Distributed process scheduling system with multi-service object cooperation

Abstract

The application belongs to the technical field of distributed cooperation, and discloses a multi-service object cooperative distributed process scheduling system, which comprises a ship end edge cooperation subsystem, a shore end global cooperation subsystem and a communication channel module; a distributed process scheduling engine decomposes a process into a directed acyclic graph node set, extracts data increments through a data incremental extraction module, generates a lightweight index data packet by using a local cooperation calculation module, calculates a scheduling strategy factor according to transmission delay and available bandwidth by means of a link adaptive scheduling module, suspends real-time transmission when being lower than a lower threshold and being full of a smooth window, constructs a priority queue temporary storage state confirmation token according to a dependence criticality factor, and preferentially sends when reaching a cooperation threshold and being full of a smooth window; and a state consistency synchronization module is used to update a global state, so that the application eliminates scheduling oscillation and head blocking caused by channel jitter, and guarantees fast convergence of a cooperative state.

Description

Technical Field

[0001] This invention relates to the field of distributed collaboration technology, and more specifically to a distributed process scheduling system for collaboration among multiple business objects. Background Technology

[0002] Existing multi-node collaborative scheduling systems typically break down complex business processes and distribute them to edge nodes and a central server for joint processing. This enables data querying, metric statistics, and status synchronization, as well as business flow and collaborative control between different network segments. Such systems mainly employ centralized cloud scheduling architectures or isolated edge monolithic architectures. Centralized cloud scheduling architectures require edge nodes to upload the collected time-series data to the central server in real time and in full. When network links are unstable or available bandwidth is limited, continuous transmission of large amounts of data can easily cause communication congestion. Isolated edge monolithic architectures, on the other hand, struggle to synchronize status with the central management node in a timely manner. This can easily lead to state disconnects in the collaborative processes between different regions. In scenarios where collaboration relies on satellite communication channels with asymmetric and intermittent fading characteristics, wave obstruction, weather changes, and satellite beam switching can cause the available bandwidth and transmission delay of the link to fluctuate frequently near the critical threshold. Existing technologies typically switch between real-time transmission mode and data storage mode based on a fixed link quality threshold. When the channel parameters fluctuate briefly near the critical threshold, the fixed threshold will be repeatedly triggered, causing the scheduling state machine to switch frequently and generate disordered oscillations. This can result in a large number of data packets being transmitted repeatedly, and may even cause the communication channel to deadlock.

[0003] Simply increasing the storage space of the local temporary queue or increasing the data transmission frequency cannot solve the problem of a large amount of high-frequency time-series log data continuously accumulating in the temporary queue. Expanding local hardware temporary storage capacity or hardware resources alone cannot resolve the technical contradiction between network instability and the system's high-performance requirements. Solutions need to be sought at the software control method and distributed workflow scheduling level. For example, Chinese invention patent CN115580622B discloses a workflow execution method, electronic device, and medium based on edge synchronization services. By setting up a synchronization service unit at the edge, node queries are completed using local template data when the cloud-edge link is disconnected, and the query is completed when the link is restored. After recovery, the unsynchronized data is transmitted to the cloud in batches according to the time sequence. However, the offline autonomy and data synchronization mechanism of this scheme are highly dependent on the binary network connectivity status determination. When applied to asymmetric fading channels where the link quality fluctuates frequently near the critical threshold due to wave obscuring or weather jitter, the binary state switching is very likely to cause disordered oscillations of the scheduling state machine and redundant data retransmission. Moreover, its data synchronization strategy only sends data in batches according to simple time sequence or quantitative batches, and fails to dynamically prioritize based on the complex dependencies and urgency between nodes in the distributed task topology. This makes high-priority status confirmation instructions very susceptible to head-of-line blocking by massive high-frequency log data, and cannot guarantee the rapid convergence of cross-network segment collaborative states.

[0004] Therefore, how to reduce the oscillation of the scheduling state machine caused by the jitter of dynamic fading channel parameters, avoid head-of-line blocking in the transmission queue, and enable the dependent states of distributed tasks to converge quickly has become the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a distributed process scheduling system for multi-service object collaboration to solve the problems of scheduling oscillations caused by channel jitter and delayed convergence of collaborative states in asymmetric intermittent network environments.

[0006] The specific technical solution adopted is as follows: A distributed process scheduling system for multi-business object collaboration, comprising: Both the ship-side edge collaboration subsystem and the shore-side global collaboration subsystem are equipped with local state storage modules and distributed process scheduling engines. The distributed process scheduling engine decomposes the business process into a set of directed acyclic graph task nodes containing dependencies, and the distributed process scheduling engine includes: The incremental data extraction module is used to perform incremental queries on the local state storage module based on business data query conditions to extract incremental time-series data. The local collaborative computing module is used to complete the statistical analysis of voyage indicators and the correlation calculation of operating conditions at the edge of the ship and generate lightweight intermediate indicator data packages. The link adaptive scheduling module calculates the scheduling strategy factor based on the transmission delay and available bandwidth of the communication channel module. When the scheduling strategy factor is lower than the preset switching threshold and the duration reaches the preset smoothing window, real-time transmission is suspended and switched to data storage mode. In data storage mode, a hierarchical priority queue is constructed based on the dependency criticality factor of the task nodes in the directed acyclic graph, and data packets containing status confirmation tokens are placed in the high-priority queue. When the scheduling strategy factor reaches the preset coordination threshold and the duration reaches the preset smoothing window, real-time transmission mode is restored and data packets in the high-priority queue are scheduled first. The state consistency synchronization module is used to receive lightweight intermediate indicator data packets, trigger shore-side business object linkage to update the global business state, and send the state confirmation token back to the ship-side edge collaboration subsystem.

[0007] Preferably, when the distributed process scheduling engine decomposes the set of task nodes in the directed acyclic graph, it extracts the dependency criticality factor of each task node in the topology. The dependency criticality factor is determined recursively based on the depth of subsequent dependency links and the urgency of synchronization of the task node in the topology. The distributed process scheduling engine generates a constraint vector for each task node that includes task identifier, set of predecessor dependencies, processing object attributes and data input relationships.

