A hierarchical retrieval method and system based on multi-format document metadata
By using a layered calling method based on multi-format document metadata, the problem of resource consumption differences in multi-format document parsing is solved, enabling accurate resource assessment and scheduling for documents of different formats and sizes, thereby improving system stability and resource utilization.
Patent Information
- Application Number
- CN202611178676.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-05
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies cannot adapt to the differences in parsing time for documents of different formats and sizes in multi-format document parsing. This leads to timeouts, retry storms, memory exhaustion, and overall response delays when parsing heavy documents is not completed. Furthermore, the lack of a graded degradation mechanism results in poor system robustness.
By extracting the initial byte stream size of the document, the format baseline decoding rate, and the inflation coefficient, the estimated parsing time and memory requirements are calculated. A spatiotemporal constraint vector is generated, a nonlinear segmented mapping function is constructed to determine the call status, and differentiated scheduling strategies are executed, including scheduling instructions under security coordination, retry storm warning, memory avalanche crisis, and dual crash conditions.
It enables precise resource assessment and scheduling for documents of different formats and sizes, avoiding retry storms and memory overflows, improving system stability and resource utilization, and ensuring second-level response for lightweight documents and high parsing success rate for heavy documents.
Smart Images

Figure CN123115199A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data retrieval technology, and in particular to a hierarchical retrieval method and system based on multi-format document metadata. Background Technology
[0002] Multi-format document parsing is a core foundational component in enterprise digital transformation, office automation, and data intelligence processing, widely applied in various business scenarios such as document management, content extraction, intelligent auditing, and data archiving. With the rapid expansion of business scale, the demand for concurrent document parsing calls is growing exponentially, and the formats of documents to be processed are becoming increasingly diverse, ranging in size from lightweight text documents of several thousand bytes to heavyweight complex documents of hundreds of megabytes.
[0003] The existing technology has the following significant drawbacks: Firstly, the fixed timeout threshold mechanism cannot adapt to the varying parsing times of documents of different formats and sizes. Heavy documents are prone to timeouts due to incomplete parsing, leading to repeated client calls, a retry storm, and excessive network bandwidth and computing resources. Secondly, the first-come-first-served scheduling model does not consider the real-time memory status of parsing nodes for resource prediction. Simultaneous scheduling of multiple heavy documents can exhaust memory resources, causing abnormal process termination and service interruption. Thirdly, it lacks differentiated processing capabilities; parsing requests for lightweight documents are blocked by queued heavy documents, significantly increasing overall response latency. Fourthly, the lack of a tiered degradation mechanism means that when system resources are strained, all requests will fail, resulting in poor system robustness. Summary of the Invention
[0004] The main objective of this invention is to provide a hierarchical calling method based on multi-format document metadata, and further to provide a hierarchical calling system based on multi-format document metadata that can run and implement the above method, effectively solving the problems mentioned in the background art.
[0005] The technical solution of the present invention is as follows:
[0006] Firstly, a hierarchical retrieval method based on multi-format document metadata is proposed, which includes the following steps:
[0007] S1. Extract the initial byte stream size of the document, document format, gateway request waiting time, maximum allowed timeout of the link, available physical memory of the node, format baseline decoding rate and format inflation coefficient from the initial call instruction to obtain the basic physical parameter set;
[0008] S2. Based on the initial byte stream size of the document, the format baseline decoding rate, and the format inflation coefficient, calculate the estimated time to complete parsing and the estimated peak memory consumption to generate a document parsing load vector;
[0009] S3. Based on the maximum allowed timeout of the link, the waiting time of the gateway request, and the available physical memory of the node, calculate the remaining waiting time to generate a spatiotemporal constraint vector for calling the gateway;
[0010] S4. Based on the document parsing load vector and the call gateway spatiotemporal constraint vector, calculate the spatiotemporal mismatch state vector, obtain the time residual and memory residual, and construct a nonlinear piecewise mapping function to determine the call state;
[0011] S5. When the call state is a secure coordination state, maintain the initial call instruction; when the call state is a retry storm warning state, generate a corrected timeout threshold based on the time residual and the maximum allowable timeout of the link; when the call state is a memory avalanche crisis state, change to asynchronous queuing to generate a queuing scheduling instruction; when the call state is a double crash state, change to degraded reading to generate a degraded scheduling instruction.
