial virtual device during the first memory performance acceptance period are acquired, and the acquired memory access operations are named N1 memory access operation, N2 memory access operation, N3 memory access operation, ... Nd memory access operation in chronological order of operation time.

[0119] It should be noted here that:

[0120] In this application, N1, N2, N3, ... Nd memory access operations are memory access operations, where N1, N2, N3, ... Nd are the sequence numbers corresponding to the memory access operations, and d is the quantity value corresponding to the memory access operation, and d is an integer greater than 0.

[0121] During the first memory performance acceptance period, the operation time value corresponding to each memory access operation is obtained to obtain the N1 memory operation time value to the Nd memory operation time value. The A1 performance analysis tool is used to map the memory access latency corresponding to the N1 memory operation time value to the Nd memory operation time value of the initial virtual device to the N1 memory access operation to the Nd memory access operation to obtain the N1 initial performance latency to the Nd initial performance latency.

[0122] It should be noted here that:

[0123] In this application, the A1 performance analysis tool described herein is capable of compensating for the time difference between memory access operation time and memory device response time;

[0124] In this application, the initial performance latency referred to herein is specifically the memory access latency corresponding to different memory operation time values ​​of the initial virtual device;

[0125] In this application, the N1 memory operation time value to the Nd memory operation time value are determined relative to the first memory performance acceptance period. If the duration of the first memory performance acceptance period is 5 minutes, then the start time value of the first memory performance acceptance period is marked as 0 seconds, and the end time value of the first memory performance acceptance period is marked as 300 seconds. If the time interval between the Ni memory access operation and the end time of the first memory performance acceptance period is 5 seconds, then the Ni memory operation time value can be marked as 5 seconds.

[0126] Using the A1 spreadsheet tool, a Cartesian coordinate system is created with the operation time value as the x-axis and the performance latency as the y-axis to obtain the first time period acceptance time coordinate system. In the first time period acceptance time coordinate system, the memory operation time value is used as the x-axis and the initial performance latency is used as the y-axis to create coordinate points, resulting in the N1 initial performance coordinate point to the Nd initial performance coordinate point.

[0127] It should be noted here that:

[0128] In this application, if the N1 memory operation time is 2.1s and the N1 initial performance latency is 0.1s, then the N1 initial performance coordinate point is (2.1, 0.1).

[0129] After the first memory performance acceptance period ends, the target host device needs to maintain the running state of the initial virtual device so that it continues to repeatedly execute the memory performance compatibility acceptance program. At the same time, an A1 target virtual device is dynamically added to the target host device, and it is ensured that the new device runs the memory performance compatibility acceptance program synchronously with the initial virtual device. The period in which the memory performance compatibility acceptance program is run is marked as the second memory performance acceptance period.

[0130] A memory performance compatibility assessment was conducted on the target acceptance equipment group during the second memory performance acceptance period. Based on the assessment results, the second memory performance acceptance period was divided into an effective compatibility period and a failure compatibility period.

[0131] Specifically as follows:

[0132] In the first time period, the performance coordinate points of the new target virtual device are created in the time period coordinate system, resulting in the newly added performance coordinate points A1 to Ad.

[0133] Please see Figure 3 In the first period of the acceptance period coordinate system, the distance between the newly added performance coordinate point A1 and the initial performance coordinate point N1 is obtained to obtain the initial performance deviation of A1. The distance between the newly added performance coordinate point A2 and the initial performance point N2 is obtained to obtain the initial performance deviation of A2. Similarly, the distance between the newly added performance coordinate point Ad and the initial performance point Nd is obtained to obtain the initial performance deviation of Ad.

[0134] The number of newly added performance coordinate points during the second memory performance acceptance period with an initial performance deviation of 0 is counted to obtain the new performance overlap. The average of the initial performance deviation from A1 to Ad is calculated to obtain the first parallel performance deviation characteristic value. The standard deviation of the initial performance deviation from A1 to Ad is calculated to obtain the second parallel performance deviation characteristic value.

[0135] The virtual device memory conflict degree corresponding to the second memory performance acceptance period is obtained by calculating the newly added performance overlap, the first parallel performance deviation feature value, and the second parallel performance deviation feature value.

