A method for hierarchical storage and retrieval of business data

Abstract

This invention discloses a method for hierarchical storage and retrieval of business data, relating to the field of hierarchical storage technology. The method includes: connecting call state information with the current storage execution position to generate a call state chain; performing call entropy perturbation analysis to identify high-entropy perturbation segments; locating and marking perturbed call segments as perturbed call segments; reconstructing stable call tracks; performing wavelet transform frequency domain layering on the stable call tracks; correcting the current storage execution position to obtain the basic storage level of the business data; defining the temporary call level of the business data; locating the call nodes of the business data; and forming abnormal call nodes. The basic storage level, temporary call level, and abnormal call nodes are then associated and connected to obtain hierarchical storage call relationships, and hierarchical storage parsing is performed to generate the call path of the business data. This invention, through call entropy perturbation analysis, avoids storage level adjustments and call path offsets caused by the mixture of abnormal fluctuations and stable call changes.

Description

Technical Field

[0001] This invention relates to the field of hierarchical storage technology, and in particular to a method for hierarchical storage and retrieval of business data. Background Technology

[0002] With the continuous development of cloud computing, distributed databases, and heterogeneous storage media, the hierarchical storage and retrieval of business data is gradually shifting from fixed media configuration to dynamic management based on access load. Existing methods typically collect access frequency, read / write latency, data lifecycle, capacity usage, and media cost to form a hot / cold attribute evaluation result. Based on the hot / cold attribute evaluation result, the data is configured to reside in cache, memory, solid-state storage, and capacity-based storage. During the data retrieval phase, business data location and retrieval are completed by relying on metadata indexing, directory mapping, cache hits, and cross-media routing. This approach is applied to cloud storage platforms, distributed file systems, and database management environments.

[0003] However, existing methods typically use access frequency, time interval, and preset popularity level to form a single stratification basis. Long-term stable access, short-term sudden access, and abnormal fluctuations are easily classified into the same change characteristics, making it difficult to accurately distinguish between continuous residence needs, temporary scheduling needs, and abnormal access interference. Mixed change characteristics can easily cause storage locations to be frequently adjusted with local fluctuations, resulting in unclear boundaries between long-term storage locations and temporary call locations. This causes access routes to deviate from the actual call order of business data, leading to inaccurate determination of storage levels and generation of call paths. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a hierarchical storage and retrieval method for business data to solve the problems of inaccurate determination of storage hierarchy and generation of retrieval path.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for hierarchical storage and retrieval of business data, comprising:

[0008] Collect the call status information of business data, connect the call status information with the current storage execution location, and generate a call status chain;

[0009] Call entropy perturbation analysis is performed on the call segments in the call state chain to obtain the call entropy perturbation value of each call segment. High entropy perturbation segments with abnormal fluctuations are identified from the call entropy perturbation values ​​and arranged as perturbation isolation tracks.

[0010] Based on the perturbation isolation track, the call segments with perturbations are found in the call state chain and marked as perturbation call segments. Each call segment not marked as a perturbation call segment is reconstructed into a stable call track. Wavelet transform frequency domain layering is performed on the stable call track to obtain low-frequency coefficient sequences and high-frequency coefficient sequences, which are then converted into low-frequency resident components and high-frequency resident components, respectively.

[0011] By using low-frequency resident components to correct the current storage execution position, the basic storage level of business data is obtained. By using high-frequency resident components to limit the temporary call level of business data, the call node of business data is located and assigned an abnormal call mark according to each disturbance call segment in the disturbance isolation track, thus forming an abnormal call node.

[0012] The basic storage level, temporary call level, and abnormal call node are associated and connected according to the call order of business data to obtain the hierarchical storage call relationship. Based on the hierarchical storage call relationship, the business data is parsed for hierarchical storage to generate the call path of the business data.

[0013] As a preferred embodiment of the hierarchical storage and retrieval method for business data described in this invention, the generation of the retrieval state chain specifically involves:

[0014] The call monitoring interface at the business data access entry point collects call status information of business data during the call process and obtains the current storage execution location of business data in each call status.

[0015] The call state information is connected to the current stored execution location in the order in which the calls occurred, to generate a call state chain.

[0016] As a preferred embodiment of the hierarchical storage and retrieval method for business data described in this invention, the arrangement is a perturbation isolation track, and the specific steps are as follows:

[0017] The call state chain is divided into several call segments, and the state difference arrangement within each call segment is performed. The change pattern of adjacent chain positions within the same call segment is extracted, and the chain positions with the same change pattern are grouped into a state arrangement group.

[0018] The proportion and distribution interval of each permutation state group within the call segment are statistically analyzed, and the proportion and distribution interval of the chain positions are discretely quantized to form the call entropy perturbation value.

[0019] By using the call entropy perturbation value, the rising and maintaining characteristics and the repeated fluctuation characteristics are screened out in the call state chain, and the corresponding call segments are identified as high entropy perturbation segments;

[0020] The high-entropy disturbance segments are arranged into disturbance isolation tracks according to the order in which they are invoked.

[0021] As a preferred embodiment of the hierarchical storage and retrieval method for business data described in this invention, the reconstruction of a stable retrieval track involves the following steps:

[0022] Using the call order of the high-entropy perturbation segment in the perturbation isolation track as the search condition, we search for call segments in the call state chain that have the same call order and the same call entropy perturbation value change pattern, and mark the found call segments as perturbation call segments.

[0023] The disturbed call segments are isolated and marked, and the discontinuous positions between adjacent call segments on both sides of the disturbed call segment are determined as link gaps;

[0024] By using link gaps to define connection boundaries and reconnecting the call segments that were not marked as disturbing call segments, a stable call track is formed.

[0025] As a preferred embodiment of the hierarchical storage and retrieval method for business data described in this invention, the specific steps of converting the data into low-frequency resident components and high-frequency resident components are as follows:

[0026] Trajectory sampling is performed on the call segments in the stable call track, and the sampling results are separated according to the connection boundary to form a stable sampling segment;

[0027] A non-decimation wavelet scaling function is used to perform scale smoothing on adjacent sampling points in a stable sampling segment, and the call components whose change amplitudes continuously decrease are organized into a low-frequency coefficient sequence.

[0028] Using the non-subtractive wavelet detail operator, the abrupt change position where the change direction is reversed between adjacent sampling points in the stable sampling segment is located, and the high-frequency detail coefficients of adjacent sampling points before and after the abrupt change position are calculated and arranged into a high-frequency coefficient sequence.

[0029] Scale merging is performed on the low-frequency coefficient sequence to extract low-frequency coefficient segments with consistent change direction and continuously decreasing amplitude in adjacent stable sampling segments. The low-frequency coefficient segments are then concatenated using the connection boundary. The concatenated low-frequency coefficients and the stable call track are classified as low-frequency resident components.

[0030] Based on the mutation position of the high-frequency coefficient sequence, the sudden call residence area in the stable call track is traced back, and the call segment state in the sudden call residence area is taken as the high-frequency residence component.

[0031] As a preferred embodiment of the hierarchical storage and retrieval method for business data described in this invention, the specific steps for obtaining the basic storage level of the business data are as follows:

[0032] Establish a call order index between the sampling location of low-frequency resident components and the current storage execution location, and concatenate storage level locations with the same call order index into a low-frequency level dwell segment;

[0033] For call segments that deviate from adjacent memory levels for a short period of time in low-frequency memory level dwell segments, perform a memory level fallback process, and retain the memory level positions of continuous dwell segments to form a corrected memory execution position;

[0034] The same storage level location repeatedly pointed to by multiple adjacent call segments in the corrected storage execution location is determined as the basic level position, and the level dwell range is limited by the start and end call segments of the low-frequency level dwell segment.

[0035] By concatenating the base level bits and the level residency range, we obtain the base storage level for business data.

[0036] As a preferred embodiment of the hierarchical storage and retrieval method for business data described in this invention, the specific steps of limiting the temporary retrieval level of business data through high-frequency resident components are as follows:

[0037] The burst call dwell area in the high-frequency dwell component is subjected to pulse clustering screening. Call segments are merged according to the adjacency relationship of the mutation position and the continuous fluctuation relationship of the call segment state to form a high-frequency dwell pulse cluster.

