Storage network token bucket scheduling algorithm based on three-party architecture and optimization method thereof

Abstract

The invention belongs to the technical field of electric power systems, and relates to a storage network token bucket scheduling algorithm based on a three-party architecture and an optimization method thereof.Data of electric power equipment are collected in real time and preprocessed, the preprocessed data are transmitted in batches through the token bucket scheduling algorithm, the data enter a storage layer step by step according to the priority, and the priority of the data is optimized. The system intelligently distributes data to different storage devices according to the importance of the data and the utilization condition of storage resources, dynamically adjusts the storage rate to achieve optimal configuration of the resources, can predict future storage requirements through a time sequence prediction model, flexibly adjusts a storage strategy according to the load condition of each storage node, and achieves the optimal configuration of the resources. The token generation rate and the data transmission priority are dynamically adjusted on the basis of real-time analysis of data transmission delay and storage capacity, and the overall storage and processing efficiency is further improved.

Description

Technical Field

[0001] This application belongs to the technical field of power systems. More specifically, it relates to a storage network token bucket scheduling algorithm based on a three-party architecture and its optimization method. Background Art

[0002] With the construction and development of smart grids, the degree of digitalization, networking, and intelligence of devices in the power system is getting higher and higher. Smart grids achieve the efficient operation and optimized management of the power system by integrating advanced communication technologies, computer technologies, and control technologies. In this context, the power system generates a large amount of real-time data, which needs to be effectively stored, managed, and analyzed to support the operation decision-making of the power system.

[0003] In the power system, the existing data storage technologies mainly include relational databases, file storage systems, and distributed storage systems. These technologies can store a large amount of data, but there are the following problems when dealing with real-time data in the power system:

[0004] The data writing speed is difficult to meet the requirements of real-time data generated at high speed;

[0005] The efficiency of data processing and storage is low, and fast retrieval and analysis of data cannot be achieved;

[0006] The utilization rate of storage resources is not high, resulting in an increase in storage costs. Summary of the Invention

[0007] The present invention provides a storage network token bucket scheduling algorithm based on a three-party architecture and its optimization method, aiming to solve the technical problems that the current data writing speed is difficult to meet the requirements of real-time data generated at high speed, the data processing and storage efficiency is low, and the utilization rate of storage resources is not high.

[0008] The storage network token bucket scheduling algorithm based on a three-party architecture and its optimization method includes the following steps:

[0009] Step 1: Collect real-time data of various devices in the power system, and perform preprocessing based on the collected real-time data to obtain preprocessed real-time data;

[0010] Step 2: Generate tokens based on the preprocessed real-time data by the token bucket scheduling algorithm. According to the allocation of tokens, the data is transmitted in batches and gradually according to priorities, and enters the storage layer;

[0011] Step 3: Allocate the data to different storage devices according to the importance of the data and the current utilization of storage resources, and schedule the data storage rate according to the token bucket algorithm;

[0012] Step 4: Predict the storage requirements for a period of time in the future based on the time series prediction model to obtain the storage requirements for a period of time in the future, and dynamically adjust the data storage strategy according to the load conditions of different storage nodes monitored;

[0013] Step 5: Dynamically adjust the token generation rate and data transmission priority based on the analysis of data transmission delay and storage capacity usage.

[0014] This token bucket scheduling algorithm and its optimization method for a storage network based on a three-party architecture effectively solve the problems of insufficient current data writing speed, low data processing and storage efficiency, and low storage resource utilization through a series of systematic steps. First, the system collects data of power equipment in real time and performs preprocessing to ensure the accuracy and timeliness of the data. Then, the token bucket scheduling algorithm is used to batch-transmit the preprocessed data, and it gradually enters the storage layer according to the priority, thus improving the data transmission efficiency. Further, the system intelligently allocates data to different storage devices according to the importance of the data and the utilization of storage resources, and dynamically adjusts the storage rate to achieve the optimal allocation of resources. In addition, through the time series prediction model, the system can foresee future storage requirements and flexibly adjust the storage strategy according to the load conditions of each storage node to ensure the efficient operation of the system. Finally, based on the real-time analysis of data transmission delay and storage capacity, the token generation rate and data transmission priority are dynamically adjusted, further improving the overall storage and processing efficiency. This technical solution not only improves the data processing ability of the power system but also realizes efficient and intelligent storage management.