[0136] The virtual device memory conflict level corresponding to the second memory performance acceptance period is calculated using the following formula: ;

[0137] Where Jrd is the virtual device memory conflict degree corresponding to the second memory performance acceptance period, Chd is the new performance overlap degree, Tzz1 is the first parallel performance deviation feature value, and Tzz2 is the second parallel performance deviation feature value.

[0138] It should be noted here that:

[0139] The newly added performance overlap (Chd) clearly reflects the degree of overlap between the newly added performance coordinate points and existing performance coordinate points during the second memory performance acceptance period. The first parallel performance deviation characteristic value (Tzz1) is calculated by averaging the initial performance deviation, reflecting the central tendency of the initial performance deviation. If this average value is large, it means that the overall initial performance deviates significantly from the expected or normal level. Such a large average deviation often suggests that there may be many performance conflicts during the operation of the virtual device. The second parallel performance deviation characteristic value (Tzz2) is calculated by calculating the standard deviation of the initial performance deviation. It is mainly used to measure the dispersion of the initial performance deviation. A large standard deviation indicates that the initial performance deviation fluctuates greatly and the performance is not stable enough, indicating that the memory performance compatibility of the target virtual device in the target host device is abnormal. The three key factors of newly added performance overlap (Chd), first parallel performance deviation characteristic value (Tzz1), and second parallel performance deviation characteristic value (Tzz2) are comprehensively considered. It not only focuses on the relationship between new performance and existing performance, but also analyzes in depth the overall level and fluctuation of initial performance deviation. By combining these factors organically through reasonable mathematical calculations, it can comprehensively and accurately assess the memory conflict degree of virtual devices during the second memory performance acceptance period.

[0140] Set a baseline range for virtual device memory conflict. If the virtual device memory conflict is within the baseline range, the second memory performance acceptance period is determined to be a valid compatibility period. If the virtual device memory conflict is not within the baseline range, the second memory performance acceptance period is determined to be a failed compatibility period.

[0141] It should be noted here that:

[0142] In this application, the effective compatibility period referred to herein includes the situation where the virtual device memory conflict level is at the boundary of the virtual device memory conflict level baseline interval;

[0143] In this application, the lower limit of the virtual device memory conflict degree benchmark interval is 0. The second memory performance acceptance period that was judged to be a valid compatibility period in history is obtained to obtain multiple second valid historical periods. The virtual device memory conflict degree corresponding to each second valid historical period is obtained, and the values ​​of the multiple virtual device memory conflict degrees are compared. The virtual device memory conflict degree with the smallest value is set as the upper limit of the virtual device memory conflict degree benchmark interval.

[0144] If the second memory performance acceptance period is classified as a failure compatibility period, then the Y1 compatibility acceptance index is determined to have compatibility anomalies. If the second memory performance acceptance period is classified as a valid compatibility period, then after the second memory performance acceptance period ends, the target host device maintains the running state of the initial virtual device and the A1 target virtual device, allowing it to continue to repeatedly execute the memory performance compatibility acceptance program. At the same time, an A2 target virtual device is dynamically added to the target host device, and it is ensured that the new device runs the memory performance compatibility acceptance program synchronously with the initial virtual device. The period in which the memory performance compatibility acceptance program is run is marked as the third memory performance acceptance period, and performance compatibility analysis is performed on the third memory performance acceptance period. If the third memory performance acceptance period is classified as a valid compatibility period, then the third memory performance acceptance period is obtained for performance compatibility analysis, and so on, until all target virtual devices are added.

[0145] If any memory performance acceptance period is a failure compatibility period, then the Y1 compatibility acceptance indicator is determined to be an abnormal compatibility indicator. If any memory performance acceptance period is a failure compatibility period during the loop, then the Y1 compatibility acceptance indicator is determined to be an abnormal compatibility indicator. If none of these periods exist, then the Y1 compatibility acceptance indicator is determined to be a normal compatibility indicator.

[0146] It should be noted here that:

[0147] In this application, a compatibility anomaly warning can be issued to the target acceptance equipment group when the first failure compatibility period occurs.

[0148] Repeat the process of judging performance compatibility anomalies for the Y1 compatibility acceptance index, and judge the compatibility anomalies for the Y2 compatibility acceptance index to the Yc compatibility acceptance index according to the order of the compatibility acceptance sorting queue to obtain performance compatibility judgment data.

