Data stream computing system and method

Abstract

The invention discloses a data stream computing system and method. The system comprises a plurality of routing nodes, a plurality of computing nodes and a data transmission link, the computing node is connected with at least one routing node through a data transmission link; the routing node comprises a plurality of flow counters and a routing decision unit; the traffic counter is used for acquiring the data traffic of the transmission port according to a preset time interval and transmitting the data traffic to the routing decision unit; the routing decision unit is used for performing flow state analysis on the transmission port based on the data flow; and if it is determined that the transmission port in the congestion state exists, selecting a transmission port in an idle state and a to-be-transferred data stream based on a preset routing adjustment strategy, and performing data transmission on the to-be-transferred data stream through the selected transmission port. According to the invention, a self-adaptive rerouting mechanism based on the real-time data flow state and the field information can be realized, and the data transmission efficiency and the link load balancing capability in a high-concurrency scene are improved.

Description

Technical Field

[0001] This invention relates to the field of integrated circuit technology, and in particular to a data flow computing system and method. Background Technology

[0002] In existing dataflow computing systems, the data transmission paths between computing nodes within the system are typically configured by a fixed strategy during system deployment or task initialization. While this statically configured routing method is applicable to relatively simple, small-scale applications and can meet basic data communication needs, as the complexity of computing tasks increases and the scale of data processing continues to expand, this type of static routing mechanism has gradually revealed many problems such as poor adaptability and slow response. It is no longer able to cope with the ever-increasing demands for system communication flexibility and operational efficiency from tasks such as large-scale parallel computing, high-speed data processing, and real-time response.

[0003] In existing dataflow computing systems, the load distribution of data flows exhibits highly dynamic fluctuations. The data traffic carried by the same routing path can change drastically at different times. Furthermore, at the same point in time, the data flow load between different paths often varies significantly. When a link or port becomes a system bottleneck due to persistent high load, a large amount of data will accumulate at that node, leading to localized congestion. This congestion not only causes a sharp increase in data transmission latency but can also severely impact the overall task scheduling order and the accuracy of the computation results.

[0004] However, traditional dataflow computing systems generally employ fixed routing strategies, lacking the ability to monitor link traffic changes in real time. They cannot monitor the instantaneous data transmission status of each port or link at any given point in time, and therefore cannot dynamically adjust routing paths based on load fluctuations during system operation. When sudden surges in traffic or uneven traffic distribution occur, the system often cannot respond promptly, ultimately leading to communication bottlenecks and performance degradation. Static routing mechanisms also limit the dynamic scheduling and utilization of idle links and computing node resources within the system, causing some channels to remain overloaded for extended periods while other channel resources are idle and wasted, severely impacting the overall utilization efficiency of computing resources and system throughput performance.

[0005] Therefore, there is an urgent need to propose a routing mechanism with traffic monitoring capabilities and dynamic adaptability, which can monitor the traffic status of each node and link in real time during operation and adaptively adjust the transmission path of the data stream based on the monitoring results.

[0006] This section is intended to provide background or context for the embodiments of the invention set forth in the claims. The description herein is not an admission that it is prior art simply because it is included in this section. Summary of the Invention

[0007] This invention provides a data stream computing system to implement an adaptive rerouting mechanism based on real-time data traffic status and field information, thereby improving data transmission efficiency and link load balancing capabilities in high-concurrency scenarios.

[0008] The data stream computing system includes: multiple routing nodes, multiple computing nodes, and data transmission links;

[0009] Adjacent routing nodes are connected via the data transmission link; the computing node is connected to at least one of the routing nodes via the data transmission link; the routing node is used to transmit data streams between different computing nodes via the data transmission link;

[0010] The routing node includes: multiple traffic counters and a routing decision unit; one end of the traffic counter is connected to the routing decision unit, and the other end is connected to the data transmission link through a transmission port, for acquiring the data traffic of the transmission port at preset time intervals and transmitting the data traffic to the routing decision unit;

[0011] The routing decision unit is used to perform traffic status analysis on the transmission port based on the data traffic; if it is determined that there is a transmission port in a congested state, it selects an idle transmission port and the data stream to be transferred based on a preset routing adjustment strategy, and transmits the data stream to be transferred through the selected transmission port.

[0012] In some embodiments, the routing decision unit includes:

[0013] The port selection module is used to determine that the transmission port is in a congested state if the data traffic exceeds a first preset threshold; and to query the traffic status of the transmission ports in the routing nodes according to a first preset order to determine that the transmission ports with data traffic less than a second preset threshold are in an idle state.

