Distributed task scheduling method in cloud computing environment

Abstract

The invention discloses a distributed task scheduling method in a cloud computing environment, which relates to the technical field of distributed task scheduling, and comprises the following steps of: firstly, monitoring all distributed rendering nodes in real time, and acquiring initialized data information generated by each distributed rendering computing node in real time in the monitoring process; through real-time monitoring of rendering nodes, dynamic generation of random seeds and intelligent task scheduling, the problem of regular distribution of noisy points caused by random seed overlapping in distributed rendering is solved. The system predicts an overlapping risk through statistical analysis and a machine learning model, adjusts random seeds, redistributes Tiles and optimizes calculation region division when the risk occurs, and ensures that a random sampling sequence of each Tiles is independent, so that the rendering quality and the calculation efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of distributed task scheduling, and in particular to a distributed task scheduling method in a cloud computing environment. Background Art

[0002] Distributed task scheduling in a cloud computing environment refers to the coordination and allocation of computing tasks across multiple computing nodes or servers to improve resource utilization, enhance computing efficiency, and ensure stable task execution. It typically involves key technologies such as task decomposition, resource scheduling, load balancing, and fault recovery to ensure that tasks are rationally allocated to computing resources and executed according to preset policies. Scheduling systems typically adopt centralized, distributed, or hybrid architectures, leveraging mechanisms such as task queues, scheduling algorithms (such as round-robin and shortest task first), auto-scaling, and elastic resource management to achieve efficient parallel computing. For example, in large-scale data processing scenarios (such as Hadoop and Spark) or container orchestration systems (such as Kubernetes), the scheduler dynamically allocates computing resources based on task requirements to ensure smooth task execution and optimize overall system performance.

[0003] Distributed rendering computing refers to the distribution of rendering tasks across multiple computing nodes (such as servers, GPU clusters, or cloud instances) to process 3D animations in parallel, accelerate computing, and optimize resource utilization. It is widely used in high-computing scenarios such as film special effects, animation production, architectural visualization, and game development. In cloud computing environments, distributed rendering computing is usually dynamically adjusted to optimize computing efficiency and resource utilization. These include automatic scaling (Auto Scaling), which dynamically adjusts computing nodes based on task load; smart task scheduling (Smart Scheduling), which optimizes task allocation based on computing performance, priority, and network bandwidth; load balancing (Load Balancing), which ensures even utilization of computing resources; distributed cache management (Distributed Caching), which shares lighting caches, texture data, and sampling modes to reduce duplicate calculations; and fault recovery (Failover Recovery), which automatically reallocates tasks when a computing node fails to prevent rendering interruptions. These adjustment mechanisms enable cloud-based distributed rendering to complete high-quality calculations in the shortest possible time, while reducing costs and improving system stability and rendering efficiency.

[0004] The existing technology has the following shortcomings: In 3D animation rendering scenarios, when using multiple computing nodes to render complex images in parallel, a single frame is usually split into multiple small blocks (Tiles) and assigned to different nodes for processing. Path Tracing, as a mainstream rendering algorithm, relies on a random number generator (RNG) for sampling to calculate physical effects such as reflection, refraction, and global illumination (GI). However, if the random seed initialization of distributed nodes overlaps, resulting in multiple Tiles using the same random number sequence, serious problems may arise. Due to the repetition of the random sampling pattern, the rendered noise distribution presents a regular pattern rather than the expected random distribution. This reduces the effectiveness of Monte Carlo sampling, making it impossible to average the noise between frames, affecting the final rendering quality.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] The present invention aims to provide a distributed task scheduling method in a cloud computing environment. By real-time monitoring of rendering nodes, dynamically generating random seeds, and intelligently scheduling tasks, this method addresses the problem of random seed overlap leading to regular noise distribution in distributed rendering. The system uses statistical analysis and machine learning models to predict overlap risks. When overlap risks occur, it adjusts the random seeds, reallocates tiles, and optimizes computational region partitioning to ensure the independence of the random sampling sequence for each tile, thereby improving rendering quality and computational efficiency and addressing the aforementioned issues.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a distributed task scheduling method in a cloud computing environment, comprising the following steps:

[0008] First, all distributed rendering nodes are monitored in real time, and the initialization data information generated by each distributed rendering computing node is obtained in real time during the monitoring process;

[0009] After obtaining the initialization data information of each node, the obtained initialization data information is preprocessed, and the preprocessed data is integrated to establish a structured data set;

[0010] Extract key indicators reflecting the overlap of random seed initialization methods of distributed nodes from the data set, and perform statistical analysis and visualization on the extracted key indicators to quantify the risk of random seed overlap;

[0011] Intelligently predict whether there is a risk of random seed overlap on distributed rendering nodes using a pre-trained machine learning model;

[0012] When the machine learning model determines that there is a risk of overlapping random seeds in tiles, it first generates a dynamic random seed based on the tile coordinates, frame index, and compute node ID to change the random number sequence of each tile. Secondly, it redistributes tiles with high overlap to different compute nodes to disperse the distribution range of the random number seeds. At the same time, it dynamically adjusts the tile size and optimizes the computational area division to change the sampling pattern between adjacent tiles.

[0013] Preferably, in a distributed rendering environment, after obtaining the initialization data information of each node, it is pre-processed and a structured data set is established. The specific steps are as follows:

[0014] First, collect initialization data from logs, API interfaces, and monitoring systems of each computing node, including hardware information, software version, load status, network connection status, and random number seeds, to ensure comprehensive data coverage.

[0015] Secondly, the acquired data is preprocessed, including data cleaning, missing value filling, data standardization, and time alignment to improve data accuracy and consistency;

[0016] Next, data is associated based on the compute node ID, task number, and timestamp, merging data from different sources and eliminating redundant information to ensure data uniqueness and integrity.