[0149] If the Y1 compatibility acceptance criterion is memory access latency, then a memory performance compatibility analysis is performed on the target acceptance equipment group. Based on the analysis results, the Y1 compatibility acceptance criterion is judged as an abnormal compatibility criterion. If it does not exist, then the Y1 compatibility acceptance criterion is judged as a normal compatibility criterion.

[0150] The acceptance feedback module provides feedback on the performance index acceptance results of the target acceptance equipment group based on the performance compatibility judgment data.

[0151] Specifically as follows:

[0152] Obtain performance compatibility judgment data, and based on the performance compatibility judgment data, obtain the compatibility acceptance results corresponding to the Y1 compatibility acceptance index to the Yc compatibility acceptance index respectively;

[0153] When the Y1 compatibility acceptance metric is a memory performance metric, if the Y1 compatibility acceptance metric is a normal compatibility metric, the target acceptance equipment group's memory performance compatibility is normal; if the Y1 compatibility acceptance metric is an abnormal compatibility metric, the target acceptance equipment group's memory performance compatibility is abnormal.

[0154] Repeat the process of providing acceptance feedback for the Y1 compatibility acceptance indicators, and provide acceptance feedback for each compatibility acceptance indicator separately.

[0155] Compared to the problems described in the background technology, this invention conducts periodic performance index load consistency analysis on different performance indicators of multiple virtual devices in the target acceptance equipment group. Based on the analysis results, these device performance indicators are prioritized for acceptance, and compatibility acceptance is carried out separately according to the ranking results. This allows computing resources to be preferentially allocated to the acceptance of device performance indicators with large fluctuations in load consistency, enabling faster discovery of compatibility issues and improving the overall efficiency of compatibility acceptance. Furthermore, when accepting individual performance indicators, this invention uses a cyclical incremental approach to gradually increase the number of verification individuals and conducts verification work based on the degree of performance index conflict generated when different device individuals are connected. This avoids the problem of excessive resource consumption caused by connecting a large number of devices for acceptance at once. In the initial stage of acceptance, only a small number of devices are used for verification, and then the number of devices is gradually increased according to the actual situation of performance index conflict, thereby achieving a more reasonable allocation and utilization of computing, storage and other resources, effectively improving resource utilization efficiency.

[0156] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An AI-assisted automatic acceptance system for data center equipment performance and system compatibility, characterized in that, include: Data acquisition module: Collects performance indicators for the target acceptance equipment group, creates a real-time performance monitoring cycle, performs performance indicator consistency analysis on the target acceptance equipment group in the real-time performance monitoring cycle, and obtains equipment performance indicator data based on the analysis results; Compatibility Acceptance Module: Based on the collected data of equipment performance indicators, the module sorts the equipment performance indicators to obtain a compatibility acceptance sorting queue. Based on the compatibility acceptance sorting queue, it performs performance indicator compatibility analysis on the target acceptance equipment group and obtains performance compatibility judgment data based on the analysis results. Acceptance Feedback Module: Provides feedback on the performance indicators of the target acceptance equipment group based on the performance compatibility judgment data.

2. The AI-assisted automatic acceptance system for data center equipment performance system compatibility according to claim 1, characterized in that, The data collected on the equipment performance indicators are as follows: Obtain the target acceptance equipment group, and create a historical analysis cycle for performance indicators during the performance compatibility acceptance process of the target acceptance equipment group; The physical devices in the target acceptance equipment group are acquired to obtain the target host device, and the virtual devices in the target acceptance equipment group are acquired to obtain multiple target virtual devices; During the performance compatibility acceptance of the target acceptance equipment group, several different types of equipment performance indicators are set, and one sample performance indicator is randomly selected from them. If the sample performance metric is CPU load, perform CPU load consistency analysis on the target virtual device, and obtain the performance consistency of the metric corresponding to the sample performance metric based on the analysis results. The performance indicators of each device are acquired and analyzed. Based on the analysis results, the performance consistency of each device's performance indicator is obtained, thus acquiring the device performance indicator data.