[0012] S6. Allocate memory and start the decoding thread based on the initial call instruction, the queuing scheduling instruction or the degraded scheduling instruction.
[0013] A further improvement of the present invention is that the specific content of S1 is: extracting the size of the initial byte stream of the document from the initial call instruction. In addition to document formatting, a timer is started to record the time elapsed for gateway requests. And read the maximum allowed timeout for the link from the gateway's initial configuration file. Read the available physical memory of the current node. The format baseline decoding rate for the document format is extracted by monitoring historical throughput. And preset the format expansion factor that represents the memory expansion ratio. To obtain the set of basic physical parameters.
[0014] A further improvement of the present invention is that step S2 includes the following specific steps:
[0015] S21. Based on the initial byte stream size of the document The format reference decoding rate and the format dilation factor Calculate the estimated time required to complete the parsing. , ;
[0016] S22. Calculate the expected peak memory consumption. , ;
[0017] S23. The estimated time required to complete the parsing. and the expected peak memory consumption Combined into a two-dimensional vector, a document parsing load vector is generated. , .
[0018] A further improvement of the present invention is that step S3 includes the following specific steps:
[0019] S31. Based on the maximum permissible timeout of the link. The gateway request has been waiting time. And the physical memory available to the node Calculate the remaining waiting time , ;
[0020] S32, the remaining waiting time With the available physical memory of the node Combined into a two-dimensional vector, it generates a spatiotemporal constraint vector for calling the gateway. , .
[0021] A further improvement of the present invention is that step S4 includes the following specific steps:
[0022] S41. Based on the document parsing load vector With the aforementioned call gateway spatiotemporal constraint vector Calculate the spatiotemporal mismatch state vector D to obtain the time residual. and memory residual , ; ; ;
[0023] S42, Based on the time residual and the memory residual A nonlinear piecewise mapping function is constructed to determine the calling status. The expression of the nonlinear piecewise mapping function is as follows:
[0024] ;
[0025] in, Indicates a secure collaborative state. This indicates a retry of the storm warning status. This indicates a memory avalanche crisis. This indicates a double crash state.
[0026] A further improvement of the present invention is that S5 includes the following specific content: when When, maintain the initial invocation instruction; when At that time, based on the time residual and the maximum allowable timeout of the link Generate the corrected timeout threshold ; in, For a very small buffer of seconds; when When the initial call instruction is intercepted, the scheduling status of the request is changed to asynchronous queuing to generate a queuing scheduling instruction; when When the initial call instruction is changed to a degraded read to generate a degraded scheduling instruction, only requesting and returning lightweight basic metadata.
[0027] Secondly, a hierarchical calling system based on multi-format document metadata is proposed. The system includes: a physical parameter acquisition module, a feature generation module, a status judgment module, and a control execution module.
[0028] The physical parameter acquisition module is used to extract the initial byte stream size of the document, document format, gateway request waiting time, maximum allowed timeout of the link, available physical memory of the node, format baseline decoding rate and format inflation coefficient from the initial call instruction, and obtain a basic set of physical parameters.
[0029] The feature generation module is used to calculate the estimated time to complete parsing and the estimated peak memory consumption based on the initial byte stream size of the document, the format baseline decoding rate, and the format inflation coefficient, so as to generate a document parsing load vector; and to calculate the remaining waiting time based on the maximum allowed timeout of the link, the waiting time of the gateway request, and the available physical memory of the node, so as to generate a spatiotemporal constraint vector for calling the gateway.
[0030] The status judgment module is used to calculate the spatiotemporal mismatch status vector based on the document parsing load vector and the spatiotemporal constraint vector of the calling gateway, obtain the time residual and memory residual, and construct a nonlinear piecewise mapping function to judge the calling status.
[0031] The control execution module is configured to maintain the initial call instruction when the call state is a safe and cooperative state; generate a corrected timeout threshold based on the time residual and the maximum allowable timeout of the link when the call state is a retry storm warning state; change to asynchronous queuing and waiting to generate a queuing scheduling instruction when the call state is a memory avalanche crisis state; change to degraded reading to generate a degraded scheduling instruction when the call state is a double crash state; allocate memory and start a decoding thread based on the initial call instruction, the queuing scheduling instruction, or the degraded scheduling instruction.