[0038] Using the concentrated segment of mutation locations within the high-frequency resident pulse cluster as the cutting boundary, the high-frequency resident pulse cluster is cut off at both ends, and the call interval from the starting call segment to the ending call segment after cutting is collected to form a temporary call trigger segment;

[0039] Map the order of calls to the temporary call triggering segment to the current storage execution location, and mark the storage level swing position caused by the temporary call triggering segment to form a temporary level free track;

[0040] The temporary level free track is excluded from the basic level, the storage level position that continuously overlaps with the basic level position is removed, and the storage level position that is located within the temporary residence range and deviates from the basic level position is screened out to form the temporary call position;

[0041] Encapsulate the temporary call bit and temporary dwell range into a temporary call layer for business data.

[0042] As a preferred embodiment of the hierarchical storage and retrieval method for business data described in this invention, the specific steps for forming an abnormal retrieval node are as follows:

[0043] The perturbation call segment in the perturbation isolation track is processed into a chain fingerprint, and the change pattern of the call entropy perturbation value is bound to the call segment state in the same segment to form a perturbation chain fingerprint.

[0044] The chain order of the disturbed call segment is arranged by the order in which the calls occur, and the fingerprint of the disturbed chain segment is written into the call status chain. Call nodes with the same business data identifier and whose chain order is within the coverage of the disturbed call segment are selected as candidate call node bands.

[0045] The candidate call node band is propagated and pruned to remove call nodes that are only connected to stable call tracks and have not written perturbation chain fingerprints, thus forming a perturbation propagation node band;

[0046] The calling node's direction is constrained by the chain position transmission relationship in the disturbance propagation node band, and the node band is converged to form an abnormal location node;

[0047] The abnormal location node is bound to the disturbed call segment by a chain mark, and the abnormal call mark is assigned to the abnormal location node to form an abnormal call node.

[0048] As a preferred embodiment of the hierarchical storage and retrieval method for business data described in this invention, the hierarchical storage and retrieval relationship is a hierarchical relationship between the basic storage level, the temporary retrieval level, and the abnormal retrieval node, recorded according to the retrieval order of the business data, including hierarchical jumps and abnormal isolation pointers.

[0049] As a preferred embodiment of the hierarchical storage and retrieval method for business data described in this invention, the specific steps for generating the retrieval path for business data are as follows:

[0050] Starting from the business data access entry point, the call nodes of the hierarchical storage call relationship are parsed sequentially according to the hierarchical jump direction, and abnormal call nodes are isolated;

[0051] Locate the associated position connecting the call node on the outgoing side of the abnormal call node from the call node on the incoming side of the abnormal call node;

[0052] The associated location is connected to the basic storage level and the temporary call level to generate the call path for business data.

[0053] The beneficial effects of this invention are as follows: By performing call entropy perturbation analysis, the discrete fluctuations of the internal state distribution of the call segment are quantified, achieving a structured separation of abnormal call components and stable call components. This avoids storage level adjustments and call path offsets caused by the mixing of abnormal fluctuations and stable call changes, reduces the interference of short-term surges and repeated fluctuations on subsequent frequency domain layering, avoids abnormal call states interfering with storage level determination, and clarifies the hierarchical boundaries between long-term resident requirements and temporary call requirements. This reduces invalid migration and repeated scheduling of business data between different storage levels, enhances the stability of hierarchical storage call relationships, the continuity of call paths, and the consistency of business data call results. The business data access entry reads the already generated business data call path during call state analysis, ensuring that the current business data call mainly executes business data call path reading and storage level access, reducing the impact of call entropy perturbation analysis and wavelet transform frequency domain layering on the business data call response time. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart illustrating the hierarchical storage and retrieval methods for business data.

[0056] Figure 2 A flowchart for creating a disturbance isolation track.

[0057] Figure 3 The flowchart shows the process of converting the components into low-frequency resident components and high-frequency resident components.

[0058] Figure 4 A flowchart for forming hierarchical storage and retrieval relationships. Detailed Implementation

[0059] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0060] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0061] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0062] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for hierarchical storage and retrieval of business data, including the following steps:

[0063] S1. Collect the call status information of business data, connect the call status information with the current storage execution location, and generate a call status chain.

[0064] S1.1. Collect call status information of business data during the call process through the call monitoring interface of the business data access entry, and obtain the current storage execution location of business data in each call status. Specifically,

[0065] The call monitoring interface of the business data access entry continuously records the call status information of the business data. The call status information is a continuous record of the status of the business data during a call, including call identifier, business data identifier, call occurrence time, call start status, call duration status, and call end status. The call end status includes call completion status and call interruption status.

[0066] For example, when e-commerce order business data is queried, the monitoring interface records the call identifier, business data mark, and query occurrence time of the order business data, and continuously records the start, duration, and completion of the query, thereby forming the call status information of the business data during the call process.

[0067] The business data access entry reads the business data identifier carried by the business data, assigns a call identifier to this call, and then retrieves the storage scheduling record using the business data identifier to read the storage level identifier and storage address of the business data.

[0068] When business data is first stored in the storage tier, the storage scheduling record records the storage tier identifier and storage address; when the storage location of business data changes or migration is completed, the storage scheduling record updates the storage tier identifier and storage address.

[0069] When the call monitoring interface forms the call start state, call continuation state, and call end state, it records the actual storage address accessed by the call. Based on the association between the storage address and the storage level identifier in the storage scheduling record, it determines the current storage execution location of the business data in each call state.

[0070] After the business data has completed the change and migration of its storage location, a new storage level identifier and storage address are formed in the storage scheduling record. When the call monitoring interface records the call start state, call duration state and call end state, it records the storage address actually accessed by this call, and determines the current storage execution location of the business data in each call state based on the association between the storage address and the storage level identifier in the storage scheduling record.

[0071] The storage scheduling record uses the business data identifier as the retrieval condition, records the storage level location where the business data is currently undertaking the call operation, and updates it as the storage location of the business data changes and as the migration results change.

[0072] For example, in the call start and call continuation states, the actual access is to the cache storage level, and the current storage execution position is recorded as the cache storage level. In the call end state, the operation is completed at the disk storage level, and the current storage execution position is recorded as the disk storage level. The complete content of the order business data is located at the disk storage level, while the recently accessed copy of the order business data is located at the cache storage level. When the call operation is completed at the cache storage level, the current storage execution position is the cache storage level; when the call operation is completed at the disk storage level, the current storage execution position is the disk storage level. The call monitoring interface uses the business data identifier to retrieve the storage scheduling record and records the determined current storage execution position under the current call identifier, thus associating the call status information with the current storage execution position and obtaining the current storage execution position of the business data in each call state.

[0073] S1.2. Connect the call state information with the current stored execution location in the order in which the calls occurred to generate a call state chain. Specifically, The call status information in the same call process is collected by the call identifier, and the call start status, call duration status and call end status are arranged according to the call occurrence time. Each call status is combined with the current storage execution position obtained at the time of call status recording to form a chain bit, and the chain bits are connected in sequence according to the order of the call occurrence to generate a call status chain.

[0074] When a business data call begins, the business data access entry point uses the business data identifier to read the already generated business data call path and accesses the storage level location according to the hierarchical jump direction defined by the business data call path; if no business data call path has been generated for the first time, the business data access entry point uses the current storage execution location recorded in the storage scheduling record to complete the business data call.

[0075] The call monitoring interface records call status information and the current storage execution position as business data is called. After the call status information under the same call identifier forms a call end state, the continuous chain from the call start state to the call end state is determined as a call segment. After multiple call segments are formed, they are connected in the order in which the calls occurred, and call entropy perturbation analysis, wavelet transform frequency domain layering, hierarchical storage call relationship formation, and business data call path generation are performed.

[0076] The current business data call is completed when the call end state is formed; after the call end state is formed, the call entropy perturbation analysis, wavelet transform frequency domain layering, hierarchical storage call relationship formation, and business data call path generation are performed to form a business data call path for subsequent business data calls to read.

[0077] S2. Perform call entropy perturbation analysis on the call segments in the call state chain to obtain the call entropy perturbation value of each call segment. Identify high-entropy perturbation segments with abnormal fluctuations from the call entropy perturbation values ​​and arrange them as perturbation isolation tracks.

[0078] S2.1 Divide the call state chain into several call segments, perform intra-chain state difference arrangement on each call segment, extract the change pattern of adjacent chain positions within the same call segment, and group chain positions with the same change pattern into a permutation state group. Specifically,

[0079] The call identifier is used to distinguish consecutive call processes in the call state chain, and consecutive chain bits under the same call identifier that continue from the call start state to the call end state are classified as a call segment.

