Real-time dynamic scheduling method for heterogeneous multipath remote procedure call channel

Abstract

The invention discloses a real-time dynamic scheduling method for a heterogeneous multipath remote procedure call channel. According to the method, a multi-dimensional path scoring mechanism fusing round-trip delay, packet loss rate and bandwidth information is provided for solving the problem that link characteristics of multi-channel parallel communication in a distributed system are remarkably different. Through the scoring mechanism, the system can evaluate the communication quality of each available channel in real time, and dynamically select an optimal path for data transmission. The method comprises the steps of constructing a heterogeneous network model, collecting performance indexes of all channels, calculating path scores, and adaptively distributing data streams based on scoring results. Under the condition that the path performance fluctuates greatly or the quality of the main path is reduced, the scheduling method provided by the invention can effectively improve the overall throughput rate of the system, and has good adaptability and robustness.

Description

Technical Field

[0001] The present invention relates to network communication and remote call technology, and more specifically, to a dynamic scheduling method for remote procedure call channels based on multi-path state perception, aiming to improve data transmission efficiency and system stability in a heterogeneous network environment. Background Art

[0002] With the large-scale deployment of network measurement and data collection platforms, crowdsourced measurement systems are playing an increasingly important role in evaluating network performance across different regions and operators. In such systems, clients often need to establish connections with multiple remote measurement nodes simultaneously for concurrent communication. Remote Procedure Calls (RPCs) are a common method for implementing these connections. Each RPC channel represents a logical link with independent network characteristics, such as bandwidth, latency, and packet loss rate.

[0003] However, due to differences in the physical location of links and the dynamic nature of the underlying network environment, significant performance differences often exist between different RPC channels. Traditional systems often use static channel selection strategies, which are unable to adapt to fluctuations in network conditions in real time. This makes it difficult to effectively schedule and allocate data transmission tasks across multiple links, limiting the system's overall throughput and leading to performance degradation during network congestion or link degradation.

[0004] Furthermore, in some highly heterogeneous or unstable network environments, single-channel transmission can significantly reduce communication efficiency and exacerbate network bottlenecks. Some research has attempted to introduce multi-path concurrency strategies to improve data acquisition capabilities. Round-Robin Scheduling (RRS) is widely adopted due to its simplicity, but it lacks link state awareness and often causes load shifts when path performance is uneven. Random Path Scheduling (RPS) avoids path fixation by introducing randomness, but suffers from performance uncertainty when the network fluctuates significantly. While Weighted Round-Robin Scheduling (WRR) can allocate tasks based on static weights, it lacks the ability to adapt to changes in link status.

[0005] To address these issues, some researchers have proposed the Delay-Aware Packet Scheduling (DAPS) algorithm. This method prioritizes paths based on round-trip delay, improving system responsiveness and scheduling efficiency. However, DAPS only considers latency, ignoring key performance factors such as bandwidth and packet loss rate. It also lacks a dynamic adjustment mechanism for scoring parameters, limiting its applicability and robustness in complex and changing network environments. Summary of the Invention

[0006] This paper proposes a dynamic scheduling method for heterogeneous multipath remote procedure call (RPC) channels. This method perceives the real-time network status of each channel and comprehensively considers metrics such as latency, packet loss rate, and available bandwidth to score and rank each RPC channel, thereby enabling adaptive data scheduling and transmission path selection. Through a multi-dimensional path evaluation mechanism and a dynamic weight allocation strategy, this method improves system throughput and stability in complex and changing network environments, addressing the static and poorly adaptable path selection issues of existing methods.

[0007] The objectives of the present invention are achieved through the following technical solutions: Step 1. In the initialization phase, a scheduling controller is configured for each RPC channel to collect path status and scheduling information; Step 2. In the path status perception phase, the network performance indicators of each channel are monitored in real time, including round-trip delay, packet loss rate and bandwidth; Step 3. In the path scoring phase, a comprehensive score of each path is calculated based on the collected performance indicators to reflect its current transmission quality; Step 4. In the weight calculation phase, the scores of each channel are normalized to obtain the corresponding data allocation ratio as the basis for scheduling; Step 5. In the scheduling execution phase, the scheduler distributes data packets according to the calculated ratio, implements a transmission strategy of prioritizing high-quality paths and avoiding low-quality paths, and continues to collect feedback information after the transmission is completed to support the next round of dynamic scheduling. Step 6: In the data transmission and feedback update phase, the system sends the data packet to the selected RPC channel and updates the path performance indicators in real time based on the information returned by the receiving end, so that the scheduler can adjust the strategy in time to achieve continuous and efficient scheduling of heterogeneous channels. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 This is an architecture diagram of the heterogeneous multi-path remote procedure call scheduler described in an embodiment of the present invention. Specific implementation