[0017] Then, the processed data is stored in a relational database, making it queryable, indexable, and scalable for efficient subsequent use.

[0018] Finally, data consistency and integrity checks are performed to ensure data quality, and the data storage structure is optimized according to business needs.

[0019] Preferably, key indicators reflecting the overlap of random seed initialization methods of distributed nodes are extracted from the data set, and the extracted indicators include the probability of collision of hash values calculated based on the hash function when the nodes distribute random seeds and the numerical distribution of the pseudo-random number generator at each computing node. The probability of collision of hash values calculated based on the hash function when the nodes distribute random seeds and the numerical distribution of the pseudo-random number generator at each computing node are statistically analyzed and visualized under the detection window, and a seed hash collision reference value and a PRNG sampling distribution consistency reference value are generated respectively. The random seed overlap risk is quantified by the seed hash collision reference value and the PRNG sampling distribution consistency reference value.

[0020] Preferably, the specific steps of generating a seed hash collision reference value by statistically analyzing and visualizing the probability of collision of hash values calculated based on the hash function when the node distributes random seeds are as follows:

[0021] Within the detection window, there are N computing nodes. Each computing node generates a random number when a random seed is assigned. The random number is then mapped to a fixed-length hash value using a hash function. To calculate the hash collision rate, the hash values of all computing nodes are compared to determine whether the nodes generate the same hash value. The hash collision matrix is defined as follows:

[0022]

[0023] Where C i，j is a hash collision matrix, which is a binary state variable indicating whether the hash values of computing nodes i and j are the same. i and H j The hash values of the random seeds generated for compute nodes i and j respectively;

[0024] Calculate the total number of collisions within the detection window. The calculation expression is as follows:

[0025]

[0026] Where P is the total number of hash collisions that occur between all computing nodes;

[0027] After obtaining the total number of collisions P, the severity of the hash collision is quantified and a seed hash collision reference value is generated. The generation formula is as follows:

[0028]

[0029] Where SHC is the seed hash collision reference value, H bin (H i ) is the hash value H of computing node i i Convert to binary vector, H bin (H i+1 ) is the hash value H of computing node i+1 i+1 Convert to a binary vector.

[0030] Preferably, the specific steps of performing statistical analysis and visualization processing on the numerical distribution of the pseudo-random number generator at each computing node under the detection window to generate a PRNG sampling distribution consistency reference value are as follows:

[0031] First, a pseudo-random number sequence is extracted from the PRNS sampling data of each computing node, and the numerical distribution characteristic entropy is calculated to measure the discrete degree of random number distribution of different nodes. The calculation expression is as follows:

[0032]

[0033] Where, X f is the characteristic entropy of the numerical distribution, H is the number of discrete intervals of the PRNG sampling value of the computing node, and P q is the normalized proportion of pseudo-random numbers in the qth numerical interval, and λ is a very small positive number used to avoid zero values in logarithmic operations;

[0034] When obtaining the numerical distribution characteristic entropy X of each computing node f Finally, in order to further quantify the similarity of random number distribution between multiple computing nodes, it is necessary to calculate the PRNG sampling distribution consistency reference value. The calculation expression is as follows:

[0035]

[0036] Where PSDU is the reference value of PRNG sampling distribution consistency, M is the total number of computing nodes, that is, the number of computing nodes in the distributed rendering environment, and B j，k is the overlap of PRNG sampling distribution between computing nodes j and k, and are the numerical distribution characteristic entropy of computing node j and computing node k respectively, ∈ is a very small positive number to prevent the denominator from being zero, is the normalization factor.

[0037] Preferably, the seed hash collision reference value and the PRNG sampling distribution consistency reference value that have undergone statistical analysis and visualization processing are input into a pre-trained machine learning model, and a random seed overlap risk coefficient is generated by the machine learning model. The random seed overlap risk coefficient is used to intelligently predict whether there is a random seed overlap risk in the distributed rendering node.

[0038] Preferably, a random seed overlap risk coefficient generated when a pre-trained machine learning model is used to intelligently predict whether a distributed rendering node has a random seed overlap risk is compared with a pre-set random seed overlap risk coefficient reference threshold to identify whether a distributed rendering node has a random seed overlap risk. The specific steps are as follows:

[0039] If the random seed overlap risk coefficient is greater than a preset random seed overlap risk coefficient reference threshold, the distributed rendering node is classified as having random seed overlap risk; if the random seed overlap risk coefficient is less than or equal to the preset random seed overlap risk coefficient reference threshold, the distributed rendering node is classified as not having random seed overlap risk.

[0040] Preferably, when the machine learning model determines that there is a risk of overlap in the random seeds of tiles, a dynamic random seed is first generated based on the tile coordinates, frame index, and compute node ID. Then, tiles with high overlap are redistributed to different compute nodes. At the same time, the specific steps for dynamically adjusting the tile size are as follows:

[0041] When the machine learning model predicts that there is a random seed overlap risk for tiles, that is, when the random seed overlap risk coefficient RSOR is greater than the preset random seed overlap risk coefficient reference threshold, the random seed for each tile should be dynamically generated first. The calculation expression is as follows:

[0042]

[0043] In the formula, Seed dynamic Is a dynamic random seed value, each tile is unique, X tile and Y tile The horizontal and vertical positions of the current Tiles in the rendered image, Frame idx Is the frame index of the current rendering task, indicating the animation frame number to which the current Tiles belongs. Node ID Is the compute node ID of the current Tiles calculation, RSOR ref is the random seed overlap risk coefficient reference threshold, e is the natural base, and Hash[·] is the hash function;

[0044] After dynamic seed generation is completed, the probability of random number sequence repetition is further reduced by dispersing the random number seed space, optimizing the spatial diversity of the sampling distribution, and reducing the similarity of sampling patterns between adjacent tiles. The calculation expression is as follows:

[0045] ,

[0046]

[0047] In the formula, Node new Is the identifier of the reallocated computing node, Node current is the node identifier to which the current Tiles originally belonged, Φ is the golden section constant, N total is the total number of computing nodes, mod is a modulo operation to ensure that the new node number falls within the valid node range, Size tile_new Is the size of the Tiles after dynamic adjustment, Size tile_original is the original default tile size, and Γ is the regularization constant used to dynamically adjust the amplitude.