3. The AI-assisted automatic acceptance system for data center equipment performance system compatibility according to claim 2, characterized in that, A consistency analysis of the sample performance metrics was performed, as follows: Based on the time slice rotation order of the target host device, the acquired multiple target virtual devices are sequentially marked as virtual device V1 to virtual device Va; Collect historical CPU load data for the V1 virtual device during the historical performance analysis period and plot the CPU load change curve. The CPU load change curve is divided into several curve segmentation points and combined to obtain multiple combinations of adjacent curve points; The existence of load change points is determined for adjacent points of the curve, and the load change rate at the time corresponding to the load change characteristic point is obtained based on the determination result. Create load change time axes from V1 to Va, and name the multiple load change feature points existing in the load change time axes as the first sequential load change feature point to the b-th sequential load change feature point; Perform time deviation analysis on the first priority load change feature point in the V1 load change time axis to obtain the load change time deviation ratio. Set a preset value for the load change time deviation ratio. If the load change time deviation ratio is greater than the preset value, the first priority load change feature point is classified as an abnormal load feature point.

4. The AI-assisted automatic acceptance system for data center equipment performance system compatibility according to claim 3, characterized in that, A consistency analysis of the sample performance metrics was performed, as follows: If the load change time deviation ratio is less than or equal to the preset value, the load change rate at time V1 and the load change rate at time V2 are collected and analyzed to obtain the load change rate deviation, and the preset load change rate deviation is set. If the load change rate deviation is greater than the preset load change rate deviation, the first priority load change feature point is classified as an abnormal load feature point. If it is less than or equal to the preset value, the first priority load change feature point is classified as an abnormal load feature point. Then, the V1 load change time axis is replaced with the V3 load change time axis, and so on. If no abnormal load feature point appears after replacing the Va load change time axis, the first priority load change feature point is marked as a normal load feature point. If the first priority load change feature point is an abnormal load feature point, then the second priority load change feature point is used to replace the first priority load change feature point, and the second priority load change feature point is divided, and so on, until the type division of the bth priority load change feature point is completed. The number of normal load characteristic points is statistically analyzed to obtain the number of normal load characteristic points. The ratio of the number of normal load characteristic points to b is calculated to obtain the consistency of the performance of the corresponding index.

5. The AI-assisted automatic acceptance system for data center equipment performance system compatibility according to claim 3, characterized in that, The load change rate at any given time is obtained as follows: Mark the two curve segmentation points in the adjacent curve combination as the first combination feature segmentation point and the second combination feature segmentation point, respectively; The historical CPU load of the V1 virtual device is collected at the first and second combination feature segmentation points to obtain the CPU load values ​​at the first and second segmentation points. If the CPU load value at the first segmentation point is equal to the CPU load value at the second segmentation point, it is determined that there are no load change points in the adjacent combination of curves. If they are not equal, the second combination feature segmentation point is set as the load change feature point. The difference between the CPU load value at the first and second segmentation points is calculated, and the ratio of the absolute value of the difference to the CPU load value at the first segmentation point is calculated to obtain the load change rate at the time corresponding to the load change feature point.

6. The AI-assisted automatic acceptance system for data center equipment performance system compatibility according to claim 1, characterized in that, The performance compatibility assessment data is obtained as follows: Acquire equipment performance index data, obtain the performance consistency of each equipment performance index based on the collected data, and name the equipment performance indexes from Y1 compatibility acceptance index to Yc compatibility acceptance index according to the performance consistency value from small to large, thus obtaining the compatibility acceptance sorting queue. If the Y1 compatibility acceptance criterion is a memory performance criterion, then a memory performance compatibility analysis is performed on the target acceptance equipment group, and the Y1 compatibility acceptance criterion is judged to be abnormal in terms of performance compatibility based on the analysis results. According to the compatibility acceptance sorting queue, compatibility anomaly judgments are performed on the Y2 compatibility acceptance indicators to the Yc compatibility acceptance indicators respectively to obtain performance compatibility judgment data.

7. The AI-assisted automatic acceptance system for data center equipment performance system compatibility according to claim 6, characterized in that, A memory performance compatibility analysis was performed on the target acceptance equipment group, as detailed below: The target host devices and multiple target virtual devices included in the target acceptance device group are acquired, and an initial virtual device is selected from the acquired multiple target virtual devices. A memory compatibility acceptance program is created, and the time period during which the initial virtual device runs the memory compatibility acceptance program is set as the first memory performance acceptance time period. The memory access operations performed by the initial virtual device during the first memory performance acceptance period are acquired, and the acquired memory access operations are named N1 memory access operation to Nd memory access operation in chronological order of operation time. The operation time value corresponding to each memory access operation is obtained to obtain the N1 memory operation time value to the Nd memory operation time value. The memory access latency corresponding to different memory operation time values ​​is mapped to the memory access operation to obtain the N1 initial performance latency to the Nd initial performance latency.