[0015] Preferably, step 2 includes the following steps:

[0016] Priority weight assignment: Assign a priority weight to each data stream. The higher the priority of the data stream, the shorter the token it obtains. Set the priority weight of the data stream as w i , and satisfy w i ∈[0,1];

[0017] Token generation rate weighting: Based on the priority of the data stream, adjust the token generation rate:

[0018] α i =α·w i ;

[0019] In the formula: α represents the global token generation rate; w i represents the priority weight of data stream i; α i represents the specific token generation rate for data stream i;

[0020] Token bucket capacity assignment: Adjust the capacity of the token bucket through weighted assignment of the token bucket:

[0021] B i = B·w i ;

[0022] In the formula: B i represents the token bucket capacity of data stream i; B represents the global token bucket capacity;

[0023] Priority scheduling for data transmission: Determine the data transmission priority according to the generated token stream and the priority of the data stream. Among them, the data stream with a higher priority obtains tokens in the token bucket faster, so data transmission is preferentially performed. If the number of tokens in the token bucket is greater than or equal to the size of the data stream, data transmission is performed, and the corresponding number of tokens is consumed after transmission. If there are insufficient tokens, the data stream will temporarily stop transmission and wait for the generation of tokens.

[0024] Preferably, step 2 further includes the following steps:

[0025] Adaptive adjustment based on network load: Adaptively adjust the token generation rate based on network load:

[0026]

[0027] In the formula: α 0 represents the initial token generation rate; β represents the control factor; Load(t) represents the real-time detection value of network load, indicating the level of the current network load; α i (t) represents the token generation rate of device i at time t.

[0028] Preferably, step 3 includes the following steps:

[0029] Load condition:

[0030]

[0031] In the formula: Current_Usage j (t) represents the actual usage of storage device j at time t; Max_Capacity j represents the maximum capacity of device j; Load j (t) represents the utilization rate of device j at time t;

[0032] Remaining storage capacity:

[0033] C j (t) = Max_Capacity j ―Current_Usage j (t);

[0034] In the formula: C j(t) represents the remaining storage capacity of storage device j at time t;

[0035] Storage priority calculation:

[0036]

[0037] Where: w i represents the priority weight of data stream i;

[0038] Dynamic storage rate scheduling:

[0039]

[0040] Where: α represents the initial storage rate factor; β represents the adjustment parameter; R j (t) represents the storage rate of storage device j at time t;

[0041] The final storage rate R of data stream i on storage device j ij (t) is obtained based on the following formula:

[0042] R ij (t) = min(R i (t), R j (t));

[0043] Where: R i (t) represents the storage rate of data stream i; R j (t) represents the storage rate of storage device j; R ij (t) represents the final storage rate of data stream i on storage device j.

[0044] Preferably, step 4 includes the following steps:

[0045] Predict future demand: Based on the LSTM model, predict the storage demand of the storage node in the next period of time;

[0046] Judge whether the demand is met: According to the load situation of each storage node and the predicted storage demand, judge whether the current storage node meets the storage demand. If the current remaining storage capacity does not meet the future storage demand, the storage strategy needs to be adjusted;

[0047] If the current remaining storage capacity meets the future storage demand, maintain the current storage strategy.