[0014] The data stream selection module is used to determine, based on the field information of the data packets in the data stream, at least one data packet in the data stream transmitted by the congested transmission port as the data stream to be transferred.

[0015] In some embodiments, the field information includes a priority field; the data stream selection module is specifically used to determine that the data packet corresponding to the priority field is the data packet of the data stream to be transferred if the priority field is a secondary priority.

[0016] In some embodiments, the field information further includes a timestamp field; the data stream selection module is specifically used to query the timestamp field of data packets in the data stream transmitted by the transmission port in a congested state, and sort the data packets according to the second preset order and the timestamp field to determine the data packets after the preset timestamp as the data packets of the data stream to be transferred.

[0017] In some embodiments, the field information further includes a dependency identifier field; the data stream selection module is specifically used to query the dependency identifier field of data packets in the data stream transmitted by the transmission port in a congested state, and determine, based on the dependency identifier field, data packets that have no dependency relationship with other data packets as data packets of the data stream to be transferred.

[0018] In some embodiments, the routing decision unit is further configured to query the field information of data packets in the data stream, and if the adjustment count field in the field information is equal to a third preset threshold, select the transmission port currently in a congested state for data transmission.

[0019] In some embodiments, the routing decision unit is further configured to perform the increment operation of the adjustment count field after determining that the data packet is a data packet of the data stream to be transferred.

[0020] This invention also provides a data flow calculation method to implement an adaptive rerouting mechanism based on real-time data flow status and field information, thereby improving data transmission efficiency and link load balancing capabilities in high-concurrency scenarios.

[0021] This data stream computing method, applied to the aforementioned data stream computing system, includes:

[0022] The routing node transmits data streams between different computing nodes through the data transmission link;

[0023] The traffic counter acquires the data traffic of the transmission port at preset time intervals and transmits the data traffic to the routing decision unit;

[0024] The routing decision unit performs traffic status analysis on the transmission port based on the data traffic; if it determines that there is a transmission port in a congested state, it selects an idle transmission port and the data stream to be transferred based on a preset routing adjustment strategy, and transmits the data stream to be transferred through the selected transmission port.

[0025] This invention also provides a chip that includes the data stream computing system described above.

[0026] This invention also provides a board card, including the chip described above.

[0027] This invention also provides an electronic device, including the aforementioned circuit board.

[0028] The data flow computing method provided in this invention, by introducing a traffic counter and a routing decision unit into the routing node, enables real-time monitoring and intelligent analysis of the traffic status of transmission ports, overcoming the transmission imbalance and link congestion problems caused by static routing path configuration in traditional data flow computing architectures. When a transmission port becomes congested, the computing system can automatically select an idle port based on a preset strategy and dynamically transfer suitable data flows, thereby effectively alleviating local congestion and improving the forwarding efficiency and path scheduling flexibility of data flows. This mechanism significantly enhances the computing system's adaptability to high-concurrency and large-scale task scenarios, improves the overall utilization of computing resources and system throughput performance, and exhibits good dynamic scalability and robustness. Attached Figure Description

[0029] 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. In the drawings:

[0030] Figure 1 This is a schematic diagram of the structure of a data stream computing system according to an embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of the structure of a routing node in an embodiment of the present invention;

[0032] Figure 3 This is a schematic diagram of the data stream computing system in another embodiment of the present invention;

[0033] Figure 4 This is a schematic diagram of the data stream computing system in another embodiment of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments and their descriptions are used to explain the present invention, but are not intended to limit the present invention. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be arbitrarily combined with each other. The acquisition, storage, use, and processing of data in the technical solutions of this application all comply with relevant laws and regulations. The user information in the embodiments of this application is obtained through legal and compliant means, and the acquisition, storage, use, and processing of user information have been authorized and agreed upon by the customer.

[0035] To facilitate understanding of the technical solution provided in this application, the relevant content of the technical solution in this application will be explained below.

[0036] To address the problems of uneven data transmission, link congestion, and low system resource utilization caused by static routing path configuration in existing dataflow computing systems, this application proposes a dataflow computing system to effectively improve the overall data transmission efficiency and resource scheduling capabilities of computing systems in scenarios involving high concurrency, large-scale, and dynamic computing tasks. In this dataflow computing system, multiple computing nodes are connected to one or more routing nodes via data transmission links. Data transmission between computing nodes is relayed and scheduled through these routing nodes. Each routing node is equipped with multiple traffic counters and a routing decision unit to monitor data transmission paths in real time and execute adaptive adjustment strategies. After receiving data traffic, the routing decision unit analyzes and judges the traffic status of each transmission port. If a transmission port is detected to be congested, it dynamically selects other currently idle transmission ports as new data output paths based on a preset routing adjustment strategy. Combining this with the field information of the data packets in the transmitted data stream, it determines a suitable data stream for path switching, thereby achieving data stream diversion from locally congested links and improving the overall dynamic balance scheduling capability of computing resources. Through the above mechanism, the data stream computing system can effectively reduce the risk of network congestion while ensuring the orderliness and correctness of data transmission. It can also improve the system's response to sudden traffic changes and throughput performance, and adapt to complex and ever-changing data stream operating environments.