[0008] Preferably, the preset switching lower limit threshold is lower than the preset coordination threshold, and the preset switching lower limit threshold is set to 0.85; when the scheduling strategy factor decreases from greater than or equal to 1.0 to less than 1.0 but greater than or equal to 0.85, the link adaptive scheduling module maintains the cross-network segment real-time transmission mode; when the scheduling strategy factor is less than 0.85 and this state is not interrupted within the preset smoothing window, the link adaptive scheduling module suspends real-time transmission and switches to data temporary storage mode; wherein, the preset smoothing window is set to 15s.

[0009] Preferably, the ship-side edge collaboration subsystem includes operational condition data acquisition business objects, fuel consumption analysis business objects, and voyage execution business objects deployed on the ship; the shore-side global collaboration subsystem includes enterprise management decision-making business objects, dispatching and command business objects, and logistics docking business objects deployed on the shore-based server; the operational condition data acquisition business objects connect to the ship-side sensor interface module and collect main engine speed, speed, draft, and fuel flow data at a sampling period of 5 seconds; the fuel consumption analysis business objects receive the data collected by the operational condition data acquisition business objects and calculate the deviation between the average nautical mile fuel consumption index and engine efficiency.

[0010] Preferably, when the link adaptive scheduling module is in data temporary storage mode, it assigns data packets with a dependency criticality factor greater than or equal to 0.90 to a high-priority queue and data packets with a dependency criticality factor less than 0.90 to a low-priority queue. The high-priority queue stores status confirmation tokens and lightweight intermediate indicator data packets containing operating condition deviation indicators, while the low-priority queue stores high-frequency raw log incremental data packets. After resuming real-time transmission mode, the link adaptive scheduling module temporarily suspends the transmission of data packets in the low-priority queue before the high-priority queue is cleared.

[0011] Preferably, the data incremental extraction module uses a sliding window mechanism based on version vectors to extract time-series data increments; the data incremental extraction module monitors the tariff policy and real-time channel quality of the communication channel module, and reduces the sampling time span of the sliding window when the channel quality deteriorates, and expands the sampling time span of the sliding window when the channel quality improves.

[0012] Preferably, the lightweight intermediate indicator data package includes fuel consumption statistics for the voyage phase, operating condition deviation indicators, and hash check codes generated from local state data; the data volume of the lightweight intermediate indicator data package is less than 5% of the original time-series data acquisition volume; after receiving the lightweight intermediate indicator data package, the state consistency synchronization module compares the hash check codes, and updates the shore-side global state storage module after the comparison is successful.

[0013] Preferably, the local state storage module includes a circular temporary queue; when the available bandwidth of the communication channel module drops to 0kbps and the scheduling strategy factor is lower than the preset switching lower limit threshold, the link adaptive scheduling module writes the generated lightweight intermediate index data packet into the circular temporary queue; after the communication channel module recovers and the scheduling strategy factor reaches the preset coordination threshold again, the link adaptive scheduling module sequentially resends the data packets in the circular temporary queue according to the order of the temporary timestamps.

[0014] Preferably, after updating the global business status, the shore-side global collaboration subsystem calls the enterprise management decision-making business object to generate a speed adjustment instruction, and calls the scheduling and command business object to encapsulate the speed adjustment instruction into a status confirmation token and send it back to the voyage execution business object of the ship-side edge collaboration subsystem. Based on the status confirmation token, the voyage execution business object removes the dependency constraints of subsequent nodes in the directed acyclic graph task node, completing the state synchronization closed loop of cross-ship and shore business objects.

[0015] The beneficial effects of this invention are: 1. In distributed process scheduling for multi-service object collaboration, by decomposing cross-network segment business processes into directed acyclic graph task nodes with clear dependencies, and having local collaborative computing modules perform statistical and correlation operations on time-series data at the edge, only transmitting the generated lightweight intermediate indicator data packets, the cross-network segment transmission volume of original time-series data can be reduced, and the system's dependence on real-time full upload can be reduced. In network environments with limited bandwidth and large transmission latency, communication congestion and data backlog can be alleviated, and key instructions and status data can be transmitted in an orderly manner.

[0016] 2. The link adaptive scheduling module adjusts the transmission mode according to the link status and continuously judges the switching conditions using preset switching lower limit threshold, preset coordination threshold and preset smoothing window. This can avoid short-term fluctuations in channel parameters from repeatedly triggering mode switching, thereby reducing frequent oscillations between real-time transmission mode and data temporary storage mode, reducing bandwidth waste caused by repeated transmission, and improving the stability of distributed process scheduling.

[0017] 3. By constructing a hierarchical priority queue based on criticality factors, data packets containing status confirmation tokens and data packets with higher urgency can be transmitted first, avoiding them from being stuck behind a large amount of raw log data for a long time. After the link is restored, critical status and control instructions can be delivered earlier, thereby reducing head-of-line congestion, accelerating the removal of dependencies between task nodes in the directed acyclic graph, and shortening the time required for the global business state to reach consistency. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the composition structure of the ship-shore collaborative distributed system of the present invention; Figure 2 This is a diagram illustrating the collaborative processing of cross-network segment service objects in this invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] A distributed process scheduling system for multi-business object collaboration, comprising: Both the ship-side edge collaboration subsystem and the shore-side global collaboration subsystem are equipped with local state storage modules and distributed process scheduling engines. The distributed process scheduling engine decomposes the business process into a set of directed acyclic graph task nodes containing dependencies, and the distributed process scheduling engine includes: The incremental data extraction module is used to perform incremental queries on the local state storage module based on business data query conditions to extract incremental time-series data. The local collaborative computing module is used to complete the statistical analysis of voyage indicators and the correlation calculation of operating conditions at the edge of the ship and generate lightweight intermediate indicator data packages. The link adaptive scheduling module calculates the scheduling strategy factor based on the transmission delay and available bandwidth of the communication channel module. When the scheduling strategy factor is lower than the preset switching threshold and the duration reaches the preset smoothing window, real-time transmission is suspended and switched to data storage mode. In data storage mode, a hierarchical priority queue is constructed based on the dependency criticality factor of the task nodes in the directed acyclic graph, and data packets containing status confirmation tokens are placed in the high-priority queue. When the scheduling strategy factor reaches the preset coordination threshold and the duration reaches the preset smoothing window, real-time transmission mode is restored and data packets in the high-priority queue are scheduled first. The state consistency synchronization module is used to receive lightweight intermediate indicator data packets, trigger shore-side business object linkage to update the global business state, and send the state confirmation token back to the ship-side edge collaboration subsystem.