[0048] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0049] The present invention successfully solves the problem of regular distribution of noise caused by random seed overlap in distributed rendering by real-time monitoring of distributed rendering nodes, dynamic generation of random seeds, and intelligent task scheduling, thereby significantly improving the quality of rendered images. First, by real-time monitoring and preprocessing of the initialization data information of each computing node, the system can accurately obtain the random seed distribution of the node and quantify the potential overlap risk through statistical analysis. Subsequently, with the help of a trained machine learning model, it intelligently predicts whether there is a risk of seed overlap, and dynamically adjusts the random seed generation method and Tiles allocation strategy when the risk occurs. Specifically, the system avoids repeated sampling patterns by generating new random seeds based on Tiles coordinates, frame index and computing node ID, while adjusting the Tiles size and computing area division to optimize computing efficiency. This solution ensures that the random sampling sequence of each Tiles is independent, eliminates the regular distribution of noise, improves rendering quality, and improves the efficiency and stability of distributed computing. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0051] Figure 1 This is a flow chart of a method for distributed task scheduling in a cloud computing environment according to the present invention. DETAILED DESCRIPTION

[0052] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0053] The present invention provides Figure 1 A distributed task scheduling method in a cloud computing environment is shown, comprising the following steps:

[0054] First, all distributed rendering nodes are monitored in real time, and the initialization data information generated by each distributed rendering computing node is obtained in real time during the monitoring process;

[0055] This information typically includes the node's hardware configuration (CPU / GPU model, memory size, etc.), operating system environment, current random seed generation strategy (such as fixed seed, timestamp seed, or hybrid), and relevant software version numbers. By acquiring this data in real time, we can fully understand the latest status of each node, providing accurate and timely foundational support for subsequent data analysis, prediction, and scheduling. In actual implementation, scripts or agents are typically deployed on each node to send the collected initialization information to a centralized data management module, ensuring the integrity and real-time nature of the information.

[0056] After obtaining the initialization data information of each node, the obtained initialization data information is preprocessed, and the preprocessed data is integrated to establish a structured data set;

[0057] In a distributed rendering environment, acquiring initialization data for each compute node involves collecting key parameters for each node at the start of a rendering task, such as hardware configuration (CPU / GPU specifications, memory size), software version (renderer, driver), network status, load, and random number generator seed. Since this data may come from various sources (such as log files, API monitoring interfaces, and system status reports), it requires preprocessing after acquisition, including data cleansing (removing outliers and filling missing values), format standardization (unifying data types and aligning timestamps), deduplication, and normalization to ensure data integrity and consistency. This preprocessed data is then consolidated—correlated and categorized based on associated information such as compute node ID, timestamp, and task number—to form a structured data set (such as a database table, distributed storage, or data warehouse) for subsequent data analysis, feature extraction, and machine learning model training. This ensures that the entire system can make intelligent decisions based on unified, standardized data, improving the efficiency of rendering task allocation and computational stability.

[0058] In a distributed rendering environment, after obtaining the initialization data information of each node, it is preprocessed and a structured data set is established. The specific steps are as follows:

[0059] First, initialization data is collected from logs, API interfaces, and monitoring systems of each computing node, including hardware information (CPU / GPU specifications, memory), software versions (rendering engine, driver), load status, network connection status, and random number seeds to ensure comprehensive data coverage.

[0060] Secondly, the acquired data is preprocessed, including data cleaning (removing outliers and correcting erroneous data), filling missing values (using mean, mode or interpolation), data standardization (normalizing numerical data and format conversion), and time alignment (ensuring that all data are aligned with a unified timestamp) to improve data accuracy and consistency;

[0061] Next, data is associated based on the compute node ID, task number, and timestamp, merging data from different sources and eliminating redundant information to ensure data uniqueness and integrity.

[0062] Then, the processed data is stored in a relational database, making it queryable, indexable, and scalable for efficient subsequent use.

[0063] Finally, data quality is ensured through data consistency checks and integrity verification, and the data storage structure is optimized according to business needs, such as establishing indexes, partitioned storage, and data compression to improve query efficiency and provide high-quality data support for subsequent machine learning analysis, task scheduling optimization, and rendering resource allocation.

[0064] Extract key indicators reflecting the overlap of random seed initialization methods of distributed nodes from the data set, and perform statistical analysis and visualization on the extracted key indicators to quantify the risk of random seed overlap;

[0065] Key indicators reflecting the overlap in the random seed initialization methods of distributed nodes are extracted from the data set. The extracted indicators include the probability of collision of hash values calculated based on the hash function when the node assigns the random seed and the numerical distribution of the pseudo-random number generator (PRNG) at each computing node. The probability of collision of hash values calculated based on the hash function when the node assigns the random seed and the numerical distribution of the pseudo-random number generator (PRNG) at each computing node are statistically analyzed and visualized under the detection window to generate a seed hash collision reference value and a PRNG sampling distribution consistency reference value respectively. The random seed overlap risk is quantified by the seed hash collision reference value and the PRNG sampling distribution consistency reference value.