[0048] Preferably, the adjustment of the storage strategy includes:

[0049] Expand storage resources: For storage nodes that do not meet the future storage demand, the system increases the storage resources, increases the hard disk capacity of the storage node according to a predetermined threshold; or shares part of the load through idle nodes;

[0050] Data migration: For storage nodes that do not meet future storage requirements, part of the data is migrated to other storage nodes for balance, and the priority of data migration is dynamically allocated according to the remaining capacity of the storage node, the current load, and the priority of the data stream:

[0051]

[0052] In the formula: S ij (t) represents the storage priority of data stream i on storage node j; C j (t) represents the remaining capacity of storage node j; D pred,j (t + k) represents the predicted storage requirement; Transfer ij (t) represents the degree to which data stream i should be migrated to storage node j;

[0053] Priority adjustment and rate limitation: For storage nodes that do not meet future storage requirements, the priority weight w of the data stream is adjusted i , and rate limitation is performed according to the priority of the data stream:

[0054]

[0055] In the formula: α i (t) represents the token generation rate of device i at time t; α represents the global token generation rate; w i represents the priority weight of data stream i; Load j (t) represents the load of storage node j; Max Load represents the maximum load capacity of storage node j;

[0056] The specific adjustment method of the priority weight w i is as follows:

[0057]

[0058] In the formula: w i (t) represents the adjusted priority weight of data stream i at time t; represents the initial priority weight of data stream i; Load j (t) represents the current load of storage node j at time t; maxLoad j represents the maximum load capacity of storage node j; β i represents the adjustment coefficient of data stream i, which is used to represent the sensitivity of the data stream to load changes.

[0059] Preferably, step 5 includes the following steps:

[0060] Feedback adjustment of the token generation rate:

[0061]

[0062] Where: Δα(t) represents the adjustment amount; α init (t) represents the token generation rate before adjustment; D(t) represents the current actual transmission delay; D target represents the set target delay; K p represents the proportional gain, adjusting the system's response to delay changes; K i represents the integral gain, adjusting the system's response to long-term cumulative deviations; K d represents the derivative gain, adjusting the system's response to instantaneous changes;

[0063] Feedback adjusts the data stream priority:

[0064]

[0065] Where: represents the data stream priority weight before adjustment; α w and β w represent the control coefficients for priority adjustment; C j (t) represents the storage capacity usage of the current storage node j; C target represents the set target storage utilization rate; D(t) represents the current actual transmission delay; D target represents the target delay; Δw i (t) represents the priority weight adjustment amount.

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

[0067] The token bucket scheduling algorithm for a storage network based on a three-party architecture and its optimization method effectively solves the problems of insufficient current data writing speed, low data processing and storage efficiency, and low storage resource utilization through a series of systematic steps. First, the system collects power equipment data in real time and preprocesses it to ensure data accuracy and timeliness. Then, the token bucket scheduling algorithm is used to batch-transmit the preprocessed data, which gradually enters the storage layer according to the priority, thus improving the data transmission efficiency. Further, the system intelligently allocates data to different storage devices according to the importance of the data and the utilization of storage resources, and dynamically adjusts the storage rate to achieve the optimal allocation of resources. In addition, through a time series prediction model, the system can foresee future storage requirements and flexibly adjust the storage strategy according to the load conditions of each storage node to ensure the efficient operation of the system. Finally, based on the real-time analysis of data transmission delay and storage capacity, the token generation rate and data transmission priority are dynamically adjusted to further improve the overall storage and processing efficiency. This technical solution not only improves the data processing ability of the power system but also enables efficient and intelligent storage management. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the following described drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0069] Figure 1 It is the overall step block diagram provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0070] In order to make the technical problems, technical solutions, and beneficial effects to be solved by the present application clearer, the following further details the present application in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0071] See Figure 1 As shown, a further description is made of the optimal embodiment of the present invention;

[0072] The token bucket scheduling algorithm for a storage network based on a three-party architecture and its optimization method include the following steps:

[0073] Step 1: Collect the real-time data of various devices in the power system, and perform preprocessing based on the collected real-time data to obtain the preprocessed real-time data; The real-time data continuously collected by various devices in the power system (such as power transformers, smart meters, line monitoring devices, etc.) mainly includes: voltage, current, frequency, load, device status (such as fault information), meteorological data (such as temperature and humidity), etc.