[0032] The technical effects of this invention are as follows:
[0033] A hierarchical invocation method based on multi-format document metadata was constructed. By quantitatively representing the time and space resource requirements of document parsing, a document parsing load vector was generated, enabling accurate assessment of the resource consumption for parsing documents of different formats and sizes, providing a quantitative basis for differentiated scheduling. A spatiotemporal constraint vector was generated by combining the remaining waiting time of the gateway and the available memory of the node. The degree of spatiotemporal mismatch was calculated through vector operations, achieving accurate judgment of the invocation status and avoiding the bias of subjective experience-based scheduling. Differentiated scheduling strategies were executed for four different invocation states: direct access in a safe state, dynamic extension of the timeout threshold when time is insufficient, asynchronous queuing when memory is insufficient, and downgrading to reading basic metadata when there is a complete mismatch. This ensures both the success rate of parsing heavy documents and the second-level response of lightweight documents. It fundamentally avoids retry storms and memory overflow problems, significantly improving system stability and resource utilization, while also possessing good scalability and adaptability to various document formats and business scenarios. Attached Figure Description
[0034] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:
[0035] Figure 1 This is a flowchart illustrating a hierarchical calling method based on multi-format document metadata according to Embodiment 1 of the present invention;
[0036] Figure 2 This is a schematic diagram of the structure of a hierarchical calling system based on multi-format document metadata according to Embodiment 2 of the present invention. Detailed Implementation
[0037] Example 1: This example constructs a hierarchical invocation method based on multi-format document metadata. By quantitatively representing the time and space resource requirements for document parsing, a document parsing load vector is generated, enabling accurate assessment of the resource consumption for parsing documents of different formats and sizes, providing a quantitative basis for differentiated scheduling. A spatiotemporal constraint vector is generated by combining the remaining gateway waiting time and available node memory. The degree of spatiotemporal mismatch is calculated through vector operations, achieving accurate judgment of the invocation status and avoiding the bias of subjective, experience-based scheduling. Differentiated scheduling strategies are executed for four different invocation states: direct access in a safe state, dynamic extension of the timeout threshold when time is insufficient, asynchronous queuing when memory is insufficient, and downgrading to read basic metadata when there is a complete mismatch. This ensures both the success rate of parsing heavy documents and the second-level response time for lightweight documents. It fundamentally avoids retry storms and memory overflow problems, significantly improving system stability and resource utilization, while also possessing good scalability and adaptability to various document formats and business scenarios.
[0038] A hierarchical retrieval method based on multi-format document metadata, such as Figure 1 As shown, the specific steps include the following:
[0039] S1. Extract the initial byte stream size of the document, document format, gateway request waiting time, maximum allowed timeout of the link, available physical memory of the node, format baseline decoding rate and format inflation coefficient from the initial call instruction to obtain the basic physical parameter set.
[0040] In this embodiment, the specific content of S1 is: extracting the initial byte stream size of the document from the initial call instruction. In addition to document formatting, a timer is started to record the time elapsed for gateway requests. And read the maximum allowed timeout for the link from the gateway's initial configuration file. Read the available physical memory of the current node. The format baseline decoding rate for the document format is extracted by monitoring historical throughput. And preset the format expansion factor that represents the memory expansion ratio. To obtain the set of basic physical parameters.
[0041] In this embodiment, the initial byte stream size of the document This indicates the size of the raw byte stream of the document to be parsed before any decoding processing, in megabytes; the gateway request has been waiting time. This represents the time elapsed from when the gateway receives the frontend call request to the current moment, in seconds; the maximum allowed timeout for the link. This indicates the maximum time, in seconds, the gateway allows requests to wait before triggering a connection termination. The preferred value for this parameter is between 10 and 60 seconds, and can be adjusted according to different business scenarios. The node's available physical memory... This represents the amount of unused memory available for document parsing currently occupied by the parsing node, in megabytes; format baseline decoding rate. This represents the average decoding rate of the parsing node for this format document under normal load, measured in megabytes per second. This parameter is calculated by averaging the decoding rates of the parsing node for the same format document processed over the past 7 days. (Format inflation factor) This represents the ratio of the space occupied in memory after the document is decoded to the size of the original byte stream. It is dimensionless, and the preferred value varies for different document formats. The preferred value for PDF documents is 3 to 5, for Word documents it is 2 to 4, and for Excel documents it is 4 to 6. This parameter is calculated by taking the arithmetic mean of the ratio of the peak memory usage of the parsing node for processing documents of the same format over the past 7 days to the size of the original byte stream.