[0080] Within each call segment, the call state and current storage execution position of adjacent chain positions are read sequentially. The adjacent transition relationships of call states and the maintenance and change relationships of current storage execution positions are recorded, and the number of chain positions with the same call state and the same current storage execution position that are continuously maintained is counted. The chain positions are then concatenated and encoded according to a fixed field order of adjacent transition relationships of call states, maintenance and change relationships of current storage execution positions, and the number of continuously maintained chain positions, forming the change pattern of adjacent chain positions. This change pattern is then marked on the subsequent chain position. Chain positions with the same call state transition relationship, the same current storage execution position change relationship, and the same number of continuously maintained chain positions are grouped into the same permutation state group; chain positions with differences in any field are grouped into different permutation state groups.

[0081] S2.2 Calculate the proportion and distribution interval of each permutation state group within the call segment, and then discretize and quantize the proportion and distribution interval to form the call entropy perturbation value. Specifically,

[0082] Statistics call segment Total number of chain positions and permutation state group The number of chain bits included, arranged in order of state groups In the calling segment The degree of occupancy of chain positions within a chain determines the chain position percentage. Arrange the same state group according to the order in which the calls occur. Each chain position within the chain is recorded, and the corresponding chain position number is recorded. The interval between adjacent chain position numbers is normalized to form a discrete distribution interval. Combined with the proportion of chain positions and the discrete quantity of the distribution interval Form a joint quantity with proportional distribution, and then process the calling segment. The joint quantity of the proportion distribution of all permutation state groups is normalized and entropy quantized to form the call entropy perturbation value, as shown in the formula:

[0083] ;

[0084] In the formula, Indicates the calling segment The call entropy perturbation value, Represents the permutation state group In the calling segment The proportion of chain bits within the chain, Represents the permutation state group In the calling segment The discrete quantity of the distribution interval within. Indicates the calling segment The total number of internal permutation state groups. Indicates the segment number to be called. Indicates the sequence number of the state group. Indicates the permutation state group number used for normalization summation.

[0085] When the permutation state group contains only one chain position, the discrete value of the distribution interval is recorded as zero; when the call segment contains only one permutation state group, the call entropy perturbation value is recorded as zero.

[0086] It should be noted that the method for normalizing the interval between adjacent chain positions is the maximum chain position span benchmark normalization. Specifically, the complete chain position span covered by the first chain position to the last chain position of the calling segment is used as a unified benchmark. The interval between each adjacent chain position in the arrangement state group is converted into a ratio relative to the complete chain position span and uniformly limited to between zero and one, forming a discrete quantity of distribution interval.

[0087] S2.3. Using the call entropy perturbation value, select the rising and maintaining characteristics and the repeated fluctuation characteristics in the call state chain, and determine the corresponding call segment as the high entropy perturbation segment. Specifically,

[0088] Arrange the call entropy perturbation values ​​of each call segment in the order of call occurrence, and record the direction of change of the call entropy perturbation values ​​of adjacent call segments.

[0089] The call entropy perturbation value increases segment by segment in consecutive adjacent call segments, which is marked as an increase-and-hold feature. The call entropy perturbation value increases compared to the previous call segment and remains no lower than the increased call entropy perturbation value in the immediately following consecutive call segments, which is also marked as an increase-and-hold feature. The call entropy perturbation value increases and decreases alternately in consecutive adjacent call segments, and the adjacent change direction reverses more than twice in a row, which is marked as a repeated fluctuation feature.

[0090] The continuous call segments that participate in forming the rise-and-hold characteristic and the continuous call segments that participate in forming the repeated fluctuation characteristic are identified as high-entropy perturbation segments, while the call segments that do not participate in forming the rise-and-hold characteristic and the repeated fluctuation characteristic are not identified as high-entropy perturbation segments.

[0091] S2.4 Arrange the high-entropy perturbation segments into perturbation isolation tracks according to the order in which they are invoked, specifically as follows: Read the call occurrence time recorded in the starting chain position of each high-entropy disturbance segment, and determine the arrangement position of each high-entropy disturbance segment according to the call occurrence time from earliest to latest. High-entropy disturbance segments with the same call occurrence time are arranged according to the order of the chain positions in the call state chain. At the same time, the starting chain position, ending chain position, call identifier and call entropy disturbance value change pattern of each high-entropy disturbance segment are retained. High-entropy disturbance segments that are separated in the call state chain are not merged, and each high-entropy disturbance segment is connected in series according to the arrangement position to form a disturbance isolation track.

[0092] It should be noted that the call entropy perturbation analysis takes the call segment that has already reached the end of the call state as the processing object, and uses the changes in the call entropy perturbation values ​​of consecutive adjacent call segments in the call state chain to form the rising and holding characteristics and the repeated fluctuation characteristics.

[0093] During the execution of the entropy perturbation analysis, newly occurring business data calls read the already generated business data call paths based on the business data identifiers, and access the storage level location according to the hierarchical jump direction defined by the business data call paths.

[0094] After a new business data call reaches the end state, the continuous chain between the start state and the end state is divided into a new call segment, and then assigned to the subsequent call state chain according to the order in which the calls occurred.

[0095] S3. Based on the perturbation isolation track, find the perturbation call segment in the call state chain and mark it as a perturbation call segment. Reconstruct each call segment that is not marked as a perturbation call segment into a stable call track. Perform wavelet transform frequency domain layering on the stable call track to obtain low-frequency coefficient sequence and high-frequency coefficient sequence, and convert them into low-frequency resident component and high-frequency resident component, respectively.

[0096] S3.1. Based on the perturbation isolation track, find the call segments with perturbations in the call state chain and mark them as perturbation call segments. Reconstruct each call segment not marked as a perturbation call segment into a stable call track. Specifically,

[0097] S3.1.1 Using the call occurrence order of the high-entropy perturbation segment in the perturbation isolation track as the search condition, search for call segments in the call state chain that have the same call occurrence order and the same call entropy perturbation value change pattern, and mark the found call segments as perturbation call segments. Specifically,

[0098] Read the call occurrence order and call entropy perturbation value change pattern of the high-entropy perturbation segment in the perturbation isolation track. Starting from the first call segment in the call state chain, read the call occurrence order sequentially. If the call occurrence order does not reach the call occurrence order of the high-entropy perturbation segment, continue reading the next call segment. If the call occurrence order exceeds the call occurrence order of the high-entropy perturbation segment, end the current search.

[0099] After locating call segments with consistent call order, read the call entropy perturbation values ​​of the call segments with consistent call order and adjacent call segments, and organize the ascending and descending order and continuous state of the call entropy perturbation values; call segments with consistent change patterns of call entropy perturbation values ​​are marked as perturbed call segments, while call segments with inconsistent change patterns of call entropy perturbation values ​​remain unmarked.

[0100] S3.1.2. Isolate and mark the disturbed call segment, and determine the discontinuous position between adjacent call segments on both sides of the disturbed call segment as the link gap, specifically,

[0101] Isolation markers are set at the start and end links of the disturbed call segment to preserve the original link order, call occurrence time and link position of the disturbed call segment, and to stop including the disturbed call segment in the continuous sampling range of the stable call track.

[0102] The connection point between the termination chain of the preceding adjacent call segment and the starting chain of the disturbance call segment is recorded as the preceding discontinuous position. The connection point between the termination chain of the disturbance call segment and the starting chain of the following adjacent call segment is recorded as the following discontinuous position. The preceding discontinuous position, the call time interval covered by the disturbance call segment, and the following discontinuous position are collectively determined as the link gap.

[0103] S3.1.3. Utilize link gaps to define connection boundaries and reconnect all call segments not marked as disturbing call segments to form a stable call track. Specifically,

[0104] The termination link of the unmarked call segment before the link gap is determined as the starting point of the connection boundary, and the starting link of the unmarked call segment after the link gap is determined as the ending point of the connection boundary.

[0105] When there are consecutive adjacent disturbance call segments, the call time intervals covered by consecutive adjacent disturbance call segments are recorded as the same link gap, and the termination link of the unmarked call segment before the foremost disturbance call segment is taken as the starting point of the connection boundary, and the starting link of the unmarked call segment after the last disturbance call segment is taken as the ending point of the connection boundary.

[0106] Unmarked call segments are arranged according to their original call occurrence time, retaining the original call occurrence time and chain position number for each chain position. Unmarked call segments on both sides of the link gap are only established with sequential pointing, without establishing continuous sampling relationships. The unmarked call segments, connection boundaries, and link gaps are collectively orchestrated into a stable call track. Only single-sided connection boundaries are retained for the link gaps at the beginning and end of the call state chain.