[0074] Preprocessing: First, perform data cleaning and filtering, such as removing outliers, filling in missing data, etc. All sensor data will be standardized to ensure that the data collected by different devices is within the same magnitude range.

[0075] Step 2: Generate tokens based on the preprocessed real-time data according to the token bucket scheduling algorithm. According to the distribution of tokens, the data is transmitted in batches and gradually according to priority, and enters the storage layer;

[0076] The said Step 2 includes the following steps:

[0077] Priority weight assignment: Assign a priority weight to each data stream. The higher the priority of the data stream, the shorter the token it obtains. Set the priority weight of the data stream as w i , and satisfy w i ∈[0,1];

[0078] Token generation rate weighting: Based on the priority of the data stream, adjust the token generation rate:

[0079] α i =α·w i ;

[0080] In the formula: α represents the global token generation rate; w i represents the priority weight of data stream i; α i represents the specific token generation rate for data stream i;

[0081] Token bucket capacity assignment: Adjust the capacity of the token bucket through token bucket weighted assignment:

[0082] B i =B·w i ;

[0083] In the formula: B i represents the token bucket capacity of data stream i; B represents the global token bucket capacity;

[0084] Priority scheduling for data transmission: Determine the transmission priority of data according to the generated token stream and the priority of the data stream. Among them, the data stream with a higher priority obtains tokens in the token bucket at a faster speed, so data transmission is prioritized. If the number of tokens in the token bucket is greater than or equal to the size of the data stream, data transmission is performed, and the corresponding number of tokens is consumed after transmission. If there are insufficient tokens, the data stream will temporarily stop transmission and wait for the generation of tokens.

[0085] Step 2 further includes the following steps:

[0086] Adaptive adjustment based on network load: Adaptive adjustment of the token generation rate based on network load:

[0087]

[0088] In the formula: α 0 represents the initial token generation rate; β represents the control factor; Load(t) represents the real-time detection value of network load, indicating the level of the current network load; α i (t) represents the token generation rate of device i at time t.

[0089] In this embodiment, by assigning priority weights to each data stream, the system can dynamically adjust the token generation rate and token bucket capacity of the data stream, so that the data stream with a higher priority is preferentially guaranteed in terms of token generation rate and capacity allocation. In this way, the data stream of critical tasks can be transmitted quickly and the delay can be reduced, while the traffic with a lower priority can be appropriately postponed to avoid occupying too many resources. This scheduling method not only improves the efficiency of data transmission, but also optimizes the utilization of storage resources, ensuring the stability and efficiency of the system. In this embodiment, the relationship between the token generation rate and the network load is automatically adjusted. When the network load is high, the token generation rate will be appropriately reduced to reduce network congestion caused by too fast transmission rate; while when the network load is low, the token generation rate can be restored to the initial value to improve data transmission efficiency. This adaptive adjustment mechanism enables the storage network to be effectively regulated according to the real-time network state, avoiding waste of resources or transmission bottlenecks, improving the stability and adaptability of the network, and thus realizing a more intelligent and efficient storage scheduling.

[0090] Step 3: Allocate data to different storage devices according to the importance of the data and the current utilization of storage resources, and schedule the data storage rate according to the token bucket algorithm;

[0091] Step 3 includes the following steps:

[0092] Load situation:

[0093]

[0094] Where: Current_Usage j (t) represents the actual usage of storage device j at time t; Max_Capacity j represents the maximum capacity of device j; Load j (t) represents the utilization rate of device j at time t;

[0095] Remaining storage capacity:

[0096] C j (t) = Max_Capacity j ―Current_Usage j (t);

[0097] Where: C j (t) represents the remaining storage capacity of storage device j at time t;

[0098] Calculation of storage priority:

[0099]

[0100] Where: w i represents the priority weight of data stream i;

[0101] Dynamic storage rate scheduling:

[0102]

[0103] Where: α represents the initial storage rate factor; β represents the adjustment parameter; R j (t) represents the storage rate of storage device j at time t;

[0104] The final storage rate R of data stream i on storage device j ij (t) is obtained based on the following formula:

[0105] R ij (t) = min(R i (t), R j (t));

[0106] Where: R i (t) represents the storage rate of data stream i; R j (t) represents the storage rate of storage device j; R ij (t) represents the final storage rate of data stream i on storage device j.