8. The AI-assisted automatic acceptance system for data center equipment performance system compatibility according to claim 7, characterized in that, A memory performance compatibility analysis was performed on the target acceptance equipment group, as detailed below: Create a coordinate system for the first period of acceptance and mark the initial performance coordinate point N1 to the initial performance coordinate point Nd; After the first memory performance acceptance period ends, the target host device is kept running in the initial virtual device state so that it continues to repeat the memory compatibility acceptance program. At the same time, an A1 target virtual device is dynamically added to the target host device, and the A1 target virtual device is ensured to run the memory compatibility acceptance program synchronously with the initial virtual device. The period in which the memory compatibility acceptance program is run is marked as the second memory performance acceptance period. A memory performance compatibility assessment was conducted on the target acceptance equipment group during the second memory performance acceptance period. Based on the assessment results, the second memory performance acceptance period was divided into an effective compatibility period and a failure compatibility period.

9. The AI-assisted automatic acceptance system for data center equipment performance system compatibility according to claim 8, characterized in that, A memory performance compatibility analysis was performed on the target acceptance equipment group, as detailed below: If the second memory performance acceptance period is classified as a failure compatibility period, then the Y1 compatibility acceptance index is determined to have compatibility anomalies. If the second memory performance acceptance period is classified as a valid compatibility period, then after the second memory performance acceptance period ends, the target host device maintains the initial virtual device and the A1 target virtual device and continues to repeatedly execute the memory compatibility acceptance program. At the same time, the A2 target virtual device is dynamically added to the target host device, and it is ensured that the new device runs the memory compatibility acceptance program synchronously with the initial virtual device. The period in which the memory compatibility acceptance program runs is marked as the third memory performance acceptance period. Performance compatibility analysis is performed on the third memory performance acceptance period. If the third memory performance acceptance period is classified as a valid compatibility period, then the third memory performance acceptance period is obtained and performance compatibility analysis is performed. This process continues until all target virtual devices have been added. If any memory performance acceptance period is a failure compatibility period, then the Y1 compatibility acceptance indicator is determined to be an abnormal compatibility indicator. If any memory performance acceptance period is a failure compatibility period during the loop, then the Y1 compatibility acceptance indicator is determined to be an abnormal compatibility indicator. If none of these periods exist, then the Y1 compatibility acceptance indicator is determined to be a normal compatibility indicator.

10. The AI-assisted automatic acceptance system for data center equipment performance system compatibility according to claim 8, characterized in that, A memory performance compatibility assessment was conducted on the target acceptance equipment group during the second memory performance acceptance phase, as detailed below: In the first time period, the performance coordinate points of the new target virtual device are created in the time period coordinate system, resulting in the newly added performance coordinate points A1 to Ad. In the first period of the acceptance period coordinate system, the distance between the newly added performance coordinate point A1 and the initial performance coordinate point N1 is obtained to obtain the initial performance deviation of A1. Similarly, the distance between the newly added performance coordinate point Ad and the initial performance point Nd is obtained to obtain the initial performance deviation of Ad. The number of newly added performance coordinate points during the second memory performance acceptance period with an initial performance deviation of 0 is counted to obtain the new performance overlap. The average of the initial performance deviation from A1 to Ad is calculated to obtain the first parallel performance deviation characteristic value. The standard deviation of the initial performance deviation from A1 to Ad is calculated to obtain the second parallel performance deviation characteristic value. The virtual device memory conflict degree corresponding to the second memory performance acceptance period is obtained by calculating the newly added performance overlap, the first parallel performance deviation feature value, and the second parallel performance deviation feature value. Set a baseline range for virtual device memory conflict. If the virtual device memory conflict is within the baseline range, the second memory performance acceptance period is determined to be a valid compatibility period. If it is not, the second memory performance acceptance period is determined to be a invalid compatibility period.

Images