[0022] Preferably, when the distributed process scheduling engine decomposes the set of task nodes in the directed acyclic graph, it extracts the dependency criticality factor of each task node in the topology. The dependency criticality factor is determined recursively based on the depth of subsequent dependency links and the urgency of synchronization of the task node in the topology. The distributed process scheduling engine generates a constraint vector for each task node that includes task identifier, set of predecessor dependencies, processing object attributes and data input relationships.

[0023] Preferably, the preset switching lower limit threshold is lower than the preset coordination threshold, and the preset switching lower limit threshold is set to 0.85; when the scheduling strategy factor decreases from greater than or equal to 1.0 to less than 1.0 but greater than or equal to 0.85, the link adaptive scheduling module maintains the cross-network segment real-time transmission mode; when the scheduling strategy factor is less than 0.85 and this state is not interrupted within the preset smoothing window, the link adaptive scheduling module suspends real-time transmission and switches to data temporary storage mode; wherein, the preset smoothing window is set to 15s.

[0024] Preferably, the ship-side edge collaboration subsystem includes operational condition data acquisition business objects, fuel consumption analysis business objects, and voyage execution business objects deployed on the ship; the shore-side global collaboration subsystem includes enterprise management decision-making business objects, dispatching and command business objects, and logistics docking business objects deployed on the shore-based server; the operational condition data acquisition business objects connect to the ship-side sensor interface module and collect main engine speed, speed, draft, and fuel flow data at a sampling period of 5 seconds; the fuel consumption analysis business objects receive the data collected by the operational condition data acquisition business objects and calculate the deviation between the average nautical mile fuel consumption index and engine efficiency.

[0025] Preferably, when the link adaptive scheduling module is in data temporary storage mode, it assigns data packets with a dependency criticality factor greater than or equal to 0.90 to a high-priority queue and data packets with a dependency criticality factor less than 0.90 to a low-priority queue. The high-priority queue stores status confirmation tokens and lightweight intermediate indicator data packets containing operating condition deviation indicators, while the low-priority queue stores high-frequency raw log incremental data packets. After resuming real-time transmission mode, the link adaptive scheduling module temporarily suspends the transmission of data packets in the low-priority queue before the high-priority queue is cleared.

[0026] Preferably, the data incremental extraction module uses a sliding window mechanism based on version vectors to extract time-series data increments; the data incremental extraction module monitors the tariff policy and real-time channel quality of the communication channel module, and reduces the sampling time span of the sliding window when the channel quality deteriorates, and expands the sampling time span of the sliding window when the channel quality improves.

[0027] Preferably, the lightweight intermediate indicator data package includes fuel consumption statistics for the voyage phase, operating condition deviation indicators, and hash check codes generated from local state data; the data volume of the lightweight intermediate indicator data package is less than 5% of the original time-series data acquisition volume; after receiving the lightweight intermediate indicator data package, the state consistency synchronization module compares the hash check codes, and updates the shore-side global state storage module after the comparison is successful.

[0028] Preferably, the local state storage module includes a circular temporary queue; when the available bandwidth of the communication channel module drops to 0kbps and the scheduling strategy factor is lower than the preset switching lower limit threshold, the link adaptive scheduling module writes the generated lightweight intermediate index data packet into the circular temporary queue; after the communication channel module recovers and the scheduling strategy factor reaches the preset coordination threshold again, the link adaptive scheduling module sequentially resends the data packets in the circular temporary queue according to the order of the temporary timestamps.

[0029] Preferably, after updating the global business status, the shore-side global collaboration subsystem calls the enterprise management decision-making business object to generate a speed adjustment instruction, and calls the scheduling and command business object to encapsulate the speed adjustment instruction into a status confirmation token and send it back to the voyage execution business object of the ship-side edge collaboration subsystem. Based on the status confirmation token, the voyage execution business object removes the dependency constraints of subsequent nodes in the directed acyclic graph task node, completing the state synchronization closed loop of cross-ship and shore business objects.

[0030] Example 1: When a container ship is sailing in the open ocean, the ship-side edge collaboration subsystem and the shore-side global collaboration subsystem establish a cross-network segment data path through a communication channel module composed of ship-shore satellite links. Both subsystems are equipped with local state storage modules and distributed process scheduling engines. The shore-side local state storage module is used to save the global business state and serves as the shore-side global state storage module. Affected by sea waves and weather obstruction, the transmission delay and available bandwidth of the satellite communication channel fluctuate frequently at the millisecond level near the critical threshold. The operational condition acquisition business object collects high-frequency raw time-series data such as engine speed, speed, draft, and fuel flow at a sampling period of 5 seconds. If this data is uploaded to the shore in real time, the limited satellite link bandwidth will be blocked and data backlog will occur. When using static channel threshold switching, the system will frequently switch between real-time transmission mode and data temporary storage mode with channel parameter jitter, generating redundant data packet retransmissions and causing high-priority status confirmation tokens to be stuck behind a large number of time-series logs, resulting in a disconnect between the collaborative state of the voyage execution business object and the enterprise management decision-making business object and a delay in topology unlocking.