[0107] S3.2. Perform wavelet transform frequency domain layering on the stable call track to obtain low-frequency coefficient sequences and high-frequency coefficient sequences, and then convert them into low-frequency resident components and high-frequency resident components, respectively. Specifically,

[0108] S3.2.1. Sampling is performed on the call segments in the stable call track, and the sampling results are separated according to the connection boundary to form a stable sampling segment. Specifically,

[0109] Stable call tracks are separated by link gaps and connection boundaries, and call segments that occur in consecutive time and do not cross link gaps are grouped into the same stable sampling segment.

[0110] The first sampling point is the time when the stable sampling segment begins to be called. Subsequent sampling points are set according to a uniform time interval. The call status and current storage execution position of the nearest chain bit before each sampling point are written into the corresponding sampling point to form sampling results at equal time intervals.

[0111] The sampling results are not rewritten for the call time interval covered by the link gap, the link positions on both sides of the link gap are not used as adjacent sampling points, and each stable sampling segment is subjected to subsequent wavelet transform frequency domain layering.

[0112] S3.2.2. A non-decimation wavelet scaling function is used to smooth the scale of adjacent sampling points in the stable sampling segment, and the call components with continuously decreasing amplitude changes are organized into a low-frequency coefficient sequence, as shown in the formula.

[0113] The call state sequence, termination marker, and storage level identifier in the stable sample segment are arranged into sample component sequences. Low-pass filter coefficients of the wavelet basis scaling function are selected, and low-pass filter coefficient sequences of different scales are formed by expanding the spacing between the low-pass filter coefficients. The low-pass filter coefficients cover adjacent sample points in the sample component sequences. Scale convolution is performed without deleting sample points. Mirror extension is used when sample positions are missing at the beginning and end of the stable sample segment. The scale smoothing coefficient satisfies the following:

[0114] ;

[0115] In the formula, Indicates the scale smoothing coefficient; Indicates the sampling component In scale and sampling location The scale smoothing coefficient at the location; Indicates the sampling component number. Indicates the scale number. Indicates the sampling location number; Indicates the index of the low-pass filter coefficient. This indicates that the summation starts from the first low-pass filter coefficient. Representing scale Number of low-pass filter coefficients Indicates the sequence number of the low-pass filter coefficient at the end. Representing scale The next number is The low-pass filter coefficients, Indicates the previous scale. This indicates the position offset of the sampling location as a result of the low-pass filter coefficient sequence; the equal sign indicates that the values ​​on both sides of the equal sign are the same, and the scale smoothing coefficient under the initial scale uses the original sampling result.

[0116] For the same stable sampling segment, the same sampling component, and the same scale, scale smoothing coefficients are arranged according to the sampling location, forming scale difference values ​​between adjacent sampling locations. The median of all scale difference values ​​is used as the difference center, forming the absolute deviation of each scale difference value relative to the difference center. The median of all absolute deviations is determined as the stationarity tolerance. When the stationarity tolerance is zero, only adjacent sampling locations with the same scale smoothing coefficient are recorded as stationary directions.

[0117] When the scale difference value is higher than the stationary tolerance, it is recorded as the increasing direction; when the scale difference value is lower than the negative limit corresponding to the stationary tolerance, it is recorded as the decreasing direction; when the absolute value of the scale difference value does not exceed the stationary tolerance, it is recorded as the stationary direction.

[0118] Starting with the first non-stationary scale difference value, subsequent scale difference values ​​are read sequentially. When the direction of change of the preceding and following scale difference values ​​is the same, and the absolute value of the subsequent scale difference value is not higher than the upper limit of the amplitude jointly defined by the absolute value of the previous scale difference value and the stationarity tolerance, the corresponding sampling position is assigned to the same low-frequency coefficient segment. When the absolute value of the subsequent scale difference value exceeds the upper limit of the amplitude, the current low-frequency coefficient segment is terminated at the previous sampling position. When the direction of the preceding and following scale difference values ​​reverses and the absolute values ​​both exceed the stationarity tolerance, the current low-frequency coefficient segment is terminated at the previous sampling position. When the subsequent scale difference value enters the stationary direction, the current sampling position is assigned to the low-frequency coefficient segment and the current low-frequency coefficient segment is terminated.

[0119] The low-frequency coefficient segments are arranged into a low-frequency coefficient sequence according to their sampling positions.

[0120] It should be noted that the wavelet transform frequency domain layering is limited to the stable sampling segments already formed in the stable call track.

[0121] The stable sampling segment consists of the call segment that has reached the end of the call state. Each stable sampling segment performs scale smoothing and high-frequency detail processing to form a low-frequency coefficient sequence and a high-frequency coefficient sequence.

[0122] During the wavelet transform frequency domain layered execution, the business data access entry reads the already generated business data call path according to the business data identifier, and completes the newly generated business data call according to the business data call path.

[0123] The low-frequency resident components and high-frequency resident components formed by wavelet transform frequency domain layering are used to determine the basic storage level and temporary call level, and further generate business data call paths for subsequent business data calls and readings.

[0124] S3.2.3. Using the non-decimation wavelet detail operator, locate the abrupt change position where the direction of change reverses between adjacent sampling points in the stable sampling segment, calculate the high-frequency detail coefficients of adjacent sampling points before and after the abrupt change position, and arrange them into a high-frequency coefficient sequence. Specifically,

[0125] The scaling coefficients formed by the call state sequence, termination marker, and storage level identifier in the stable sample segment at the previous scale are used as the processing objects. High-pass filter coefficients with the same wavelet basis as the non-decimated wavelet scaling function are selected, and zero-coefficient interval expansion is used to form the scale. The non-extractable wavelet detail operator.

[0126] Three consecutive adjacent sampling points are read according to the sampling location, and the scale difference values ​​of the preceding and following adjacent intervals are read for each sampling component. If the preceding scale difference is in the increasing direction and the following scale difference is in the decreasing direction, and the absolute values ​​of both scale difference values ​​exceed the stationarity tolerance, the central sampling location is determined as the abrupt change location. Similarly, if the preceding scale difference is in the decreasing direction and the following scale difference is in the increasing direction, and the absolute values ​​of both scale difference values ​​exceed the stationarity tolerance, the central sampling location is also determined as the abrupt change location. If at least one sampling component satisfies the direction reversal condition, the central sampling location and the sampling component that satisfies the condition are retained; if the direction reversal condition is not met, the central sampling location is not determined as the abrupt change location. The sampling locations of the sampling points adjacent to the abrupt change location before and after it are respectively taken as... The high-frequency detail coefficients are obtained using the following formula:

[0127] ;

[0128] In the formula, Indicates high-frequency detail coefficients; Indicates the sampling component In scale and sampling location High-frequency detail coefficients at the location; Indicates the sample component number; Indicates the current scale number; Indicates the sampling location sequence; the equals sign indicates that the results on both sides of the equals sign are the same; Indicates the index of the high-pass filter coefficient; This indicates that the summation starts from the first high-pass filter coefficient; Indicates the number of high-pass filter coefficients; Representing scale The number of high-pass filter coefficients below; subscript This indicates the scale to which the number of high-pass filter coefficients belong; Representing scale The high-pass filter coefficients at the bottom end; Indicates the non-decimated wavelet high-pass filter coefficients; Representing scale The next number is Non-decimated wavelet high-pass filter coefficients; subscript Indicates the scale to which the non-decimated wavelet high-pass filter coefficients belong; (The text in parentheses is incomplete and cannot be translated.) Indicates the position of the non-decimated wavelet high-pass filter coefficients; Indicates the scale smoothing coefficient; Indicates the sampling component The scale smoothing coefficient at the previous scale and the offset sampling position; Indicates the previous scale of the current scale; This indicates the position offset of the current sampling position based on the high-pass filter coefficient sequence.

[0129] When the beginning and end of a stable sampling segment lack coverage, mirror extension is used. The high-frequency detail coefficients of the adjacent sampling points before the abrupt change position, the abrupt change position, and the adjacent sampling points after the abrupt change position are calculated using the non-subtractive wavelet detail operator and arranged into a high-frequency coefficient sequence according to the sampling position.