[0066] When nodes are assigned random seeds, the probability of collisions in hash values calculated using hash functions increases significantly, indicating a risk of overlap in random seed initialization. In distributed rendering environments, random seeds are used to initialize pseudo-random number generators (PRNGs), which generate identical random number sequences given the same seed input. If hash values collide when initializing random seeds for multiple compute nodes, this means they may have been assigned the same or highly similar random seeds, resulting in overlapping random number sequences. Because rendering tasks rely on random numbers to determine key parameters such as ray tracing paths, global illumination calculations, and shadow generation, overlapping random seeds can directly lead to similar sampling patterns across tiles, causing periodic noise, lighting errors, and non-uniformity in path tracing sampling. Furthermore, an increase in the seed hash collision rate may also reflect deficiencies in the random seed generation strategy, such as a too-small seed generation range, an allocation algorithm that fails to consider global distribution, or low random entropy, further increasing the risk of randomness loss. Therefore, when the hash collision reference value increases significantly, it means that multiple computing nodes may share the same random number pattern, and there is a risk of overlapping in the random seed initialization method. It is necessary to optimize through dynamic seed allocation, task rescheduling, etc. to ensure the uniformity and independence of the random number distribution.

[0067] The specific steps for generating a seed hash collision reference value by statistically analyzing and visualizing the probability of hash value collisions calculated based on the hash function when the node assigns random seeds are as follows:

[0068] Within the detection window, there are N computing nodes. Each computing node generates a random number when a random seed is assigned. A hash function (such as SHA-256 or MurmurHash) is then used to map the random number into a fixed-length hash value. To calculate the hash collision rate, the hash values of all computing nodes are compared to determine whether the nodes generate the same hash value. The hash collision matrix is defined as follows:

[0069]

[0070] Where C i，j is a hash collision matrix, which is a binary state variable indicating whether the hash values of computing node i and computing node j are the same, i.e., the elements of the hash collision matrix, H i and H j The hash values of the random seeds generated for compute nodes i and j respectively;

[0071] C i，j =1,H i and H j Hash collision occurs, that is, their random seeds are the same;

[0072] C i，j =0, indicating that H i and H j The generated hash values are different.

[0073] Calculate the total number of collisions within the detection window. The calculation expression is as follows:

[0074]

[0075] Where P is the total number of hash collisions that occur between all computing nodes;

[0076] After obtaining the total number of collisions P, the severity of the hash collision is quantified and a seed hash collision reference value is generated. The generation formula is as follows:

[0077]

[0078] Where SHC is the seed hash collision reference value, H bin (H i ) is the hash value H of computing node i i Convert to binary vector, H bin (H i+1 ) is the hash value H of computing node i+1 i+1 Convert to a binary vector.

[0079] The Seed Hash Collision Reference Value, generated by statistically analyzing and visualizing the probability of collisions among hash values calculated using a hash function when assigning random seeds to nodes within a monitoring window, indicates a higher risk of overlap in random seed initialization. Conversely, a lower Seed Hash Collision Reference Value indicates a lower risk of overlap in random seed initialization. This is because the Seed Hash Collision Reference Value is a metric derived by analyzing the probability of collisions among hash values calculated using a hash function when assigning random seeds to nodes within the monitoring window. An increase in this metric indicates that multiple nodes generated the same or very similar hash values when initializing random seeds, potentially leading to these nodes sharing the same random seed. Overlapping random seeds can lead to different nodes using the same random number sequence during rendering, potentially causing issues such as periodic noise, uneven lighting, and uneven sampling in the rendered results.

[0080] Uneven distribution of pseudo-random number generator (PRNG) values across compute nodes may indicate a risk of random seed initialization overlap. From the perspective of random seed initialization overlap, the random number sequence generated by the PRNG is entirely determined by the initial seed. If multiple compute nodes are initialized using the same or highly similar seeds, the random number sequences generated by their PRNGs will also be highly similar or identical. In this case, even if different compute nodes are running at different times or in different tasks, their sampling distributions may still overlap, causing the value distribution to tend toward specific patterns rather than the ideal uniform random distribution. Furthermore, PRNGs inherently exhibit periodicity. If the seed distribution is inappropriate, multiple compute nodes may enter the same pseudo-random number cycle, resulting in periodic sampling artifacts or reduced randomness in certain computational regions. This not only affects the sampling uniformity of path tracing rendering but can also lead to deviations in global illumination calculations, ultimately affecting the consistency of rendering results. Therefore, when the numerical distribution of PRNG in each computing node is abnormally uneven, it is necessary to detect whether there is a risk of random seed overlap, and adopt strategies such as dynamic seed perturbation, seed distribution optimization, or low-discrepancy sequence replacement PRNG to ensure true randomness and improve rendering accuracy and stability.