[0031] The distributed process scheduling engine configured in the ship's edge collaboration subsystem receives fuel consumption assessment and speed adjustment instructions for each voyage phase. It decomposes the cross-business process into a set of directed acyclic graph task nodes containing dependencies, extracts the depth of subsequent dependency links and synchronization urgency of each task node in the topology, and recursively calculates the dependency criticality factor from subsequent nodes. The distributed process scheduling engine also generates constraint vectors for each task node, including task identifiers, sets of prerequisite dependencies, processing object attributes, and data input relationships. Among these, the constraint vectors for the task nodes corresponding to the data incremental extraction module... The value is 0.3, corresponding to the voyage index statistics, operational condition correlation calculation, and data packet generation task nodes of the local collaborative calculation module. The value is 0.6, corresponding to the task node for the shore-side speed adjustment command feedback and status confirmation token confirmation. The version vector, with a value of 0.95, contains two integer fields: the ship's local data update sequence number and the shore-acknowledged receipt sequence number. During network disconnection, the ship increments its local data update sequence number by 1 each time it completes local data extraction, while the shore-acknowledged receipt sequence number remains unchanged, forming a unidirectionally increasing version number. After network recovery, the ship sends the current version vector to the shore along with the data packet. The shore compares the received version vector with its own stored state vector. If the ship's local data update sequence number is greater than the shore-acknowledged receipt sequence number, it triggers incremental retransmission and state overwrite arbitration starting from the breakpoint to complete state synchronization under asymmetric disconnection conditions. The incremental data extraction module performs a sliding window incremental query based on the version vector on the ship's local state storage module according to the business data query conditions, extracting time-series data increments containing 2400KB of data. The local collaborative calculation module calls the fuel consumption analysis business object to complete the voyage index statistics and operating condition correlation calculations at the ship's edge, generating a 12KB file containing the average nautical mile fuel consumption index. Operating condition deviation index The lightweight intermediate index data packet with hash checksum is used, and the data extraction and compression ratio is 99.5%, reducing the cross-network segment transmission volume of the original time-series collected data, which is beneficial for calculating the average nautical mile fuel consumption index. First, the sampled fuel flow rate is accumulated second by second to obtain the total fuel consumption within the set time period. Then, the total distance is calculated by multiplying the speed and time during the same period. Finally, the total fuel consumption is divided by the total distance to calculate the deviation index of the operating conditions. At that time, the engine speed and draft are read, and the reference fuel consumption rate corresponding to the combination of engine speed and draft is retrieved from the pre-stored standard performance surface lookup table of the engine. Then, the difference between the actual fuel consumption rate and the reference fuel consumption rate is calculated. The difference is divided by the reference fuel consumption rate and multiplied by 100% to obtain the percentage value representing the deviation of engine efficiency, so as to complete the conversion of multi-dimensional physical acquisition data into operating condition evaluation indicators.

[0032] The link adaptive scheduling module monitors the transmission delay and available bandwidth of the communication channel module in real time and presets a minimum guaranteed bandwidth. 64kbps, maximum tolerable latency preset. The timeout is 2000ms, and the scheduling strategy factor is calculated according to the following formula: ,in, Indicates the scheduling strategy factor. Indicates the currently available bandwidth. This indicates the preset minimum guaranteed bandwidth. Indicates the current link transmission delay. This represents the preset maximum tolerable delay, calculated when the ship enters a dynamic fading channel, where the available bandwidth drops to 80kbps and the transmission delay rises to 1500ms. The value is 0.59046, which is lower than the preset collaboration threshold of 1.0 and the preset switching lower limit threshold of 0.85. When the value drops from greater than or equal to 1.0 to below 1.0 but still greater than or equal to 0.85, the link adaptive scheduling module maintains real-time transmission mode; when Continuous timing begins when the value drops below 0.85, and only during a preset smoothing window of 15 seconds in this state. If no interruption occurs, real-time transmission is suspended and switched to data buffering mode. To handle the frequent millisecond-level jitter of channel parameters, the system continuously acquires raw bandwidth and transmission delay data with a sampling period of 100ms before link evaluation, and performs weighted averaging within a 5s time sliding window. The link adaptive scheduling module then applies the filtered data... Perform continuous judgment; during the 15-second timing process, as long as If the value rises to 0.85 or above, the timing is interrupted and cleared. Timing restarts when the value falls below 0.85 again. Under the channel conditions of this embodiment, After the value remains below 0.85 for 15 seconds, the link adaptive scheduling module switches to data buffering mode. In data buffering mode, the link adaptive scheduling module schedules data packets according to the directed acyclic graph task nodes corresponding to them. Construct a hierarchical priority queue, Packets with a value greater than or equal to 0.90 are placed in a high-priority queue. Data packets with a priority level below 0.90 are placed in a low-priority queue. In this embodiment, high-priority data packets are... The value is 0.95. The high-priority queue stores status confirmation tokens and includes operating condition deviation indicators. Lightweight intermediate metric data packets; stored in a low-priority queue. The high-frequency raw log incremental data packet is 0.3. After the satellite channel is restored, when When the status reaches 1.0 again and remains there for 15 seconds, the link adaptive scheduling module resumes real-time transmission mode, prioritizes sending data packets in the high-priority queue, and temporarily suspends sending data packets in the low-priority queue before the high-priority queue is cleared. This allows the status confirmation token to be transmitted across network segments within 1.5 seconds, avoiding head-of-queue congestion caused by a large amount of log data.

[0033] After receiving the lightweight intermediate indicator data packet, the distributed process scheduling engine configured in the shore-side global collaboration subsystem compares the hash check code in the data packet with the data packet. If the comparison is successful, the state consistency synchronization module updates the shore-side global state storage module, which is the local state storage module configured in the shore-side global collaboration subsystem. Subsequently, the shore-side business object linkage is triggered. The enterprise management decision business object generates a speed adjustment instruction based on the updated global business state. The scheduling command business object encapsulates the speed adjustment instruction into a state confirmation token and sends it back to the voyage execution business object in the ship-side edge collaboration subsystem. The voyage execution business object releases the dependency constraints of subsequent nodes in the directed acyclic graph task node based on the state confirmation token, completing the state synchronization closed loop across ship and shore business objects.

[0034] Example 2: The ship-side edge server, shore-side cloud-based server, and satellite link simulator constitute a distributed hardware testing platform. The ship-side edge server is equipped with an 8-core processor and 16GB of memory, deploys the ship-side edge collaboration subsystem, and connects to the operational condition acquisition business object, fuel consumption analysis business object, and voyage execution business object. The shore-side cloud-based server is equipped with a 16-core processor and 32GB of memory, deploys the shore-side global collaboration subsystem, which includes enterprise management decision-making business objects, scheduling and command business objects, and logistics docking business objects. The satellite link simulator connects the ship-side and shore-side network ports, and dynamically loads the bandwidth limit, transmission delay, and bit error rate of the satellite communication channel through a wide area network impairment model to test the data transmission, mode switching, and global business status convergence under different scheduling methods.