[0130] S3.2.4. Scale merging is performed on the low-frequency coefficient sequence to extract low-frequency coefficient segments with consistent change directions and continuously decreasing amplitudes from adjacent stable sampling segments. These segments are then concatenated using connection boundaries. The concatenated low-frequency coefficients and stable call tracks are classified as low-frequency resident components. Specifically,

[0131] For the same stable sampling segment and the same sampling component, the low-frequency coefficient segments and scale difference values ​​at different scales are read in ascending order of scale, using the sampling position as the index.

[0132] Starting from the smallest scale, scale difference values ​​at the same sampling location at adjacent scales are continuously compared. If the scale difference value of the current scale is in the same direction as the scale difference value of the previous scale, and the absolute value of the scale difference value of the current scale does not exceed the upper limit of the amplitude formed by relaxing the absolute value of the scale difference value of the previous scale by one current scale stability tolerance upwards, the scale smoothing coefficient corresponding to the current scale is retained.

[0133] The scale difference value at the current scale is in the opposite direction to the scale difference value at the previous scale. The scale difference value at the current scale enters a stationary direction. When the absolute value of the scale difference value at the current scale exceeds the upper limit of the amplitude, scale merging at the current sampling position is stopped.

[0134] If the same sampling location satisfies the scale merging condition at least in two consecutive scales, the retained results corresponding to the same sampling location will be included in the low-frequency permanent candidate location; sampling locations that satisfy the scale merging condition only in a single scale will not be included in the low-frequency permanent candidate location.

[0135] Consecutive adjacent low-frequency resident candidate positions are grouped into low-frequency coefficient segments. The chain positions retained by the sampling positions are used to associate the low-frequency coefficient segments with the corresponding chain positions in the stable call track to form low-frequency resident components.

[0136] Different stable sampling segments are scaled separately. Low-frequency coefficient segments located on both sides of the link gap remain separated, are not connected across the link gap, and are not merged using the connection boundary.

[0137] When the low-frequency coefficient sequence contains only one scale, scale merging is not performed, and the low-frequency coefficient sequence is retained until the scale smoothing coefficients for the next scale are formed before reprocessing.

[0138] S3.2.5. Based on the abrupt change positions of the high-frequency coefficient sequence, backtrack the burst call residency area in the stable call track, and use the call segment state within the burst call residency area as the high-frequency residency component. Specifically,

[0139] For the same stable sampling segment, the same sampling component, and the same scale, the high-frequency detail coefficients corresponding to all sampling positions of the stable sampling segment are arranged according to the sampling position.

[0140] All high-frequency detail coefficients are arranged from low to high values, and the center of the high-frequency detail coefficients is determined by centering the values. The absolute distance of each high-frequency detail coefficient from the center of the high-frequency detail coefficients is read, and they are arranged from low to high absolute distances. The high-frequency detail stability tolerance is determined by centering the values.

[0141] When the high-frequency detail stationarity tolerance is zero and all high-frequency detail coefficients are the same, all sampling positions are recorded as high-frequency stationar positions; when the high-frequency detail stationarity tolerance is zero and different high-frequency detail coefficients exist, the minimum value among all non-zero absolute distances is determined as the high-frequency detail stationarity tolerance.

[0142] When the absolute distance between the high-frequency detail coefficient and the center of the high-frequency detail coefficient does not exceed the high-frequency detail stability tolerance, the corresponding sampling position is recorded as the high-frequency stable position; when it exceeds the high-frequency detail stability tolerance, the corresponding sampling position is recorded as the high-frequency fluctuation position.

[0143] Using the mutation location as the starting point for backtracking, high-frequency detail coefficients are read from each sampling location in the direction of earlier and later call times, respectively.

[0144] If, along any backtracking direction, the absolute value of the high-frequency detail coefficient at the current sampling position does not exceed the upper limit of the amplitude formed by widening a high-frequency detail stability tolerance upwards from the absolute value of the high-frequency detail coefficient at the previous sampling position, backtracking continues in the same direction.

[0145] When the absolute value of the high-frequency detail coefficient at the current sampling position exceeds the amplitude limit, the previous sampling position is determined as the backtracking stop position.

[0146] When two consecutive sampling positions are both recorded as high-frequency stable positions, the high-frequency stable position closest to the abrupt change position is determined as the backtracking stop position. If only a single high-frequency stable position appears, and the sampling position immediately re-enters a high-frequency fluctuation position, backtracking continues, without terminating the backtracking using a single high-frequency stable position.

[0147] If the backtracking reaches the beginning and end of a stable sampling segment without finding a backtracking stop position, the beginning and end of the stable sampling segment are determined as the backtracking stop positions in the corresponding directions.

[0148] The continuous sampling positions between the backtracking stop position in the direction where the call occurred earlier and the backtracking stop position in the direction where the call occurred later are determined as the burst call dwell area.

[0149] When sampling locations overlap within the same stable sampling segment, the same sampling component, and the same scale of the burst call residence area, the overlapping portion and the continuous sampling locations on both sides of the overlapping portion are merged; when no sampling locations overlap, the burst call residence areas are kept separate from each other.

[0150] By utilizing the sequential positions of the chain bits retained at each sampling location within the burst call residency area, the corresponding chain bits in the stable call track are traced back, and the call status and current storage execution position of the call segment to which the corresponding chain bit belongs are retained, forming a high-frequency residency component.

[0151] When there are no abrupt change locations in the high-frequency coefficient sequence, no sudden call dwell area is formed.

[0152] S4. Use low-frequency resident components to correct the current storage execution position, obtain the basic storage level of business data, limit the temporary call level of business data through high-frequency resident components, locate the call node of business data and assign abnormal call mark according to each disturbance call segment in the disturbance isolation track, and form an abnormal call node.

[0153] S4.1. Utilize low-frequency resident components to correct the current storage execution location and obtain the basic storage level of the business data. Specifically,

[0154] S4.1.1 Establish a call order index between the sampling location of the low-frequency resident component and the current storage execution location, and concatenate storage level locations with the same call order index into a low-frequency level dwell segment, specifically,

[0155] Starting from the sampling position of each low-frequency coefficient in the low-frequency resident component, the current storage execution position is located in the stable call track by using the chain position reserved by the sampling position. The sampling position of the low-frequency resident component and the current storage execution position are assigned the same call sequence number to form a call sequence index.

[0156] The current execution position is arranged according to the call sequence number. If consecutive call sequence numbers point to the same storage level position, they are connected into the same low-frequency level dwell segment. If the storage level position changes, the current low-frequency level dwell segment is terminated and the next low-frequency level dwell segment is connected from the changed storage level position.

[0157] S4.1.2. For call segments that briefly deviate from adjacent memory levels within low-frequency memory level dwell segments, perform memory level fallback processing, and retain the continuously dwelling memory level positions to form the corrected memory execution position. Specifically,

[0158] According to the call order, check the storage level position in the low-frequency level dwell segment. If the continuous call segment deviates from the storage level position of the previous continuous dwell segment, and the end of the deviated call segment points back to the same storage level position as the previous continuous dwell segment, and the number of deviated call segments is less than the number of the previous continuous dwell call segments and less than the number of the subsequent continuous dwell call segments, it is determined to be a short-term deviated call segment.

[0159] Replace the storage level location of the short-term offset call segment with the storage level location that is pointed to by both the preceding and following consecutive stay call segments, and then connect the preceding consecutive stay call segment, the replaced short-term offset call segment, and the following consecutive stay call segment.

[0160] The deviated call segments that point to different storage level positions from the consecutive call segments on the front and the consecutive call segments on the back retain their original storage level positions. The number of deviated call segments that is not less than the number of consecutive call segments on either side also retains their original storage level positions. The storage level positions of the consecutive calls after the completion of the level fallback process are retained according to the call order, forming the corrected storage execution position.

[0161] S4.1.3. The same storage level location repeatedly pointed to by multiple adjacent call segments in the corrected storage execution location is determined as the base level location. The level residency range is limited by the start and end call segments of the low-frequency level dwell segment. Specifically,

[0162] Read the corrected storage execution location according to the call order. If consecutive adjacent call segments repeatedly point to the same storage level location, determine the same storage level location as the base level location.

[0163] The first call segment that repeatedly points to the same storage level location is determined as the starting call segment of the low-frequency level dwell segment, and the last call segment that repeatedly points to the same storage level location is determined as the ending call segment of the low-frequency level dwell segment. The call sequence segment covered by the starting call segment to the ending call segment is determined as the level dwell range. If the storage level location changes, the current level dwell range ends, and the base level position and level dwell range are re-determined from the changed storage level location.