[0009] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments, and do not constitute a limitation of the present invention. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0010] The dynamic scheduling method for heterogeneous multi-path RPC channels described in an embodiment of the present invention includes the following steps: 1) Initialization phase: Establish multiple RPC channels on the client side to form a path set Each path Corresponding to an independent RPC channel, and bound to a specific channel identifier. In this stage, the path scoring parameters are pre-configured for each path, including the round-trip delay weight. , packet loss weight , bandwidth enhancement factor , and the basic smoothing coefficient , used for subsequent comprehensive scoring and scheduling decisions. 2) Path state awareness phase: Whenever the client sends a data packet to the remote service node, the sending timestamp is recorded , record the receiving timestamp after receiving the confirmation response , calculate the round trip delay as: To prevent jitter and instantaneous fluctuations from affecting the overall scheduling effect, an exponentially weighted moving average is used to update the smoothing delay: The packet loss rate is calculated by periodically counting the number of packets sent. and the number of successful confirmations Calculation yields: Bandwidth estimation is based on the time window Number of data bytes received within the confirmation Calculate available bandwidth: . 3) Path scoring stage: After obtaining the three performance indicators of each path, calculate the comprehensive path score : ,in A small positive number introduced to prevent the denominator from being zero. 4) Weight normalization stage: After normalizing the scores of each path, the data distribution ratio is obtained. : Make , used to guide the scheduler's proportional distribution strategy, where represents the sum of all path scores, Indicates the data allocation ratio that the path should bear in this scheduling cycle, ensuring that the sum of the scheduling probabilities of each path is 1. 5) Scheduling decision stage: After the path allocation ratio is calculated, the system scheduling core will Trigger the actual transmission scheduling operation. Use the weighted random algorithm to allocate the current data flow to the corresponding path according to the ratio. , prioritizing the use of high-quality paths. This mechanism effectively implements path priority control, enabling rapid switching and traffic adjustment in the face of dynamic changes in path performance, thereby avoiding congestion, packet loss, or increased latency caused by long-term reliance on inefficient paths. 6) Data transmission and feedback update phase: The scheduler sends the data packet to the selected RPC channel. After the receiving end confirms the response, the scheduler immediately updates the delay in the path state cache. , packet loss and bandwidth This periodic performance update mechanism ensures that the scheduling strategy adapts to network status changes in real time, improving the stability and throughput of the entire system in heterogeneous network environments. This paper proposes a dynamic scheduling method for heterogeneous multipath RPC channels. Addressing the poor adaptability and unstable performance of existing path selection strategies in complex network environments, a multidimensional path scoring model is constructed that integrates latency, packet loss rate, and bandwidth. Through a real-time feedback mechanism and a dynamic weight adjustment strategy, the flexibility of data scheduling and the overall throughput performance of the system are effectively improved. Furthermore, the scheduling architecture design is versatile and scalable, adapting to various distributed network measurement scenarios and possessing practical application value for improving the stability of RPC communication systems in heterogeneous networks.