[0035] The test data consisted of time-series logs of main engine speed, speed, draft, and fuel flow collected from ocean-going container ships during specific voyages. To reproduce channel jitter in industrial electromagnetic environments and high sea states, Gaussian white noise with a signal-to-noise ratio of 20dB and 50Hz power frequency interference were superimposed on the input signal source. The sampling period for the operational data acquisition service was set to 5s, the preset minimum guaranteed bandwidth was set to 64kbps, the preset maximum tolerable delay was set to 2000ms, the preset lower threshold for switching was set to 0.85, and the preset smoothing window was set to 15s. Three systems were set up for testing: control group A adopted a static threshold single-unit switching architecture, control group B adopted a dynamic scheduling architecture without hysteresis smoothing gating and hierarchical priority queues, and the experimental group adopted a distributed process scheduling system with hysteresis smoothing gating and a hierarchical priority queue based on criticality.

[0036] Within 0 to 120 seconds, the satellite link simulator maintains the available bandwidth at 256 kbps and the transmission delay at 800 ms. Substituting these two inputs into the link evaluation function yields the scheduling policy factor. The value is 2.68128. The communication channel module is in real-time transmission mode. During the 120-second channel jitter period (121-240 seconds), the satellite link simulator injects transient fading and wave swaying disturbances, causing the available bandwidth to fluctuate between 50kbps and 120kbps, and the transmission delay to fluctuate between 1200ms and 1900ms. The calculated value is... It varies between 0.42152 and 0.92318, and fluctuates repeatedly around the preset switching lower limit threshold of 0.85. Within 241 to 360 seconds, the simulated channel recovers stability and the available bandwidth rises back to 256kbps.

[0037] During the aforementioned testing process, the incremental data extraction module performed incremental extraction on the 2400KB raw acquisition logs generated within a 1-hour window, while the local collaborative calculation module called the fuel consumption analysis business object to complete the voyage index statistics and operating condition correlation calculations, generating an index including average nautical mile fuel consumption. Operating condition deviation index The lightweight intermediate index data packet with hash check code is 12KB in size. Compared with the original 2400KB collection log, the link data compression rate is 99.5%. Control group A adopts a centralized full upload method and does not perform the above incremental extraction and local collaborative calculation on the ship end. Its link data compression rate is 0.0%. Control group B and the experimental group both perform incremental extraction and local collaborative calculation, and the link data compression rate is 99.5% for both.

[0038] The incremental data extraction module adjusts the sampling time span of the sliding window based on the real-time channel quality. When the channel quality is within the normal range, the sampling time span remains at 60 seconds; when... When the value drops to between 0.85 and 1.0, the sampling time span is proportionally reduced to 30 seconds; when When the value is further reduced to below 0.85, the sampling time span is reduced to 10 seconds to decrease the amount of data extracted per time. This is especially important when channel quality improves. When the value is raised back to 1.0 or higher, the sampling time span is increased by 10 seconds every 15 seconds until it is restored to 60 seconds.

[0039] During the channel jitter period of 121 to 240 seconds, control group B lacked a hysteresis smoothing gating system comprised of a preset handover lower limit threshold and a preset smoothing window. Transient changes in channel parameters triggered mode switching multiple times, accumulating 14 handovers within 120 seconds, accompanied by handshake packet retransmissions, communication channel congestion, and data transmission delays. The experimental group... When the value dropped to 0.78125 in 125 seconds, the link adaptive scheduling module started to continuously time the state below 0.85. After this state lasted for 15 seconds, the system switched from real-time transmission mode to data storage mode in 140 seconds. Only one mode switch occurred during the entire channel jitter period. Control group A did not complete an effective scheduling mode switch after the centralized full upload was blocked, and the number of recorded mode switches was 0.

[0040] In the data temporary storage mode, the experimental group's link adaptive scheduling module uses the dependency criticality factor of the task nodes in the directed acyclic graph. Construct a hierarchical priority queue, relying on status confirmation tokens with a criticality factor of 0.95 and including operating condition deviation indicators. Lightweight intermediate indicator data packets are placed in the high-priority queue, while high-frequency raw log incremental data packets with a dependency criticality factor of 0.3 are placed in the low-priority queue. The satellite communication channel recovers after 241 seconds. After reaching 1.0 again and maintaining it for 15 seconds, the experimental group resumed real-time transmission mode at 256 seconds, first sending data packets in the high-priority queue, and temporarily suspending the sending of data packets in the low-priority queue before the high-priority queue was cleared. The shore-side global coordination subsystem received the priority-sent status confirmation token at 257.5 seconds, with a network segment penetration delay of 1.5 seconds. Control group B used a first-in-first-out queue, and the status confirmation token remained after the backlog of raw logs, with a network segment penetration delay of 42.8 seconds. Control group A was affected by centralized full upload blocking, and its network segment penetration delay reached 108.5 seconds.

[0041] After receiving the lightweight intermediate index data packet, the shore-side global coordination subsystem compares the hash check code in the packet with the global service status. After the comparison is successful, the global service status is updated. The enterprise management decision-making business object generates a speed adjustment command based on the updated global service status. The scheduling and command business object encapsulates the speed adjustment command into a status confirmation token and sends it back to the voyage execution business object. The voyage execution business object then removes the dependency constraints of subsequent nodes in the directed acyclic graph task node. After repeating the above channel change and state synchronization process, the global coordination state convergence success rate of control group A is 45.2%, that of control group B is 68.5%, and that of the experimental group is 99.8%.