[0081] The specific steps for statistically analyzing and visualizing the distribution of pseudo-random number generator values at each computing node within the detection window to generate a PRNG sampling distribution consistency reference value are as follows:

[0082] First, extract the pseudo-random number sequence from the PRNG sampling data of each computing node and calculate the numerical distribution characteristic entropy to measure the degree of dispersion of the random number distribution of different nodes. The higher the entropy, the more uniform the random number distribution. The lower the entropy, the less randomness there is and the risk of seed overlap. The calculation expression is as follows:

[0083]

[0084] Where, X f is the numerical distribution characteristic entropy, which indicates the distribution diversity of PRNG sampling values within a given window and is used to measure whether the random number distribution of different nodes is convergent. H is the number of discrete intervals of the PRNG sampling value of the calculation node, and P q is the normalized proportion of pseudo-random numbers in the qth numerical interval, calculated as the frequency of occurrence of the sampled values in the interval, and λ is a very small positive number (e.g. 10 -10 ), used to avoid zero values in logarithmic operations and improve calculation stability;

[0085] When obtaining the numerical distribution characteristic entropy X of each computing node fFinally, to further quantify the similarity of random number distributions between multiple computing nodes, it is necessary to calculate the PRNG sampling distribution consistency reference value, which is used to measure whether the pseudo-random number generation patterns between different nodes are highly similar. The calculation expression is as follows:

[0086]

[0087] Where PSDU is the reference value of PRNG sampling distribution consistency, M is the total number of computing nodes, that is, the number of computing nodes in the distributed rendering environment, and B j，k is the overlap of PRNG sampling distribution between computing nodes j and k, and They are the numerical distribution characteristic entropy of computing node j and computing node k, and ∈ is a very small positive number (such as 10 -10 ), prevent the denominator from being zero and improve calculation stability, It is a normalization factor used to prevent the influence of the numerical distribution characteristic entropy between computing nodes on the similarity calculation, making the calculation more stable.

[0088] The PRNG Sampling Distribution Consistency Reference Value, generated by statistically analyzing and visualizing the distribution of pseudo-random number generator values across compute nodes within a monitoring window, generally indicates a risk of overlap in random seed initialization. Conversely, a smaller value indicates relatively independent initialization and a lower risk of overlap. In a distributed rendering environment, pseudo-random number generators (PRNGs) rely on random seeds for initialization to generate pseudo-random sequences. If the random seeds of multiple compute nodes overlap or are highly similar, the random sequences generated by their PRNGs will tend to be identical, potentially leading to highly similar sampling paths and results when performing ray tracing, global illumination calculations, or sampling. The PRNG Sampling Distribution Consistency Reference Value measures the similarity of the pseudo-random number distributions across multiple compute nodes within a monitoring window. A high value indicates consistent random number distribution patterns across multiple compute nodes, suggesting they may be using the same or highly similar seeds. This can lead to insufficient randomness and can cause issues such as periodic sampling artifacts, global illumination deviations, and fixed noise distribution patterns. On the contrary, if the PRNG sampling distribution consistency reference value is low, it means that the random number sequences of different computing nodes have good discreteness and independence, the random seed initialization is more reasonable, the randomness is higher, and it is not easy to produce overlapping risks.

[0089] Intelligently predict whether there is a risk of random seed overlap on distributed rendering nodes using a pre-trained machine learning model;

[0090] The seed hash collision reference value and PRNG sampling distribution consistency reference value that have undergone statistical analysis and visualization are input into a pre-trained machine learning model. The random seed overlap risk coefficient is generated by the machine learning model, and the random seed overlap risk coefficient is used to intelligently predict whether there is a random seed overlap risk on the distributed rendering node.

[0091] A pre-trained machine learning model (Pre-trained Machine Learning Model) refers to a model that has been trained with a large amount of historical data and has optimized parameters before its formal application. It can be directly used for analysis and prediction of new data without rebuilding and training the entire model. In a distributed rendering environment, the main task of the model is to generate a random seed overlap risk coefficient by reasoning based on key indicators obtained from statistical analysis (such as seed hash collision reference value and sampling distribution consistency reference value). During the training phase, the model uses a large amount of rendering node data in different scenarios to learn the relationship between the random seed initialization method and the final numerical distribution, so as to have the ability to identify under what circumstances seeds may overlap and under what circumstances sampling patterns may be abnormally similar. Pre-trained models can use supervised learning (such as classification or regression models) or unsupervised learning (such as clustering or anomaly detection models). The specific choice depends on the characteristics of the data and the target application.

[0092] In actual applications, the machine learning model inputs new data (such as the seed hash collision index and sampling distribution consistency index after statistical analysis) and uses the trained model parameters to calculate a random seed overlap risk coefficient. This coefficient is used to evaluate whether the current distributed rendering system has a random seed overlap problem. Since the model has been trained and optimized on a large amount of data, it can quickly adapt to new data and provide highly accurate prediction results. In addition, the model can be continuously updated and adaptively optimized, that is, incremental training is performed through new rendering task data to continuously improve prediction capabilities. The pre-trained model can not only automatically detect high-risk nodes, but also provide optimization suggestions, such as whether the random seed initialization method needs to be adjusted, whether the computing tasks need to be redistributed, etc., thereby helping to optimize the stability and rendering quality of the distributed rendering system.

[0093] The machine learning model is not limited here. Any machine learning model that can perform a comprehensive analysis of the seed hash collision reference value SHC and the PRNG sampling distribution consistency reference value PSDU to generate a random seed overlap risk coefficient RSOR can be used. In order to realize the technical solution of the present invention, the present invention provides a specific implementation method.

[0094] The random seed overlap risk coefficient RSOR generation formula is as follows; RSOR = n1·SHC+n2·PSDU, where n1 and n2 are the preset proportional coefficients of the seed hash collision reference value SHC and the PRNG sampling distribution consistency reference value PSDU, respectively, and both n1 and n2 are greater than 0.

[0095] The preset proportional coefficient here refers to a weight factor, which is used to balance the relative importance of the seed hash collision reference value SHC and the PRNG sampling distribution consistency reference value PSDU in the final risk assessment when calculating the random seed overlap risk coefficient PRSO.

[0096] n1 and n2 are preset proportional coefficients used to adjust the contribution of the seed hash collision reference value SHC and the PRNG sampling distribution consistency reference value PSDU to the random seed overlap risk coefficient RSOR calculation result. If a certain component (such as SHC) plays a more decisive role in the random seed overlap risk, it can be given a higher weight (i.e., n1>n2); if the PRNG sampling distribution consistency has a greater impact on the final result, the value of n2 can be appropriately increased.