Claims

1. A real-time dynamic scheduling method for heterogeneous multi-path remote procedure call channels, characterized in that: The method is applicable to transmitting data between multiple remote procedure call channels with heterogeneous performance, and comprises the following steps: Step 1: During the initialization phase, a scheduling controller is configured for each RPC channel to collect path status and scheduling information, providing the necessary data basis for subsequent scheduling calculations and path selection. Step 2: During the path state awareness phase, the scheduler monitors the performance of multiple RPC channels, obtaining the real-time round-trip delay, packet loss rate, and estimated available bandwidth for each channel to characterize the dynamic changes in link status. Step 3: Path scoring calculation. Based on the collected metrics, a multi-dimensional path scoring model is used to comprehensively score each heterogeneous channel. This scoring model combines bandwidth advantages with penalty factors for high-latency and high-packet-loss paths to achieve a unified quantitative assessment of link quality. Step 4: Normalization of allocation weights: normalize the score of each channel, calculate the corresponding data packet allocation ratio, and form a selection probability distribution that reflects the quality of the link; Step 5: In the path scheduling decision phase, the scheduler adaptively selects the target path from multiple RPC channels using a weighted sampling strategy based on the normalized weight results for the current round of data packet transmission. Step 6: Data transmission and feedback update phase: data packets are sent to the selected channel, and path performance indicators are updated in real time based on received feedback to ensure that subsequent scheduling strategies can respond to network status changes in a timely manner and achieve efficient dynamic scheduling of heterogeneous RPC channels.

2. The dynamic scheduling method for heterogeneous multi-path RPC channels according to claim 1, characterized in that: The path state awareness phase of step 1 includes: The sending time of each data packet and the receiving time of the corresponding response are recorded for each RPC channel, and the original timestamps and data interaction information during the communication process are collected. This information is not used directly to calculate performance indicators, but serves as basic data for evaluating parameters such as round-trip delay, packet loss rate, and bandwidth in subsequent steps. By continuously recording and updating this type of information, the system can historically track and dynamically observe the operating status of each channel, thereby supporting the adjustment of subsequent scheduling strategies and the optimization of path selection.

3. The dynamic scheduling method for heterogeneous multi-path RPC channels according to claim 1, characterized in that: The path state awareness phase of step 2 includes: The scheduler sends each RPC channel to the sender. Perform performance monitoring and record the The sending timestamp of each data packet and receipt confirmation timestamp , and the round-trip delay is calculated as: ,in, Indicates the Packets in the path The instantaneous round-trip delay on Indicates the sending timestamp. Indicates the receiving timestamp. To eliminate the impact of network jitter on the measurement results, the scheduler applies an exponentially weighted moving average update to the above delay: ,in, is a smoothing factor used to control the weighted ratio of historical delay to newly measured delay. Total number of data packets sent in the previous time window and the number of received confirmations , and the packet loss rate is calculated accordingly: ,in, Indicates the total number of packets sent, Indicates the number of packets successfully acknowledged. Total number of bytes transferred within the record , the estimated available bandwidth of the path is: ,in, Indicates the path bandwidth (in bps), is the amount of bytes transferred, The duration of the statistics window.

4. The dynamic scheduling method for heterogeneous multi-path RPC channels according to claim 1, characterized in that: The path score calculation phase of step 3 includes: The scheduler comprehensively evaluates the path based on the path's bandwidth, round-trip delay, and packet loss rate. The transmission quality of the ,is calculated as follows: in, For path Rating, To estimate bandwidth, is the delay after smoothing, is the packet loss rate, 、 are the weight coefficients of delay and packet loss respectively, is the bandwidth amplification factor, To prevent extremely small positive numbers whose denominator is zero.

5. The dynamic scheduling method for heterogeneous multi-path RPC channels according to claim 1, characterized in that: The allocation weight normalization stage of step 4 includes: The scheduler first sums the scores of all paths: , and then normalize the score of each path to get the path The data packet distribution ratio is: in, Indicates the data allocation ratio that the path should bear in this round of scheduling cycle. is the score of the path, Sum the scores for all paths.

6. The dynamic scheduling method for heterogeneous multi-path RPC channels according to claim 1, characterized in that: The path scheduling decision stage of step 5 includes: The scheduler allocates the same amount of traffic to each path , using a weighted random sampling strategy, from the path set Select the target path , used for the current data packet transmission task. Among them, the selection of the target path satisfies the priority relationship reflected by the path score and the allocation ratio, and maintains the flexibility and load balance of the scheduling process.

7. The dynamic scheduling method for heterogeneous multi-path RPC channels according to claim 1, characterized in that: The data transmission and feedback update phase of step 6 includes: The sender sends the data packet through the selected path After sending, the receiving end returns confirmation information, and the scheduler updates the parameters such as the sending time, receiving time, confirmation count and transmission byte volume of the path, and recalculates 、 、 Indicators are used to build a real-time feedback loop for path scoring and allocation decisions in the next scheduling cycle.

Images