[0042] Example 3: During the transoceanic voyage of a 10,000-ton ocean-going container ship, the ship-side edge collaboration subsystem continuously collects time-series data on main engine speed, ship speed, draft, and fuel flow. The shore-side global collaboration subsystem receives status information and issues speed adjustment commands through the communication channel module. Affected by weather obstruction and sea swaying, the satellite communication link frequently experiences intermittent interruptions, reducing the available bandwidth to 0kbps and increasing the transmission delay. At this time, the local state storage module of the ship-side edge collaboration subsystem stores the unsent incremental time-series data and lightweight intermediate index data packets through a circular temporary queue. If the data packets to be transmitted are still sent in the order they entered the queue, the status confirmation token is likely to be stuck after the high-frequency raw log incremental data packets, causing the dependency topology unlocking of the directed acyclic graph task nodes to lag, and causing the state of cross-ship shore business objects to lose synchronization.

[0043] After receiving a business process trigger command, the distributed process scheduling engine configured in the ship-side edge collaboration subsystem decomposes the complex business process across business objects into a set of directed acyclic graph task nodes containing dependencies. It then extracts the business data query conditions and calculation logic corresponding to each task node and calculates the dependency criticality factor of each task node. This factor is recursively calculated based on the depth of subsequent dependency links and synchronization urgency of task nodes in the topology. When decomposing task nodes, the distributed process scheduling engine also generates a constraint vector for each task node, which includes task identifier, set of prerequisite dependencies, processing object attributes and data input relationships.

[0044] In calculation During this process, the distributed process scheduling engine traverses the directed acyclic graph topology to determine all reachable paths from the target task node to the endpoint node. The total number of nodes corresponding to the path with the most nodes is used as the subsequent dependency link depth. Then, starting from the endpoint node, the process recursively reverses, setting the initial value of the endpoint node to 1.0. For any non-endpoint node, the dependency values ​​of each direct successor node are multiplied by their corresponding edge connection weights, summed, and then divided by the total number of direct successor nodes to obtain the initial link value. This value is then multiplied by the subsequent dependency link depth to obtain the topology base score. Task nodes associated with critical status confirmation or secure command issuance have a synchronization urgency value of 0.5; task nodes associated with ordinary data transmission have a synchronization urgency value of 0.1. The topology base score is added to the synchronization urgency value, and then normalized to a value range of 0 to 1.0 to obtain the corresponding task node's... .

[0045] The link adaptive scheduling module calculates the scheduling strategy factor based on the transmission delay and available bandwidth of the communication channel module. ,when When the value drops from above 1.0 to below 1.0, but remains greater than or equal to the preset switching lower limit threshold of 0.85, the system maintains real-time transmission mode; when When the value is below 0.85, the link adaptive scheduling module starts timing continuously according to a preset smoothing window of 15 seconds, only when... The system will only suspend real-time transmission and switch to data storage mode if the value remains below 0.85 for 15 seconds without interruption.

[0046] exist During the transition phase where the signal strength is below 0.85 but the duration has not yet reached 15 seconds, the system activates a pre-protective double buffer mechanism. At this time, the real-time transmission mode continues to operate, the link adaptive scheduling module reduces the transmission frequency of high-frequency raw log incremental data packets, writes non-critical time-series log data into a temporary secondary buffer, and only allows high-priority data packets to continue attempting to be transmitted through the main channel. Furthermore, the method for confirming successful transmission is changed from single confirmation to multi-path redundant retransmission. If the value recovers to 0.85 or higher within 15 seconds, the system interrupts the current timing and releases the log data in the temporary secondary buffer; if... If the value remains below 0.85 for 15 seconds, the data in the temporary secondary buffer will be sequentially merged into the circular temporary storage queue, and the system will switch to data temporary storage mode.

[0047] In data temporary storage mode, the link adaptive scheduling module determines the task node corresponding to the data packet to be transmitted. Construct a hierarchical priority queue. Place status confirmation tokens and corresponding highly dependent critical task nodes, along with lightweight intermediate indicator data packets containing operational deviation metrics, into the high-priority queue. Place high-frequency raw log incremental data packets into the low-priority queue. After the communication channel module recovers, when... When the system reaches 1.0 again and remains so for 15 seconds, it resumes real-time transmission mode, prioritizes sending data packets in the high-priority queue, and performs breakpoint resumption on data packets in the circular temporary queue according to the order of the temporary timestamps; before the high-priority queue is cleared, data packets in the low-priority queue are temporarily suspended from transmission.

[0048] After the shore-side global collaboration subsystem receives the lightweight intermediate indicator data packet, the state consistency synchronization module extracts the hash check code and compares it. After the comparison is successful, the state consistency synchronization module updates the shore-side global state storage module, triggering the enterprise management decision-making business object and the scheduling and command business object to jointly update the global business state, and sends the state confirmation token back to the ship-side edge collaboration subsystem, completing the state synchronization closed loop of cross-ship and shore business objects.

[0049] Example 4: This example combines Figures 1 to 2 This section describes a distributed process scheduling system that enables collaboration among multiple business objects, such as... Figure 1 As shown, the ship-side edge collaboration subsystem includes a working condition acquisition business object, a fuel consumption analysis business object, a local state storage module, a data increment extraction module, a local collaboration calculation module, a link adaptive scheduling module, and a voyage execution business object. The working condition acquisition business object collects main engine speed, speed, draft, and fuel flow data at a sampling period of 5 seconds and transmits the collected data to the fuel consumption analysis business object. The fuel consumption analysis business object receives the data collected by the working condition acquisition business object, calculates the deviation between the average nautical mile fuel consumption index and engine efficiency, and stores the calculated data in the local state storage module, which includes a circular temporary storage queue. The data increment extraction module performs incremental queries on the local state storage module according to the business data query conditions, extracts the time-series data increments, and inputs them into the local collaboration calculation module. The local collaboration calculation module completes the voyage index statistics and working condition correlation calculations at the ship-side edge, generating lightweight intermediate index data packets. The link adaptive scheduling module calculates the scheduling strategy factor based on the transmission delay and available bandwidth of the communication channel module, and constructs a hierarchical priority queue when in data temporary storage mode, sending the lightweight intermediate index data packets to the local state storage module. The data is sent to the communication channel module, which receives lightweight intermediate indicator data packets and transmits them to the shore-side global collaboration subsystem. The shore-side global collaboration subsystem includes a state consistency synchronization module, a shore-side local state storage module that stores the global business state, enterprise management decision-making business objects, scheduling and command business objects, and logistics docking business objects. After receiving the lightweight intermediate indicator data packets sent by the communication channel module, the state consistency synchronization module triggers a linkage to update the global business state stored in the shore-side local state storage module and sends back a state confirmation token. The enterprise management decision-making business object generates a speed adjustment instruction based on the updated global business state. The logistics docking business object provides feedback on the port loading and unloading urgency time constraint index to adjust the synchronization urgency. The scheduling and command business object encapsulates the speed adjustment instruction into a state confirmation token based on the synchronization urgency and sends the state confirmation token and speed adjustment instruction back to the ship-side edge collaboration subsystem through the communication channel module. The voyage execution business object releases the dependency constraints of subsequent nodes in the directed acyclic graph task node based on the state confirmation token sent back by the communication channel module, completing the state synchronization closed loop.