[0097] The values of these coefficients are typically determined through empirical settings or machine learning optimization to ensure that the RSOR accurately reflects the true extent of seed overlap risk. In addition, n1 and n2 must be greater than 0 to ensure that both SHC and PSDU have a real impact on the RSOR calculation and are not ignored.

[0098] It can be seen from the random seed overlap risk coefficient that the larger the seed hash collision reference value generated after statistical analysis and visualization of the probability of collision of hash values calculated based on the hash function when the node assigns random seeds under the detection window, and the larger the PRNG sampling distribution consistency reference value generated after statistical analysis and visualization of the numerical distribution of the pseudo-random number generator at each computing node under the detection window, the larger the random seed overlap risk coefficient generated when the pre-trained machine learning model is used to intelligently predict whether there is a random seed overlap risk for the distributed rendering node, indicating a greater probability of whether there is a random seed overlap risk for the distributed rendering node. Conversely, the smaller the probability of whether there is a random seed overlap risk for the distributed rendering node.

[0099] The random seed overlap risk coefficient generated by the pre-trained machine learning model when intelligently predicting whether a distributed rendering node has random seed overlap risk is compared with the pre-set random seed overlap risk coefficient reference threshold to identify whether a distributed rendering node has random seed overlap risk. The specific steps are as follows:

[0100] If the random seed overlap risk coefficient is greater than a preset random seed overlap risk coefficient reference threshold, the distributed rendering node is classified as having random seed overlap risk; if the random seed overlap risk coefficient is less than or equal to the preset random seed overlap risk coefficient reference threshold, the distributed rendering node is classified as not having random seed overlap risk.

[0101] When the machine learning model determines that there is a risk of overlapping random seeds in tiles, it first generates a dynamic random seed based on the tile coordinates, frame index, and compute node ID to change the random number sequence of each tile. Secondly, it redistributes tiles with high overlap to different compute nodes to disperse the distribution range of the random number seed. At the same time, it dynamically adjusts the tile size and optimizes the computational area division to change the sampling pattern between adjacent tiles.

[0102] The purpose of this step is to reduce the risk of random seed overlap, ensure the diversity of path tracing samples, optimize rendering quality, and avoid problems such as periodic noise, lighting anomalies, and inter-frame flickering caused by repeated random numbers.

[0103] First, generating a dynamic random seed based on tile coordinates, frame index, and compute node ID ensures that different tiles have independent random number sequences, reducing the problem of pseudo-random number distribution similarity caused by seed duplication. If multiple tiles share the same or highly similar random seeds, the random sequences generated by their PRNGs will also be highly similar, resulting in periodic artifacts in path tracing sampling directions, lighting calculations, shadow mapping, and other processes. Therefore, by introducing a dynamic random seed, we can break the fixed random number pattern, improve the randomness of global illumination, reflection, and refraction calculations, and reduce the predictability of noise distribution.

[0104] Secondly, redistributing highly overlapping Tiles to different compute nodes to disperse the distribution of random number seeds can reduce the risk of adjacent Tiles generating similar random numbers on the same compute node. Since compute nodes may use similar seed initialization methods across consecutive frames or adjacent Tiles, if left unadjusted, this can lead to inconsistent lighting calculations between frames, resulting in noticeable inter-frame flickering. Through intelligent task scheduling and dynamic tile allocation, we can balance the seed distribution across compute nodes, prevent localized random number aggregation, and improve the global stability of rendering.

[0105] Finally, dynamically adjusting the tile size and optimizing the computational area division to change the sampling pattern between adjacent tiles can reduce computational redundancy between tiles and optimize the spatial distribution of path tracing. If the tiles are too small, the compute nodes may generate similar random seeds within multiple small areas, causing the sampling patterns between adjacent tiles to converge and resulting in spatial unevenness in lighting, shadow, and reflection calculations. On the other hand, if the tiles are too large, this can lead to an unbalanced computational load, affecting rendering efficiency. Therefore, dynamic tile adjustment can ensure that the similarity between local tiles in complex lighting scenes is reduced, improving the randomness of the computation, while optimizing the allocation of computing resources and improving rendering efficiency.

[0106] In summary, this step ensures a balanced distribution of random numbers in the distributed rendering system through dynamic random seed generation, intelligent task scheduling, and tile size optimization. This improves the diversity of path tracing calculations, reduces periodic sampling artifacts, and enhances the physical consistency of global illumination, reflection, and shadow calculations, thereby improving the realism and stability of the final rendering results.

[0107] When the machine learning model determines that the random seeds of tiles are at risk of overlapping, it first generates a dynamic random seed based on the tile coordinates, frame index, and compute node ID. Then, it redistributes the tiles with high overlap to different compute nodes. The specific steps for dynamically adjusting the tile size are as follows:

[0108] When the machine learning model predicts that there is a risk of random seed overlap in tiles, that is, when the random seed overlap risk coefficient RSOR is greater than the preset random seed overlap risk coefficient reference threshold, the random seed for each tile should be dynamically generated first to change the random sequence of each tile and fundamentally reduce the random seed overlap risk. The calculation expression is as follows:

[0109]

[0110] In the formula, Seed dynamic Is a dynamic random seed value, each tile is unique, X tile and Y tile The horizontal and vertical positions of the current Tiles in the rendered image, Frame idx Is the frame index of the current rendering task, indicating the animation frame number to which the current Tiles belongs. Node ID Is the compute node ID of the current compute Tiles, which is the unique number of each compute node, RSOR refis the random seed overlap risk coefficient reference threshold, e is a natural base, and Hash[·] is a hash function (such as SHA-256, MurmurHash, etc.) used to ensure the security and distribution uniformity of generated seeds;

[0111] By generating dynamic random seeds based on tile coordinates, frame index and compute node ID, different tiles use independent random number sequences, thus breaking the sampling pattern similarity caused by random seed overlap and reducing the generation of periodic artifacts. Secondly, the random seed overlap risk coefficient RSOR and its reference threshold RSOR are introduced. ref Exponential regulation is performed to dynamically adjust the calculation method of the random seed based on the risk level, making the distribution of random number seeds more dispersed in high-risk situations, thereby optimizing the uniformity of path tracing sampling and improving the stability of rendering calculations.