[0042] S2. Based on the initial byte stream size of the document, the format baseline decoding rate, and the format inflation coefficient, calculate the estimated time to complete parsing and the estimated peak memory consumption to generate a document parsing load vector.
[0043] In this embodiment, step S2 includes the following specific steps:
[0044] S21. Based on the initial byte stream size of the document The format reference decoding rate and the format dilation factor Calculate the estimated time required to complete the parsing. , ;
[0045] S22. Calculate the expected peak memory consumption. , ;
[0046] S23. The estimated time required to complete the parsing. and the expected peak memory consumption Combined into a two-dimensional vector, a document parsing load vector is generated. , .
[0047] In this embodiment, the estimated time required to complete the parsing is... This indicates the estimated time, in seconds, required for the parsing node to complete the decoding process of the document; the estimated time to complete the parsing. This indicates the peak memory size required by the parsing node during the decoding of the document, in megabytes.
[0048] S3. Based on the maximum allowed timeout of the link, the time the gateway request has been waiting for, and the available physical memory of the node, calculate the remaining waiting time to generate a spatiotemporal constraint vector for calling the gateway.
[0049] In this embodiment, step S3 includes the following specific steps:
[0050] S31. Based on the maximum permissible timeout of the link. The gateway request has been waiting time. And the physical memory available to the node Calculate the remaining waiting time , ;
[0051] S32, the remaining waiting time With the available physical memory of the node Combined into a two-dimensional vector, this generates a spatiotemporal constraint vector representing the gateway's scheduling tolerance threshold. , .
[0052] S4. Based on the document parsing load vector and the spatiotemporal constraint vector of the calling gateway, calculate the spatiotemporal mismatch state vector, obtain the time residual and memory residual, and construct a nonlinear piecewise mapping function to determine the calling state.
[0053] In this embodiment, step S4 includes the following specific steps:
[0054] S41. Based on the document parsing load vector With the aforementioned call gateway spatiotemporal constraint vector Calculate the spatiotemporal mismatch state vector D to obtain the time residual. and memory residual , ; ; ;
[0055] S42, Based on the time residual and the memory residual A nonlinear piecewise mapping function is constructed to determine the calling status. The expression of the nonlinear piecewise mapping function is as follows:
[0056] ;
[0057] in, Indicates a secure collaborative state. This indicates a retry of the storm warning status. This indicates a memory avalanche crisis. This indicates a double crash state.
[0058] In this embodiment, the time residual This represents the difference between the time required for parsing and the remaining waiting time at the gateway, in seconds; memory residual. This represents the difference between the peak memory required for parsing and the available memory of the node, in megabytes.
[0059] S5. When the call state is a secure collaborative state, maintain the initial call instruction; when the call state is a retry storm warning state, generate a corrected timeout threshold based on the time residual and the maximum allowable timeout of the link; when the call state is a memory avalanche crisis state, change to asynchronous queuing to generate a queuing scheduling instruction; when the call state is a double crash state, change to degraded reading to generate a degraded scheduling instruction.
[0060] In this embodiment, S5 includes the following specific content: when When, maintain the initial invocation instruction; when At that time, based on the time residual and the maximum allowable timeout of the link Generate the corrected timeout threshold ; in, This is a very small buffer period in seconds, preferably 0.5 to 1 second, used to compensate for time errors caused by network transmission latency and node load fluctuations; when When the initial call instruction is intercepted, the scheduling status of the request is changed to asynchronous queuing to generate a queuing scheduling instruction; when When the initial call instruction is changed to a degraded read to generate a degraded scheduling instruction, only requesting and returning lightweight basic metadata.
[0061] S6. Allocate memory and start the decoding thread based on the initial call instruction, the queuing scheduling instruction or the degraded scheduling instruction.