[0050] like Figure 2As shown, the communication channel module connects the ship-side edge collaboration subsystem and the shore-side global collaboration subsystem. The operational condition acquisition service collects data on main engine speed, ship speed, draft, and fuel flow rate, and transmits the collected data to the data increment extraction module. The data increment extraction module extracts the time-series data increments and outputs them to the local collaborative calculation module. The local collaborative calculation module generates lightweight intermediate indicator data packets based on the time-series data increments and transmits them to the link adaptive scheduling module. The link adaptive scheduling module calculates the scheduling strategy factor based on transmission delay and available bandwidth, and outputs the lightweight intermediate indicator data packets to the communication channel module. The communication channel module then processes the lightweight intermediate indicator data packets... The system transmits data across network segments and delivers it to the state consistency synchronization module. After receiving the lightweight intermediate indicator data packet, the state consistency synchronization module updates the global business state and triggers the enterprise management decision-making business object to generate a speed adjustment command. The enterprise management decision-making business object sends the generated speed adjustment command to the dispatch and command business object. The dispatch and command business object encapsulates the speed adjustment command into a state confirmation token and inputs it to the communication channel module. The communication channel module sends the state confirmation token back to the voyage execution business object. The voyage execution business object releases the dependency constraint based on the state confirmation token, completing the state synchronization closed loop across ship-shore business objects and realizing the state synchronization closed loop with the working condition acquisition business object.

[0051] Example 5: When a 10,000-ton ocean-going container ship first connects to the distributed process scheduling system for multi-business object collaboration and prepares for transoceanic voyage, the ship-side edge collaboration subsystem performs on-site deployment calibration to determine the preset maximum tolerable delay used by the link adaptive scheduling module. With preset minimum guaranteed bandwidth .

[0052] When the ship is berthed and the satellite communication channel is in an unobstructed open environment, the local cooperative computing module continuously sends probe data packets containing fixed-byte payloads to the shore-side global cooperative subsystem. The state consistency synchronization module records the round-trip time of the shore-side return of the state confirmation token, forming round-trip time data for 1000 consecutive sampling periods. During calibration, the extreme values ​​at both ends of this data set are first removed, and the mean of the remaining data is calculated. This mean is then multiplied by a coefficient of 1.2, which includes a 20% environmental degradation margin, to obtain a preliminary estimate of the preset maximum tolerable delay. Subsequently, the 99th percentile of the round-trip time in the 1000 sampling periods after removing the extreme values ​​is taken as the baseline value for the upper bound of communication jitter. This baseline value is then multiplied by the coefficient 1.2 to obtain the final value. Therefore, the initial estimate is used to verify the calibration results, and the final value covers the upper limit of the delay when the business control link is in an acceptable state.

[0053] In calibration During this period, the operational data acquisition module collects operating data under simulated full-load operation of the ship's main engine, and measures the minimum data throughput rate required for the fuel consumption analysis module and the voyage execution module to maintain basic state synchronization. The incremental data extraction module multiplies this minimum data throughput rate by a safety concurrency redundancy coefficient, and calibrates the result as follows: This ensures that the calculation benchmark of the scheduling strategy factor matches the communication load of the ship. After parameter calibration, the data incremental extraction module continuously monitors the tariff strategy and real-time channel quality of the communication channel module. When the channel quality deteriorates, the sampling time span of the sliding window is reduced, and when the channel quality improves, the sampling time span is expanded.

[0054] Before the shore-based global coordination subsystem issues a constraint vector containing speed adjustment commands, the distributed process scheduling engine considers the dependency criticality factors of each task node. Offline calibration is performed. The shore-based distributed process scheduling engine imports cross-business object collaboration logs from historical voyages, extracts incremental operational data to identify the frequency of historical state disconnection between nodes and enterprise management decision-making business objects, and schedules and commands business objects to traverse the directed acyclic graph topology to calculate the longest physical interaction path hop count from the task node to the process termination node, and uses it as the depth of subsequent dependency links.

[0055] The enterprise management decision-making business object normalizes the frequency of historical state disconnection and the number of hops in the longest physical interaction path. The normalized frequency of historical state disconnection is multiplied by 0.6, and the normalized number of hops in the longest physical interaction path is multiplied by 0.4. The two results are then added together to obtain the initial topology weight. The state consistency synchronization module reads the port loading and unloading urgency time constraint index fed back by the logistics docking business object and adjusts the synchronization urgency accordingly: when there are less than 12 hours left before the scheduled loading and unloading time, the compensation coefficient is set to 0.2; when there are 12 hours to 24 hours left before the scheduled loading and unloading time, the compensation coefficient is set to 0.1; when there are more than 24 hours left before the scheduled loading and unloading time, the compensation coefficient is set to 0. The initial topology weight is added to the corresponding compensation coefficient. If the value exceeds 1.0, it is truncated to 1.0 to form the synchronization urgency used for recursive calculation of task nodes.