[0112] After dynamic seed generation is completed, the probability of random number sequence repetition is further reduced by dispersing the random number seed space, optimizing the spatial diversity of the sampling distribution, and reducing the similarity of sampling patterns between adjacent tiles. The calculation expression is as follows:

[0113] ,

[0114]

[0115] In the formula, Node new Is the identifier of the reallocated computing node, used to determine the node to which each high-risk Tiles should be rescheduled. current is the node identifier to which the current Tiles originally belonged, Φ is the golden section constant (≈1.618), which introduces the non-periodic characteristics of irrational numbers to effectively avoid simple periodic allocation, N total is the total number of computing nodes, used to ensure the validity of new node allocation, mod operation to ensure that the new node number falls within the valid node range, Size tile_new It is the dynamically adjusted Tiles size, which is used to change the sampling area of Tiles on the rendering screen, optimize the calculation area division, and reduce the spatial correlation of random number seeds. tile_original is the original default Tiles size, Γ is the regularization constant used to dynamically adjust the amplitude, which is used to avoid the denominator approaching 0 and ensure the stability of resizing.

[0116] By redistributing highly overlapping tiles to different compute nodes, the distribution of random seeds is effectively dispersed across compute nodes, reducing the duplication of random number seeds on the same node. At the same time, the tile size is dynamically adjusted to optimize the division of the computational area, ensuring more diverse sampling patterns between adjacent tiles, thereby reducing sampling similarity and further reducing the risk of random seed overlap.

[0117] The present invention successfully solves the problem of regular distribution of noise caused by random seed overlap in distributed rendering by real-time monitoring of distributed rendering nodes, dynamic generation of random seeds, and intelligent task scheduling, thereby significantly improving the quality of rendered images. First, by real-time monitoring and preprocessing of the initialization data information of each computing node, the system can accurately obtain the random seed distribution of the node and quantify the potential overlap risk through statistical analysis. Subsequently, with the help of a trained machine learning model, it intelligently predicts whether there is a risk of seed overlap, and dynamically adjusts the random seed generation method and Tiles allocation strategy when the risk occurs. Specifically, the system avoids repeated sampling patterns by generating new random seeds based on Tiles coordinates, frame index and computing node ID, while adjusting the Tiles size and computing area division to optimize computing efficiency. This solution ensures that the random sampling sequence of each Tiles is independent, eliminates the regular distribution of noise, improves rendering quality, and improves the efficiency and stability of distributed computing.

[0118] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0119] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0120] It should be noted that, in this document, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0121] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0122] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0123] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0124] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0125] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0126] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0127] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. A distributed task scheduling method in a cloud computing environment, characterized in that: The following steps are involved: First, all distributed rendering nodes are monitored in real time, and the initialization data information generated by each distributed rendering computing node is obtained in real time during the monitoring process; After obtaining the initialization data information of each node, the obtained initialization data information is preprocessed, and the preprocessed data is integrated to establish a structured data set; Extract key indicators reflecting the overlap of random seed initialization methods of distributed nodes from the data set, and perform statistical analysis and visualization on the extracted key indicators to quantify the risk of random seed overlap; Intelligently predict whether there is a risk of random seed overlap on distributed rendering nodes using a pre-trained machine learning model; When the machine learning model determines that there is a risk of overlapping random seeds in tiles, it first generates a dynamic random seed based on the tile coordinates, frame index, and compute node ID to change the random number sequence of each tile. Secondly, it redistributes tiles with high overlap to different compute nodes to disperse the distribution range of the random number seeds. At the same time, it dynamically adjusts the tile size and optimizes the computational area division to change the sampling pattern between adjacent tiles.

2. The distributed task scheduling method in a cloud computing environment according to claim 1, characterized in that: In a distributed rendering environment, after obtaining the initialization data information of each node, it is preprocessed and a structured data set is established. The specific steps are as follows: First, collect initialization data from logs, API interfaces, and monitoring systems of each computing node, including hardware information, software version, load status, network connection status, and random number seeds, to ensure comprehensive data coverage. Secondly, the acquired data is preprocessed, including data cleaning, missing value filling, data standardization, and time alignment to improve data accuracy and consistency; Next, data is associated based on the compute node ID, task number, and timestamp, merging data from different sources and eliminating redundant information to ensure data uniqueness and integrity. Then, the processed data is stored in a relational database, making it queryable, indexable, and scalable for efficient subsequent use. Finally, data consistency and integrity checks are performed to ensure data quality, and the data storage structure is optimized according to business needs.

3. The distributed task scheduling method in a cloud computing environment according to claim 1, characterized in that: Key indicators reflecting the overlap in the random seed initialization methods of distributed nodes are extracted from the data set. The extracted indicators include the probability of collision of hash values calculated based on the hash function when the nodes assign random seeds and the numerical distribution of the pseudo-random number generator in each computing node. The probability of collision of hash values calculated based on the hash function when the nodes assign random seeds and the numerical distribution of the pseudo-random number generator in each computing node are statistically analyzed and visualized under the detection window, and a seed hash collision reference value and a PRNG sampling distribution consistency reference value are generated respectively. The random seed overlap risk is quantified by the seed hash collision reference value and the PRNG sampling distribution consistency reference value.