[0107] Step 4: Based on the time series prediction model, predict the storage requirements for a future period of time, obtain the storage requirements for the future period of time, and dynamically adjust the data storage strategy according to the monitored load conditions of different storage nodes;

[0108] Step 4 includes the following steps:

[0109] Predict future demand: Based on the LSTM model, predict the storage demand of the storage node within a certain period in the future;

[0110] Judge whether the demand is met: According to the load condition of each storage node and the predicted storage demand, judge whether the current storage node meets the storage demand. If the remaining storage capacity currently does not meet the future storage demand, the storage strategy needs to be adjusted;

[0111] If the remaining storage capacity currently meets the future storage demand, maintain the current storage strategy.

[0112] The said adjustment of the storage strategy includes:

[0113] Expand storage resources: For the storage nodes that do not meet the future storage demand, the system increases the storage resources, increasing the hard disk capacity of the storage node according to a predetermined threshold; or sharing part of the load through idle nodes;

[0114] Data migration: For the storage nodes that do not meet the future storage demand, migrate part of the data to other storage nodes for balance, where the priority of data migration is dynamically allocated according to the remaining capacity of the storage node, the current load, and the priority of the data stream:

[0115]

[0116] In the formula: S ij (t) represents the storage priority of data stream i on storage node j; C j (t) represents the remaining capacity of storage node j; D pred,j (t + k) represents the predicted storage demand; Transfer ij (t) represents the degree to which data stream i should be migrated to storage node j;

[0117] Priority adjustment and rate limitation: For the storage nodes that do not meet the future storage demand, by adjusting the priority weight w i of the data stream, perform rate limitation according to the priority of the data stream:

[0118]

[0119] In the formula: α i (t) represents the token generation rate of device i at time t; α represents the global token generation rate; w i represents the priority weight of data stream i; Load j (t) represents the load of storage node j; Max Load represents the maximum load capacity of storage node j;

[0120] The priority weight w i is adjusted as follows:

[0121]

[0122] In the formula: w i (t) represents the adjusted priority weight of data stream i at time t; represents the initial priority weight of data stream i; Load j (t) represents the current load of storage node j at time t; maxLoad j represents the maximum load capacity of storage node j; β i represents the adjustment coefficient of data stream i, which is used to represent the sensitivity of the data stream to load changes.

[0123] In this embodiment, the system can allocate appropriate storage resources for different data streams according to the priority of each data stream and the actual load of the storage device. At the same time, the scheduling of the storage rate also takes into account the load of the storage device to ensure that high-load storage devices are not overused, while low-load devices can fully utilize their storage capabilities. Through this dynamic scheduling strategy, the system can more intelligently respond to load fluctuations, optimize the allocation of storage resources, thereby avoiding waste or shortage of storage resources, and improving the utilization rate of storage resources and the overall performance of the system.

[0124] Step 5: Dynamically adjust the token generation rate and data transmission priority based on the analysis of data transmission delay and storage capacity usage.

[0125] The said Step 5 includes the following steps:

[0126] Feedback adjustment of the token generation rate:

[0127]

[0128] α(t) = α init (t) + Δα(t);

[0129] In the formula: Δα(t) represents the adjustment amount; α init (t) represents the token generation rate before adjustment; D(t) represents the current actual transmission delay; D target represents the set target delay; K p represents the proportional gain, which adjusts the system's response to delay changes; K i represents the integral gain, which adjusts the system's response to long-term cumulative deviations; K d represents the derivative gain, which adjusts the system's response to instantaneous changes;

[0130] Feedback adjustment of data stream priority:

[0131] Δw i (t) = α w ·(C j (t) ― C target ) + β w ·(D(t) ― D target );

[0132]

[0133] Where: represents the data stream priority weight before adjustment; α w and β w represent the control coefficients for priority adjustment; C j (t) represents the storage capacity usage of the current storage node j; C target represents the set target storage utilization rate; D(t) represents the current actual transmission delay; D target represents the target delay; Δw i (t) represents the priority weight adjustment amount.