[0164] S4.1.4. Connect the basic level bits and the level residency range to obtain the basic storage level of the business data, specifically,

[0165] Using business data identifiers as the association benchmark, the basic level bits are combined with the start call segment, end call segment, and call order segment in the level dwell range. This allows the basic level bits to limit the storage level where business data resides continuously, and the level dwell range to limit the call segment where the basic level bits continuously undertake call operations.

[0166] When the same basic level bits are distributed in mutually separate hierarchical residence ranges, each hierarchical residence range is retained separately, and different basic level bits are combined with their respective hierarchical residence ranges to form the basic storage hierarchy of business data.

[0167] S4.2. Temporary call levels for business data are limited by high-frequency residing components. Specifically,

[0168] S4.2.1. For the burst call dwell area in the high-frequency dwell component, pulse clustering is performed. Call segments are merged according to the adjacency of the abrupt change position and the continuous fluctuation relationship of the call segment state to form a high-frequency dwell pulse cluster. Specifically,

[0169] The mutation positions in the high-frequency resident components are arranged according to the sampling position, and the call segment covered by the mutation position is located by using the chain position reserved by the sampling position.

[0170] Starting from the first mutation position, read the next mutation position. The two mutation positions are located in the same call segment or adjacent call segments that are directly connected end to end. The state of the call segment within the coverage area of ​​the two mutation positions is constantly changing and they are classified into the same pulse cluster.

[0171] If a mutation location crosses a call segment that has not undergone a state change, the current pulse cluster is terminated and the pulse cluster is regrouped from the subsequent mutation location; if adjacent mutation locations do not cover consecutive call segments, they remain separated; call segments within the same pulse cluster are connected according to the order in which they occur, forming a high-frequency resident pulse cluster.

[0172] S4.2.2. Using the concentrated segment of mutation locations within the high-frequency resident pulse cluster as the cutting boundary, the high-frequency resident pulse cluster is cut off at both ends, and the call interval from the cut start call segment to the end call segment is collected to form a temporary call trigger segment, specifically,

[0173] The mutation positions within the high-frequency resident pulse cluster are arranged according to the order of call occurrence. Consecutive adjacent mutation positions that are not separated by non-mutation call segments are grouped into the same mutation position set segment. The starting chain position of the call segment containing the earliest mutation position is selected as the first-end trimming boundary, and the ending chain position of the call segment containing the latest mutation position is selected as the last-end trimming boundary. Call segments outside the first-end trimming boundary and outside the last-end trimming boundary that do not participate in the mutation position set segment are removed. The continuous call segments covered by the first-end trimming boundary to the last-end trimming boundary are retained and grouped into call intervals according to the order of call occurrence to form temporary call trigger segments.

[0174] S4.2.3 Map the call sequence of the temporary call triggering segment to the current memory execution location, and mark the memory level swing position caused by the temporary call triggering segment, forming a temporary level free track, specifically,

[0175] Using the order of calls within the temporary call trigger segment as an index, locate the chain position with the same call order in the call state chain, and read the current stored execution position recorded in the chain position.

[0176] The current storage execution position is arranged according to the order of the calls. If the storage level position of consecutive call segments changes and the subsequent call segment returns to the storage level position before the change, the call segment covered by the first switch chain position to the return chain position is marked as the storage level swing position.

[0177] If a continuous call segment remains at the same storage level, it is not marked as a storage level swing position. Similarly, if a segment remains at the changed storage level after a storage level position switch, it is also not marked as a storage level swing position. The storage level swing positions are connected in the order of the call occurrence to form a temporary level free track.

[0178] S4.2.4. Perform basic level exclusion on temporary level free tracks, remove storage level positions that continuously overlap with basic level positions, and filter out storage level positions that are within the temporary residence range and deviate from the basic level positions to form temporary call positions. Specifically,

[0179] The temporary residence range is defined by the start and end call segments of the temporary call trigger segment, and the swing position of each storage level in the temporary level free track is read according to the order of call occurrence.

[0180] If the storage level swing position points to the base level position in consecutive call segments, it is excluded from the temporary level free track. When the storage level swing position switches from the base level position to a storage level position different from the base level position, the call segment where the switch occurs is determined as the starting call segment of the deviated segment, and subsequent call segments are read continuously according to the order of call occurrence.

[0181] When a storage level location is redirected to a base level location, the call segment preceding the redirection to the base level location is designated as the terminating call segment of the offset segment. The starting call segment, terminating call segment, call order segment, and storage level location corresponding to the offset segment are retained. The call segment redirected to the base level location is preserved in the original call order as an interval between adjacent offset segments and is not included in the temporary call segment.

[0182] When the storage level location deviates from the base level position again after it has been redirected to the base level position, a new deviation segment is established from the call segment that has deviated from the base level position again. The previous and subsequent deviation segments are retained separately, and no connection is made across the call segment that points to the base level position.

[0183] After the basic level is excluded, the call segments whose call order is within the temporary residence range and whose storage level position is different from the basic level position are retained. Storage level positions whose call order exceeds the temporary residence range are not included in the filtering results. Call segments whose storage level position is the same as the basic level position are also not included in the filtering results. The retained storage level positions are collected according to the order of call occurrence to form temporary call positions.

[0184] S4.2.5. Encapsulate the temporary call bit and temporary dwell range into a temporary call layer for business data, specifically,

[0185] Using the business data identifier as the association benchmark, the temporary call bit is combined with the start call segment, end call segment and call order segment in the temporary residence range. This makes the temporary call bit limit the storage level location used by the business data in the temporary residence range, and the temporary residence range limit the call segment in which the temporary call bit participates in the call.

[0186] When the same temporary call bit is distributed in separate temporary residence ranges, each temporary residence range is retained. Different temporary call bits are combined with their corresponding temporary residence ranges to form a temporary call hierarchy for business data.

[0187] S4.3. Based on each disturbance call segment in the disturbance isolation track, locate the call nodes of business data and assign abnormal call markers to form abnormal call nodes. Specifically,

[0188] S4.3.1. Perform chain-bit fingerprinting on the perturbation call segment in the perturbation isolation track, and bind the change pattern of the call entropy perturbation value to the call segment state in the same segment to form a perturbation chain-bit fingerprint. Specifically,

[0189] Arrange the chain bits within the perturbation call segment according to their sequential position, and combine the chain bit sequence number, call identifier, business data identifier, call status, and current storage execution position into a chain bit feature string. The coverage area of ​​the chain bit feature string is limited by the starting chain bit and the ending chain bit of the perturbation call segment.

[0190] The call entropy perturbation value change pattern is linked to the call start state, call duration state, and call end state within the coverage segment to the same call identifier, so that the call entropy perturbation value change pattern is only associated with the call segment state of the perturbation call segment to which it belongs, and does not cross adjacent perturbation call segments; the chain bit feature rate, coverage segment, call entropy perturbation value change pattern, and call segment state together form the perturbation chain bit fingerprint.

[0191] S4.3.2. Arrange the chain order of the disturbed call segments according to the order of call occurrence, and write the fingerprint of the disturbed chain position into the call status chain. Select the call nodes with the same business data identifier and whose chain position order is within the coverage of the disturbed call segments as candidate call node bands. Specifically,

[0192] The order of the chain positions is determined by the call occurrence time recorded in each chain position within the disturbance call segment. If the call occurrence times are the same, the order is determined by the original chain position in the call status chain. Then, the disturbance chain position fingerprint is attached to each chain position covered by the start chain position to the end chain position of the disturbance call segment. The chain position in the call status chain that records the business data identifier and the call status is used as the call node.

[0193] Call nodes are filtered based on the business data identifier and chain position coverage area of ​​the disturbed call segment. Only call nodes with consistent business data identifiers and chain position order within the coverage area from the start chain position to the end chain position are retained. Call nodes with inconsistent business data identifiers are excluded, as are call nodes whose chain position order exceeds the chain position coverage area. The retained call nodes are then connected according to the chain position order to form a candidate call node band.

[0194] S4.3.3. Perform propagation pruning on the candidate call node band, removing call nodes that are only connected to stable call tracks and have not been written with the perturbation chain fingerprint, forming a perturbation propagation node band. Specifically,

[0195] Read the chain connection relationship and the perturbation chain fingerprint carrying status from both ends of the candidate call node band node by node; if the call node does not carry a perturbation chain fingerprint and both the forward chain connection and the backward chain connection point only to the stable call track, delete it from the candidate call node band, and continue to check the call nodes adjacent to the deletion position.