4. The distributed task scheduling method in a cloud computing environment according to claim 3, characterized in that: The specific steps for generating a seed hash collision reference value by statistically analyzing and visualizing the probability of hash value collisions calculated based on the hash function when the node assigns random seeds are as follows: Within the detection window, there are N computing nodes. Each computing node generates a random number when a random seed is assigned. The random number is then mapped to a fixed-length hash value using a hash function. To calculate the hash collision rate, the hash values of all computing nodes are compared to determine whether the nodes generate the same hash value. The hash collision matrix is defined as follows: Where C i，j is a hash collision matrix, which is a binary state variable indicating whether the hash values of computing nodes i and j are the same. i and H j The hash values of the random seeds generated for compute nodes i and j respectively; Calculate the total number of collisions within the detection window. The calculation expression is as follows: Where P is the total number of hash collisions that occur between all computing nodes; After obtaining the total number of collisions P, the severity of the hash collision is quantified and a seed hash collision reference value is generated. The generation formula is as follows: Where SHC is the seed hash collision reference value, H bin (H i ) is the hash value H of computing node i i Convert to binary vector, H bin (H i+1 ) is the hash value H of computing node i+1 i+1 Convert to a binary vector.

5. The distributed task scheduling method in a cloud computing environment according to claim 3, characterized in that: The specific steps for statistically analyzing and visualizing the distribution of pseudo-random number generator values at each computing node within the detection window to generate a PRNG sampling distribution consistency reference value are as follows: First, extract the pseudo-random number sequence from the PRNG sampling data of each computing node and calculate the numerical distribution characteristic entropy to measure the discrete degree of random number distribution of different nodes. The calculation expression is as follows: Where, X f is the characteristic entropy of the numerical distribution, H is the number of discrete intervals of the PRNG sampling value of the computing node, and P q is the normalized proportion of pseudo-random numbers in the qth numerical interval, and λ is a very small positive number used to avoid zero values in logarithmic operations; When obtaining the numerical distribution characteristic entropy X of each computing node f Finally, in order to further quantify the similarity of random number distribution between multiple computing nodes, it is necessary to calculate the PRNG sampling distribution consistency reference value. The calculation expression is as follows: Where PSDU is the reference value of PRNG sampling distribution consistency, M is the total number of computing nodes, that is, the number of computing nodes in the distributed rendering environment, and B j，k is the overlap of PRNG sampling distribution between computing nodes j and k, and are the numerical distribution characteristic entropy of computing node j and computing node k respectively, ∈ is a very small positive number to prevent the denominator from being zero, is the normalization factor.

6. The distributed task scheduling method in a cloud computing environment according to claim 3, characterized in that: The seed hash collision reference value and PRNG sampling distribution consistency reference value that have undergone statistical analysis and visualization are input into a pre-trained machine learning model. The random seed overlap risk coefficient is generated by the machine learning model, and the random seed overlap risk coefficient is used to intelligently predict whether there is a random seed overlap risk on the distributed rendering node.

7. The distributed task scheduling method in a cloud computing environment according to claim 6, characterized in that: The random seed overlap risk coefficient generated by the pre-trained machine learning model when intelligently predicting whether a distributed rendering node has random seed overlap risk is compared with the pre-set random seed overlap risk coefficient reference threshold to identify whether a distributed rendering node has random seed overlap risk. The specific steps are as follows: If the random seed overlap risk coefficient is greater than a preset random seed overlap risk coefficient reference threshold, the distributed rendering node is classified as having a random seed overlap risk; If the random seed overlap risk coefficient is less than or equal to a preset random seed overlap risk coefficient reference threshold, the distributed rendering node is classified as having no random seed overlap risk.

8. The distributed task scheduling method in a cloud computing environment according to claim 7, characterized in that: When the machine learning model determines that the random seeds of tiles are at risk of overlapping, it first generates a dynamic random seed based on the tile coordinates, frame index, and compute node ID. Then, it redistributes the tiles with high overlap to different compute nodes. The specific steps for dynamically adjusting the tile size are as follows: When the machine learning model predicts that there is a random seed overlap risk for tiles, that is, when the random seed overlap risk coefficient RSOR is greater than the preset random seed overlap risk coefficient reference threshold, the random seed for each tile should be dynamically generated first. The calculation expression is as follows: In the formula, Seed dynamic Is a dynamic random seed value, each tile is unique, X tile and Y tile The horizontal and vertical positions of the current Tiles in the rendered image, Frame idx Is the frame index of the current rendering task, indicating the animation frame number to which the current Tiles belongs. Node ID Is the compute node ID of the current Tiles calculation, RSOR ref is the random seed overlap risk coefficient reference threshold, e is the natural base, and Hash[·] is the hash function; After dynamic seed generation is completed, the probability of random number sequence repetition is further reduced by dispersing the random number seed space, optimizing the spatial diversity of the sampling distribution, and reducing the similarity of sampling patterns between adjacent tiles. The calculation expression is as follows: In the formula, Node new Is the identifier of the reallocated computing node, Node current is the node identifier to which the current Tiles originally belonged, Φ is the golden section constant, N total is the total number of computing nodes, mod is a modulo operation to ensure that the new node number falls within the valid node range, Size tile_new Is the size of the tiles after dynamic adjustment, Size tile_original is the original default tile size, and Γ is the regularization constant used to dynamically adjust the amplitude.

Images