[0062] Example 2: This example proposes a hierarchical retrieval system based on multi-format document metadata, such as... Figure 2 As shown, it includes: a physical parameter acquisition module, a feature generation module, a status judgment module, and a control execution module;
[0063] The physical parameter acquisition module is used to extract the initial byte stream size of the document, document format, gateway request waiting time, maximum allowed timeout of the link, available physical memory of the node, format baseline decoding rate and format inflation coefficient from the initial call instruction, and obtain a basic set of physical parameters.
[0064] The feature generation module is used to calculate the estimated time to complete parsing and the estimated peak memory consumption based on the initial byte stream size of the document, the format baseline decoding rate, and the format inflation coefficient, so as to generate a document parsing load vector; and to calculate the remaining waiting time based on the maximum allowed timeout of the link, the waiting time of the gateway request, and the available physical memory of the node, so as to generate a spatiotemporal constraint vector for calling the gateway.
[0065] The status judgment module is used to calculate the spatiotemporal mismatch status vector based on the document parsing load vector and the spatiotemporal constraint vector of the calling gateway, obtain the time residual and memory residual, and construct a nonlinear piecewise mapping function to judge the calling status.
[0066] The control execution module is configured to maintain the initial call instruction when the call state is a safe and cooperative state; generate a corrected timeout threshold based on the time residual and the maximum allowable timeout of the link when the call state is a retry storm warning state; change to asynchronous queuing and waiting to generate a queuing scheduling instruction when the call state is a memory avalanche crisis state; change to degraded reading to generate a degraded scheduling instruction when the call state is a double crash state; allocate memory and start a decoding thread based on the initial call instruction, the queuing scheduling instruction, or the degraded scheduling instruction.
[0067] The parameters and steps of each unit module in the hierarchical calling system based on multi-format document metadata of the present invention described above can be referred to the parameters and steps in the embodiment of the hierarchical calling method based on multi-format document metadata in Embodiment 1 above.
[0068] Example 3: This example provides an electronic device, including a processor and a memory, wherein the memory stores a computer program that can be called by the processor; the processor executes the above-described hierarchical calling method based on multi-format document metadata by calling the computer program stored in the memory.
[0069] This electronic device can vary considerably depending on its configuration or performance. It may include one or more Central Processing Units (CPUs) and one or more memories, wherein the memory stores at least one computer program, which is loaded and executed by the processor to implement the hierarchical calling method based on multi-format document metadata provided in the above-described embodiment. The electronic device may also include other components for implementing device functions; for example, it may have wired or wireless network interfaces and input / output interfaces for data input and output. Further details are omitted here.
[0070] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.
[0071] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.
[0072] This invention is described with reference to flowchart illustrations and block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and block diagrams, as well as combinations of blocks in the flowchart illustrations and block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0073] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and boxes Figure 1 The steps of the function specified in one or more boxes.
[0074] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.
Claims
1. A method for hierarchical invocation based on metadata of multi-format documents, characterized in that: The specific steps include the following: S1. Extract the initial byte stream size of the document, document format, gateway request waiting time, maximum allowed timeout of the link, available physical memory of the node, format baseline decoding rate and format inflation coefficient from the initial call instruction to obtain the basic physical parameter set; S2. Based on the initial byte stream size of the document, the format baseline decoding rate, and the format inflation coefficient, calculate the estimated time to complete parsing and the estimated peak memory consumption to generate a document parsing load vector; S3. Based on the maximum allowed timeout of the link, the waiting time of the gateway request, and the available physical memory of the node, calculate the remaining waiting time to generate a spatiotemporal constraint vector for calling the gateway; S4. Based on the document parsing load vector and the call gateway spatiotemporal constraint vector, calculate the spatiotemporal mismatch state vector, obtain the time residual and memory residual, and construct a nonlinear piecewise mapping function to determine the call state; S5. When the call status is a secure collaboration status, maintain the initial call instruction; when the call status is a retry storm warning status, generate a corrected timeout threshold based on the time residual and the maximum allowable timeout of the link. When the call status is a memory avalanche crisis state, it is changed to asynchronous queuing to generate a queuing scheduling instruction; When the call status is a double crash state, it is changed to a degraded read to generate a degraded scheduling instruction; S6. Allocate memory and start the decoding thread based on the initial call instruction, the queuing scheduling instruction or the degraded scheduling instruction.