[0056] The distributed process scheduling engine recursively proceeds from the endpoint of the directed acyclic graph, calculating the final result based on the depth of subsequent dependency links and synchronization urgency of each task node. and will each core task node The static mapping lookup table is written to local memory. After channel fading and switching to data buffering mode, the ship's edge cooperative subsystem directly reads the data from the static mapping lookup table. Construct a hierarchical priority queue to determine the priority of data packets to be transmitted and arbitrate the transmission tokens accordingly.

[0057] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A distributed process scheduling system for multi-business object collaboration, characterized in that, Include: Both the ship-side edge collaboration subsystem and the shore-side global collaboration subsystem are equipped with local state storage modules and distributed process scheduling engines. The distributed process scheduling engine decomposes the business process into a set of directed acyclic graph task nodes containing dependencies, and the distributed process scheduling engine includes: The incremental data extraction module is used to perform incremental queries on the local state storage module based on business data query conditions to extract incremental time-series data. The local collaborative computing module is used to complete the statistical analysis of voyage indicators and the correlation calculation of operating conditions at the edge of the ship and generate lightweight intermediate indicator data packages. The link adaptive scheduling module is used to calculate the scheduling strategy factor based on the transmission delay and available bandwidth of the communication channel module. When the scheduling strategy factor is lower than the preset switching threshold and the duration reaches the preset smoothing window, real-time transmission is suspended and switched to data storage mode. In data storage mode, a hierarchical priority queue is constructed based on the dependency criticality factor of the task nodes in the directed acyclic graph, and data packets containing status confirmation tokens are placed in the high-priority queue. When the scheduling strategy factor reaches the preset coordination threshold and the duration reaches the preset smoothing window, real-time transmission mode is restored and data packets in the high-priority queue are scheduled first. The state consistency synchronization module is used to receive lightweight intermediate indicator data packets, trigger shore-side business object linkage to update the global business state, and send the state confirmation token back to the ship-side edge collaboration subsystem.

2. The distributed process scheduling system for multi-business object collaboration according to claim 1, characterized in that, When the distributed process scheduling engine decomposes the set of task nodes in a directed acyclic graph, it extracts the dependency criticality factor of each task node in the topology. The dependency criticality factor is determined recursively based on the depth of subsequent dependency links and the urgency of synchronization of task nodes in the topology. The distributed process scheduling engine generates a constraint vector for each task node, which includes task identifier, set of preceding dependencies, processing object attributes, and data input relationships.

3. The distributed process scheduling system for multi-business object collaboration according to claim 1, characterized in that, The preset handover lower limit threshold is lower than the preset coordination threshold, and the preset handover lower limit threshold is set to 0.85; when the scheduling strategy factor decreases from greater than or equal to 1.0 to less than 1.0 but greater than or equal to 0.85, the link adaptive scheduling module maintains the cross-network segment real-time transmission mode; when the scheduling strategy factor is less than 0.85 and the status is not interrupted within the preset smoothing window, the link adaptive scheduling module suspends real-time transmission and switches to data temporary storage mode; where the preset smoothing window is set to 15s.

4. A distributed process scheduling system for multi-business object collaboration according to claim 1, characterized in that, The ship-side edge collaboration subsystem includes operational condition data collection, fuel consumption analysis, and voyage execution business objects deployed on the ship; the shore-side global collaboration subsystem includes enterprise management decision-making, dispatching and command, and logistics docking business objects deployed on shore-based servers. The operating condition data acquisition service object connects to the ship's sensor interface module and collects main engine speed, speed, draft, and fuel flow data at a sampling period of 5 seconds; the fuel consumption analysis service object receives the data collected by the operating condition data acquisition service object and calculates the deviation between the average nautical mile fuel consumption index and engine efficiency.

5. A distributed process scheduling system for multi-business object collaboration according to claim 1, characterized in that, When the link adaptive scheduling module is in data temporary storage mode, it assigns data packets with a dependency criticality factor greater than or equal to 0.90 to a high-priority queue and data packets with a dependency criticality factor less than 0.90 to a low-priority queue. The high-priority queue stores status confirmation tokens and lightweight intermediate indicator data packets containing operating condition deviation indicators, while the low-priority queue stores high-frequency raw log incremental data packets. After resuming real-time transmission mode, the link adaptive scheduling module temporarily suspends the transmission of data packets in the low-priority queue before the high-priority queue is cleared.

6. A distributed process scheduling system for multi-business object collaboration according to claim 1, characterized in that, The incremental data extraction module uses a sliding window mechanism based on version vectors to extract incremental time-series data. The incremental data extraction module monitors the tariff policy and real-time channel quality of the communication channel module, and reduces the sampling time span of the sliding window when the channel quality deteriorates, and expands the sampling time span of the sliding window when the channel quality improves.

7. A distributed process scheduling system for multi-business object collaboration according to claim 1, characterized in that, The lightweight intermediate index data package includes fuel consumption statistics for each voyage phase, operating condition deviation indicators, and hash check codes generated from local state data. The data volume of the lightweight intermediate index data package is less than 5% of the original time-series data volume. After receiving the lightweight intermediate index data package, the state consistency synchronization module compares the hash check codes, and updates the shore-side global state storage module after the comparison is successful.

8. A distributed process scheduling system for multi-business object collaboration according to claim 1, characterized in that, The local state storage module includes a circular temporary queue; when the available bandwidth of the communication channel module drops to 0kbps and the scheduling strategy factor is lower than the preset switching lower limit threshold, the link adaptive scheduling module writes the generated lightweight intermediate index data packet into the circular temporary queue. After the communication channel module recovers and the scheduling strategy factor reaches the preset coordination threshold again, the link adaptive scheduling module resends the data packets in the circular temporary queue in the order of the temporary timestamps.

9. A distributed process scheduling system for multi-business object collaboration according to claim 4, characterized in that, After updating the global business status, the shore-side global collaboration subsystem calls the enterprise management decision-making business object to generate a speed adjustment command, and calls the scheduling and command business object to encapsulate the speed adjustment command into a status confirmation token and send it back to the voyage execution business object of the ship-side edge collaboration subsystem. Based on the status confirmation token, the voyage execution business object removes the dependency constraints of subsequent nodes in the directed acyclic graph task node, completing the state synchronization closed loop of cross-ship and shore business objects.

Images