[0134] In this embodiment, based on the prediction of future storage requirements by the LSTM model, the storage strategy is adjusted in combination with the load conditions of the storage nodes. This prediction model can identify in advance the load pressure that the storage nodes may face in a future period of time, so as to make preparations in advance. By judging whether the current storage capacity can meet the future storage requirements, the system can take preventive measures before the storage requirements surge, such as adjusting the storage strategy or optimizing the storage allocation of the data stream. This forward-looking storage scheduling strategy significantly improves the system's resource planning ability, avoids storage shortages or performance bottlenecks caused by sudden storage requirements, and improves the reliability and elasticity of the storage network.

[0135] The above are only the preferred embodiments of the present application, and are not intended to limit the present application. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A storage network token bucket scheduling algorithm and its optimization method based on a tripartite architecture, characterized in that: The following steps are involved: Step 1: Collect real-time data of various devices in the power system, and pre-process the collected real-time data to obtain pre-processed real-time data; Step 2: The pre-processed real-time data generates tokens based on the token bucket scheduling algorithm. According to the allocation of tokens, the data is gradually transmitted in batches and by priority and enters the storage layer; Step 3: Allocate data to different storage devices based on the importance of the data and the current utilization of storage resources, and schedule the data storage rate based on the token bucket algorithm; Step 4: Predict the storage demand in the future based on the time series prediction model, obtain the storage demand in the future, and dynamically adjust the data storage strategy according to the load of different monitored storage nodes; Step 5: Dynamically adjust the token generation rate and data transmission priority based on the analysis of data transmission latency and storage capacity usage.

2. According to the storage network token bucket scheduling algorithm based on the tripartite architecture and its optimization method according to claim 1, it is characterized in that: The step 2 comprises the following steps: Priority weight allocation: Assign a priority weight to each data flow. The higher the priority of the data flow, the shorter the token obtained. Set the priority weight of the data flow to w i , and satisfy w i ∈[0,1]; Token generation rate weighting: Based on the priority of the data flow, adjust the token generation rate: α i ＝α·w i ； Where: α represents the global token generation rate; w i represents the priority weight of data stream i; α i It is expressed as the specific token generation rate of data stream i; Token bucket capacity allocation: The capacity of the token bucket is adjusted through token bucket weighted allocation: B i =B·w i ; Where: B i represents the token bucket capacity of data stream i; B represents the global token bucket capacity; Priority scheduling of data transmission: Determine the data transmission priority based on the generated token stream and the priority of the data stream. The data stream with high priority obtains tokens in the token bucket quickly, so data transmission is given priority. If the number of tokens in the token bucket is greater than or equal to the size of the data stream, data transmission is performed and the corresponding number of tokens is consumed after the transmission. If there are insufficient tokens, the data stream will temporarily stop transmitting and wait for the generation of tokens.

3. According to the storage network token bucket scheduling algorithm based on the tripartite architecture and its optimization method according to claim 2, it is characterized in that: The step 2 further comprises the following steps: Adaptive adjustment based on network load: Adaptive adjustment of token generation rate based on network load: Where: α0 represents the initial token generation rate; β represents the control factor; Load(t) represents the real-time detection value of the network load, indicating the current network load level; α i (t) represents the token generation rate of device i at time t.