[0196] Calling nodes with perturbation chain fingerprints are retained, and those without perturbation chain fingerprints but with at least one chain connection pointing to a retained calling node are also retained. Calling nodes that meet the deletion conditions at both ends of the candidate calling node band are repeatedly deleted until no more calling nodes that are only connected to the stable calling track and do not have perturbation chain fingerprints appear at both ends of the candidate calling node band. Then, the retained calling nodes are connected in the original chain order to form a perturbation propagation node band.

[0197] S4.3.4. Constrain the calling node's direction using the chain position transmission relationship in the disturbance propagation node band, and perform node convergence on the disturbance propagation node band to form anomaly location nodes. Specifically,

[0198] Starting from the call node with the earliest call time and carrying the perturbation chain fingerprint in the perturbation propagation node band, the perturbation chain fingerprint is read node by node according to the chain transmission relationship from the earlier chain to the later chain.

[0199] If the perturbation chain fingerprint continues to exist in subsequent calling nodes, the transmission direction is maintained, and subsequent calling nodes that only continue the same perturbation chain fingerprint are merged into the tracking starting point; if the perturbation chain fingerprint stops appearing, the current tracking is terminated; if the perturbation chain fingerprint reappears after being interrupted, the tracking starting point is re-determined from the position of reappearance.

[0200] The calling node that first carries the fingerprint of the perturbation chain position in each tracking is retained to form the abnormal location node.

[0201] S4.3.5. Bind the anomaly location node to the disturbed call segment using a chain-linked tag, and assign an anomaly call tag to the anomaly location node to form an anomaly call node. Specifically,

[0202] Read the business data identifier, call identifier, and chain position order carried by the abnormal location node, and search for the disturbed call segment in the disturbance isolation track where the business data identifier and call identifier are the same and the starting chain position to the ending chain position covers the chain position order of the abnormal location node.

[0203] If the search result is unique, the anomaly location node and the disturbed call segment are kept in the same call state chain segment, and the call occurrence order and call entropy disturbance value change pattern of the disturbed call segment are attached to the anomaly location node to complete the same chain tag binding.

[0204] Abnormal location nodes with inconsistent business data identifiers, inconsistent call identifiers, or chain position order exceeding the coverage area of ​​the disturbed call segment will not be bound to the same chain tag; abnormal location nodes that have completed the same chain tag binding will be given an abnormal call tag, thus forming an abnormal call node.

[0205] S5. Associate and connect the basic storage level, temporary call level, and abnormal call node according to the call order of business data to obtain the hierarchical storage call relationship. Based on the hierarchical storage call relationship, perform hierarchical storage parsing on the business data to generate the call path of the business data.

[0206] S5.1. According to the order of business data calls, record the hierarchical relationship of hierarchical jumps and exception isolation pointers between the basic storage level, temporary call level, and exception call section. Specifically,

[0207] Based on the business data identifier, the hierarchical residency range of the basic storage layer, the temporary residency range of the temporary call layer, and the chain order of the abnormal call nodes are arranged according to the order of the call occurrence.

[0208] A hierarchical jump pointer is established from the initial call segment of the temporary residency scope to the basic storage level, and a return pointer is established from the terminating call segment of the temporary residency scope to the subsequent basic storage level.

[0209] When an abnormal call node is located in the coverage area of ​​the basic storage level or the coverage area of ​​the temporary call level, the connection between the incoming side of the abnormal call node and its storage level is preserved, and the call node is read step by step from the chain position immediately adjacent to the outgoing side of the abnormal call node, in the original call occurrence order.

[0210] If the next call node still carries an exception call flag, continue reading the next node in the chain; stop reading when the first call node without an exception call flag is encountered, and designate the first call node without an exception call flag as the continuation call position. During the reading process, only consecutive adjacent exception call nodes are crossed, and no call nodes without exception call flags are crossed.

[0211] Read the hierarchical residency boundary and temporary residency boundary between the abnormal call node and the continuation call location; if the hierarchical boundary is not crossed, connect the storage level where the abnormal call node is located on the incoming side to the continuation call location; if the temporary residency start segment or end segment is crossed, retain the hierarchical jump pointers from the basic storage level to the temporary call level and the return pointers from the temporary call level to the subsequent basic storage level in the original call occurrence order, and then connect the last hierarchical pointer to the continuation call location, prohibiting the establishment of direct connections across hierarchical boundaries.

[0212] If there is no call node without an exception call marker after the exception call node, the continuation relationship on the outgoing side of the exception call node is not established, and the consecutive exception call nodes after the exception call node are retained in the exception isolation pointer.

[0213] S5.2. Based on the hierarchical storage and call relationship, perform hierarchical storage parsing on the business data to generate the call path for the business data. Specifically,

[0214] S5.2.1 Starting from the business data access entry point, parse each call node of the hierarchical storage call relationship sequentially according to the hierarchical jump direction, and isolate abnormal call nodes. Specifically,

[0215] Starting from the first call node connected to the business data access entry point, the subsequent call nodes are read one by one according to the hierarchical jump pointers in the hierarchical storage call relationship, while preserving the connection order between the basic storage level and the temporary call level.

[0216] When a call node with an abnormal call flag is read, the chain positions of the incoming and outgoing call nodes of the abnormal call node are recorded. The connection of the abnormal call node to participate in the normal hierarchical jump is disconnected, and the abnormal call node is kept separately in the abnormal isolation pointer. Call nodes without an abnormal call flag maintain the original hierarchical jump connection until the end call node is resolved.

[0217] S5.2.2, Locate the associated position connecting the outgoing call node of the abnormal call node from the incoming call node, specifically,

[0218] Starting from the inbound call node immediately adjacent to the abnormal call node, read the business data identifier and backward hierarchical jump pointer node by node in the order of call occurrence towards the earlier node.

[0219] If the business data identifier is the same as the outgoing call node of the abnormal call node, and the target storage level location pointed to by the backward hierarchical jump is the same as the storage level location of the outgoing call node of the abnormal call node, it is reserved as a connectable location.

[0220] If the business data identifiers are different or the target storage level locations are different, continue reading the previous inbound call node. If there are multiple connectable locations, select the connectable location whose call order is closest to the abnormal call node, and determine the backward chain position of the call node containing the connectable location as the associated location.

[0221] S5.2.3. Connect the associated location with the basic storage level and the temporary call level to generate the call path for business data. Specifically, Use the associated location as the starting point for continuation after isolating the abnormal call node, and read the backward jump pointer of the associated location.

[0222] The backward hierarchical jump points to the temporary scheduling location, connects the associated location to the starting call node of the temporary scheduling location, and then connects the ending call node of the temporary scheduling location to the subsequent resident storage location.

[0223] The backward hierarchical jump points to the resident storage location, directly connecting the associated location to the resident storage location; the original call order from the business data access entry to the associated location is preserved, and the associated location, temporary scheduling location and resident storage location are connected in sequence to form a business data call path that avoids abnormal call nodes.

[0224] After the business data call path is generated, the business data call path is saved with the business data identifier as the association benchmark, so that the business data identifier points to the current hierarchical jump direction, basic storage level, temporary call level and exception isolation point of the business data.

[0225] When subsequent business data calls begin, the business data access entry reads the saved business data call path based on the business data identifier, and accesses the basic storage level or temporary call level according to the hierarchical jump direction in the business data call path.

[0226] After the subsequent call segment completes call entropy perturbation analysis, wavelet transform frequency domain layering, and hierarchical storage parsing, it saves the newly generated business data call path; during the formation of the newly generated business data call path, the business data access entry continuously reads the original business data call path.

[0227] Once the newly generated business data call path is saved, subsequent business data calls will read the newly generated business data call path, and business data calls that have already started will be completed according to the business data call path read at the start of the call.

[0228] This embodiment also provides a computer device applicable to the hierarchical storage and retrieval method of business data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the hierarchical storage and retrieval method of business data as proposed in the above embodiment.