2. The hierarchical retrieval method based on multi-format document metadata according to claim 1, characterized in that: The specific content of the S1 is: extracting the document initial byte stream size in the initial calling instruction and the document format, starting a timer to record the gateway request waiting time and reading the link maximum allowed timeout from the gateway initial configuration file ; reading the current node available physical memory extracting the format benchmark decoding rate for the document format through historical throughput rate monitoring and presetting the format expansion coefficient representing the memory expansion multiple obtaining the basic physical parameter set.
3. The method of claim 2, wherein: S2 includes the following specific steps: S21、based on the document initial byte stream size , the format reference decoding rate and the format inflation coefficient , calculate the time consumption expected to complete the parsing , ; S22, calculating the predicted consumed peak memory , ; S23, the predicted completion parsing time and the predicted peak memory consumption are combined into a two-dimensional vector, generating a document parsing load vector , .
4. The method of claim 3, wherein: S3 includes the following specific steps: S31、based on the link maximum allowed timeout , the gateway request waiting time and the node available physical memory , calculate the remaining waiting time , ; S32, the remaining waiting time With the available physical memory of the node Combined into a two-dimensional vector, it generates a spatiotemporal constraint vector for calling the gateway. , .
5. A hierarchical retrieval method based on multi-format document metadata according to claim 4, characterized in that, S4 includes the following specific steps: S41. Based on the document parsing load vector With the aforementioned call gateway spatiotemporal constraint vector Calculate the spatiotemporal mismatch state vector D to obtain the time residual. and memory residual , ; ; ; S42, Based on the time residual and the memory residual A nonlinear piecewise mapping function is constructed to determine the calling status. The expression of the nonlinear piecewise mapping function is as follows: ; in, Indicates a secure collaborative state. This indicates a retry of the storm warning status. This indicates a memory avalanche crisis. This indicates a double crash state.
6. The hierarchical retrieval method based on multi-format document metadata according to claim 5, characterized in that, S5 includes the following specific content: when When, maintain the initial invocation instruction; when At that time, based on the time residual and the maximum allowable timeout of the link Generate the corrected timeout threshold ; in, For a very small buffer of seconds; when When the initial call instruction is intercepted, the scheduling status of the request is changed to asynchronous queuing to generate a queuing scheduling instruction; when When the initial call instruction is changed to a degraded read to generate a degraded scheduling instruction, only requesting and returning lightweight basic metadata.
7. A hierarchical retrieval system based on multi-format document metadata, implemented based on the hierarchical retrieval method based on multi-format document metadata as described in any one of claims 1-6, characterized in that, The system includes: a physical parameter acquisition module, a feature generation module, a state judgment module, and a control execution module; The physical parameter acquisition module is used to extract the initial byte stream size of the document, document format, gateway request waiting time, maximum allowed timeout of the link, available physical memory of the node, format baseline decoding rate and format inflation coefficient from the initial call instruction, and obtain a basic set of physical parameters. The feature generation module is used to calculate the estimated time to complete parsing and the estimated peak memory consumption based on the initial byte stream size of the document, the format baseline decoding rate, and the format inflation coefficient, so as to generate a document parsing load vector; and to calculate the remaining waiting time based on the maximum allowed timeout of the link, the waiting time of the gateway request, and the available physical memory of the node, so as to generate a spatiotemporal constraint vector for calling the gateway. The status judgment module is used to calculate the spatiotemporal mismatch status vector based on the document parsing load vector and the spatiotemporal constraint vector of the calling gateway, obtain the time residual and memory residual, and construct a nonlinear piecewise mapping function to judge the calling status. The control execution module is configured to maintain the initial call instruction when the call state is a safe and cooperative state; generate a corrected timeout threshold based on the time residual and the maximum allowable timeout of the link when the call state is a retry storm warning state; change to asynchronous queuing and waiting to generate a queuing scheduling instruction when the call state is a memory avalanche crisis state; change to degraded reading to generate a degraded scheduling instruction when the call state is a double crash state; allocate memory and start a decoding thread based on the initial call instruction, the queuing scheduling instruction, or the degraded scheduling instruction.