4. According to the storage network token bucket scheduling algorithm based on the tripartite architecture and its optimization method according to claim 1, it is characterized in that: The step 3 comprises the following steps: Load conditions: In the formula: Current_Usage j (t) represents the actual usage of storage device j at time t; Max_Capacity j Indicates the maximum capacity of device j; Load j (t) represents the utilization rate of equipment j at time t; Storage capacity remaining: C j (t)=Max_Capacity j ―Current_Usage j (t); Where: C j (t) represents the remaining storage capacity of storage device j at time t; Storage priority calculation: Where: w i represents the priority weight of data stream i; Dynamic storage rate scheduling: Where: α represents the initial storage rate factor; β represents the adjustment parameter; R j (t) represents the storage rate of storage device j at time t; The final storage rate R of data stream i on storage device j ij (t) is obtained based on the following formula: R ij (t)=min(R i (t),R j (t)); Where: R i (t) represents the storage rate of data stream i; R j (t) represents the storage rate of storage device j; R ij (t) represents the final storage rate of data stream i on storage device j.

5. According to the storage network token bucket scheduling algorithm based on the tripartite architecture and its optimization method according to claim 1, it is characterized in that: The step 4 comprises the following steps: Predict future demand: Predict the storage demand of storage nodes in the future based on the LSTM model; Determine whether the demand is met: Based on the load of each storage node and the predicted storage demand, determine whether the current storage node meets the storage demand. If the remaining storage capacity does not meet the future storage demand, the storage strategy needs to be adjusted. If the current remaining storage capacity meets future storage needs, the current storage strategy is maintained.

6. The storage network token bucket scheduling algorithm and optimization method based on the tripartite architecture according to claim 5 is characterized in that: The adjusting storage strategy includes: Expand storage resources: For storage nodes that do not meet future storage needs, the system will increase storage resources by increasing the hard disk capacity of the storage node according to the predetermined threshold; or share part of the load through idle nodes; Data migration: For storage nodes that do not meet future storage needs, some data will be migrated to other storage nodes for balancing. The priority of data migration is dynamically allocated based on the remaining capacity of the storage node, the current load, and the priority of the data flow: Where: S ij (t) represents the storage priority of data stream i on storage node j; C j (t) represents the remaining capacity of storage node j; D pred,j (t+k) represents the predicted storage demand; Transfer ij (t) indicates whether data stream i should be migrated to storage node j; Priority adjustment and rate limiting: For storage nodes that do not meet future storage requirements, the priority weight w of the data flow is adjusted. i , rate limiting is performed based on the priority of the data flow: Where: α i (t) represents the token generation rate of device i at time t; α represents the global token generation rate; w i Indicates the priority weight of data stream i; Load j (t) represents the load of storage node j; Max Load represents the maximum load capacity of storage node j; The priority weight w i The specific adjustment methods are as follows: Where: w i (t) represents the adjusted priority weight of data stream i at time t; Indicates the initial priority weight of data stream i; Load j (t) represents the current load of storage node j at time t; maxLoad j represents the maximum load capacity of storage node j; β i The adjustment coefficient of data stream i is used to indicate the sensitivity of the data stream to load changes.

7. The storage network token bucket scheduling algorithm and optimization method based on the tripartite architecture according to claim 1 is characterized in that: The step 5 comprises the following steps: Feedback adjusts the token generation rate: Where: Δα(t) represents the adjustment amount; α init (t) represents the token generation rate before adjustment; D(t) represents the current actual transmission delay; D target Indicates the target delay set; K p Represents the proportional gain, which adjusts the system's response to delay changes; K i Indicates the integral gain, which adjusts the system's response to long-term accumulated deviations; K d It represents the differential gain, which adjusts the system's response to instantaneous changes; Feedback adjusts data flow priority: Where: Indicates the priority weight of the data flow before adjustment; α w and β w Indicates the control coefficient of priority adjustment; C j (t) represents the current storage capacity usage of storage node j; C target represents the target storage utilization rate; D(t) represents the current actual transmission delay; D target represents the target delay; Δw i (t) represents the priority weight adjustment amount.

Images