[0229] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0230] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the hierarchical storage and retrieval method for business data as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0231] In summary, this invention quantifies the discrete fluctuations in the internal state distribution of a call segment by performing call entropy perturbation analysis, achieving a structured separation of abnormal and stable call components. This avoids storage level adjustments and call path offsets caused by the mixing of abnormal fluctuations and stable call changes, reduces the interference of short-term surges and repeated fluctuations on subsequent frequency domain layering, and prevents abnormal call states from interfering with storage level determination. Simultaneously, it clarifies the hierarchical boundaries between long-term resident requirements and temporary call requirements, reducing invalid migration and repeated scheduling of business data between different storage levels. This enhances the stability of hierarchical storage call relationships, the continuity of call paths, and the consistency of business data call results. During call state analysis, the business data access entry point reads the already generated business data call path, ensuring that the current business data call primarily executes business data call path reading and storage level access, thus reducing the impact of call entropy perturbation analysis and wavelet transform frequency domain layering on the business data call response time.

[0232] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for hierarchical storage and retrieval of business data, characterized in that: include, Collect the call status information of business data, connect the call status information with the current storage execution location, and generate a call status chain; Call entropy perturbation analysis is performed on the call segments in the call state chain to obtain the call entropy perturbation value of each call segment. High entropy perturbation segments with abnormal fluctuations are identified from the call entropy perturbation values ​​and arranged as perturbation isolation tracks. Based on the perturbation isolation track, the call segments with perturbations are found in the call state chain and marked as perturbation call segments. Each call segment not marked as a perturbation call segment is reconstructed into a stable call track. Wavelet transform frequency domain layering is performed on the stable call track to obtain low-frequency coefficient sequences and high-frequency coefficient sequences, which are then converted into low-frequency resident components and high-frequency resident components, respectively. By using low-frequency resident components to correct the current storage execution position, the basic storage level of business data is obtained. By using high-frequency resident components to limit the temporary call level of business data, the call node of business data is located and assigned an abnormal call mark according to each disturbance call segment in the disturbance isolation track, thus forming an abnormal call node. The basic storage level, temporary call level, and abnormal call node are associated and connected according to the call order of business data to obtain the hierarchical storage call relationship. Based on the hierarchical storage call relationship, the business data is parsed for hierarchical storage to generate the call path of the business data.

2. The hierarchical storage and retrieval method for business data as described in claim 1, characterized in that: The generation of the call state chain is specifically as follows: The call monitoring interface at the business data access entry point collects call status information of business data during the call process and obtains the current storage execution location of business data in each call status. The call state information is connected to the current stored execution location in the order in which the calls occurred, to generate a call state chain.

3. The hierarchical storage and retrieval method for business data as described in claim 2, characterized in that: The arrangement is a disturbance isolation rail, and the specific steps are as follows: The call state chain is divided into several call segments, and the state difference arrangement within each call segment is performed. The change pattern of adjacent chain positions within the same call segment is extracted, and the chain positions with the same change pattern are grouped into a state arrangement group. The proportion and distribution interval of each permutation state group within the call segment are statistically analyzed, and the proportion and distribution interval of the chain positions are discretely quantized to form the call entropy perturbation value. By using the call entropy perturbation value, the rising and maintaining characteristics and the repeated fluctuation characteristics are screened out in the call state chain, and the corresponding call segments are identified as high entropy perturbation segments; The high-entropy disturbance segments are arranged into disturbance isolation tracks according to the order in which they are called.

4. The hierarchical storage and retrieval method for business data as described in claim 3, characterized in that: The reconstruction to a stable call track involves the following steps: Using the call order of the high-entropy perturbation segment in the perturbation isolation track as the search condition, we search for call segments in the call state chain that have the same call order and the same call entropy perturbation value change pattern, and mark the found call segments as perturbation call segments. The disturbed call segments are isolated and marked, and the discontinuous positions between adjacent call segments on both sides of the disturbed call segment are determined as link gaps; By using link gaps to define connection boundaries and reconnecting the call segments that were not marked as disturbing call segments, a stable call track is formed.

5. The hierarchical storage and retrieval method for business data as described in claim 4, characterized in that: The conversion into low-frequency resident components and high-frequency resident components involves the following specific steps: Trajectory sampling is performed on the call segments in the stable call track, and the sampling results are separated according to the connection boundary to form a stable sampling segment; A non-decimation wavelet scaling function is used to perform scale smoothing on adjacent sampling points in a stable sampling segment, and the call components whose change amplitudes continuously decrease are organized into a low-frequency coefficient sequence. Using the non-subtractive wavelet detail operator, the abrupt change position where the change direction is reversed between adjacent sampling points in the stable sampling segment is located, and the high-frequency detail coefficients of adjacent sampling points before and after the abrupt change position are calculated and arranged into a high-frequency coefficient sequence. Scale merging is performed on the low-frequency coefficient sequence to extract low-frequency coefficient segments with consistent change direction and continuously decreasing amplitude in adjacent stable sampling segments. The low-frequency coefficient segments are then concatenated using the connection boundary. The concatenated low-frequency coefficients and the stable call track are classified as low-frequency resident components. Based on the mutation position of the high-frequency coefficient sequence, the sudden call residence area in the stable call track is traced back, and the call segment state in the sudden call residence area is taken as the high-frequency residence component.

6. The hierarchical storage and retrieval method for business data as described in claim 5, characterized in that: The specific steps for obtaining the basic storage layer of business data are as follows: Establish a call order index between the sampling location of low-frequency resident components and the current storage execution location, and concatenate storage level locations with the same call order index into a low-frequency level dwell segment; For call segments that deviate from adjacent memory levels for a short period of time in low-frequency memory level dwell segments, perform a memory level fallback process, and retain the memory level positions of continuous dwell segments to form a corrected memory execution position; The same storage level location repeatedly pointed to by multiple adjacent call segments in the corrected storage execution location is determined as the basic level position, and the level dwell range is limited by the start and end call segments of the low-frequency level dwell segment. By concatenating the base level bits and the level residency range, we obtain the base storage level for business data.

7. The hierarchical storage and retrieval method for business data as described in claim 6, characterized in that: The specific steps for limiting the temporary invocation level of business data by using high-frequency resident components are as follows: The burst call dwell area in the high-frequency dwell component is subjected to pulse clustering screening. Call segments are merged according to the adjacency relationship of the mutation position and the continuous fluctuation relationship of the call segment state to form a high-frequency dwell pulse cluster. Using the concentrated segment of mutation locations within the high-frequency resident pulse cluster as the cutting boundary, the high-frequency resident pulse cluster is cut off at both ends, and the call interval from the starting call segment to the ending call segment after cutting is collected to form a temporary call trigger segment; The order of calls to the temporary call triggering segment is mapped to the current storage execution location, and the storage level swing position caused by the temporary call triggering segment is marked, forming a temporary level free track; The temporary level free track is excluded from the basic level, the storage level position that continuously overlaps with the basic level position is removed, and the storage level position that is located within the temporary residence range and deviates from the basic level position is screened out to form the temporary call position; Encapsulate the temporary call bit and temporary dwell range into a temporary call layer for business data.

8. The hierarchical storage and retrieval method for business data as described in claim 7, characterized in that: The specific steps for forming an abnormal call node are as follows: The perturbation call segment in the perturbation isolation track is processed into a chain fingerprint, and the change pattern of the call entropy perturbation value is bound to the call segment state in the same segment to form a perturbation chain fingerprint. The chain order of the disturbed call segment is arranged by the order in which the calls occur, and the fingerprint of the disturbed chain segment is written into the call status chain. Call nodes with the same business data identifier and whose chain order is within the coverage of the disturbed call segment are selected as candidate call node bands. The candidate call node band is propagated and pruned to remove call nodes that are only connected to stable call tracks and have not written perturbation chain fingerprints, thus forming a perturbation propagation node band; The calling node's direction is constrained by the chain position transmission relationship in the disturbance propagation node band, and the node band is converged to form an abnormal location node; The abnormal location node is bound to the disturbed call segment by a chain mark, and the abnormal call mark is assigned to the abnormal location node to form an abnormal call node.

9. The hierarchical storage and retrieval method for business data as described in claim 8, characterized in that: The hierarchical storage call relationship is a hierarchical relationship recorded according to the order of business data calls, showing the hierarchical jumps and exception isolation points between the basic storage level, the temporary call level, and the abnormal call node.

10. The hierarchical storage and retrieval method for business data as described in claim 9, characterized in that: The specific steps for the call path to generate business data are as follows: Starting from the business data access entry point, the call nodes of the hierarchical storage call relationship are parsed sequentially according to the hierarchical jump direction, and abnormal call nodes are isolated; Locate the associated position connecting the call node on the outgoing side of the abnormal call node from the call node on the incoming side of the abnormal call node; The associated location is connected to the basic storage level and the temporary call level to generate the call path for business data.

Images