[0037] like Figure 1 As shown, this application provides a data flow computing system, including: multiple routing nodes, multiple computing nodes, and data transmission links.

[0038] Adjacent routing nodes are connected via data transmission links. Computing nodes are connected to at least one routing node via data transmission links. Routing nodes are used to transmit data streams between different computing nodes via data transmission links.

[0039] The routing node includes multiple traffic counters and routing decision units.

[0040] One end of the traffic counter is connected to the routing decision unit, and the other end is connected to the data transmission link through the transmission port. It is used to obtain the data traffic of the transmission port at preset time intervals and transmit the data traffic to the routing decision unit.

[0041] The routing decision unit is used to analyze the traffic status of transmission ports based on data traffic. If it is determined that there are transmission ports in an overloaded state, it selects an idle transmission port and the data stream to be transferred based on a preset routing adjustment strategy, and transmits the data stream to be transferred through the selected transmission port.

[0042] According to the above embodiments, by introducing a traffic counter and a routing decision unit into the routing node, real-time monitoring and intelligent analysis of the traffic status of the transmission port can be achieved, overcoming the transmission imbalance and link congestion problems caused by the static configuration of routing paths in traditional data flow computing architectures. When a transmission port becomes congested, the computing system can automatically select an idle port based on a preset strategy and dynamically transfer suitable data flows, thereby effectively alleviating local congestion and improving the forwarding efficiency and path scheduling flexibility of data flows. The above mechanism significantly enhances the computing system's adaptability to high-concurrency and large-scale task scenarios, improves the overall utilization of computing resources and system throughput performance, and has good dynamic scalability and robustness.

[0043] In this embodiment of the invention, the number of computing nodes and routing nodes can be flexibly set by those skilled in the art according to specific application requirements, and the invention does not limit this number.

[0044] For ease of explanation, we will use a computing system architecture containing 9 routing nodes (i.e., routing nodes 1 to 9) and 16 computing nodes (i.e., computing nodes 1 to 16) as an example.

[0045] like Figure 1 As shown, the nine routing nodes are arranged in a 3×3 matrix structure, designated as routing node 1 to routing node 9. Each routing node connects to a data transmission link (…). Figure 1 (As shown by the black line with the double-headed arrows) It connects with adjacent routing nodes to form a highly interconnected and redundant routing network to support high-concurrency and low-latency data stream transmission.

[0046] Based on the aforementioned routing network, each routing node is further connected to multiple computing nodes via an independent data transmission link to achieve distributed deployment and localized scheduling of computing resources. In this embodiment, each routing node is connected to four computing nodes, totaling 16 computing nodes distributed around the routing node. Each computing node establishes a communication connection with its corresponding routing node via a dedicated data transmission link to enable bidirectional data transmission between computing nodes.

[0047] Computing nodes 1 through 16 are used to execute the defined computing tasks. After completing the corresponding computing task, each computing node sends the computing results as a data stream to the routing node it is connected to. The routing node schedules the received data stream according to the current routing policy and forwards it to the target computing node for subsequent computing processing.

[0048] like Figure 2 As shown, in this embodiment, the internal structure of the routing node includes multiple traffic counters and a routing decision unit. The routing node has multiple transmission ports; in the illustrated embodiment, it includes eight transmission ports, labeled as transmission port 1 to transmission port 8. The number and numbering of the ports are merely examples; the specific configuration can be adjusted according to actual application requirements, and this invention does not impose any limitations on this. Each transmission port is connected to a corresponding traffic counter, used to monitor the data traffic on its corresponding transmission port in real time, and to send the monitored data traffic to the routing decision unit at preset time intervals.

[0049] The routing decision unit, as the core control module in the routing node, is connected to all traffic counters. It receives the data traffic transmitted by each traffic counter and performs real-time analysis of the traffic status of each transmission port based on this data traffic. When the routing decision unit determines that the data traffic of a certain transmission port exceeds a first preset threshold, it determines that the transmission port is in a congested state. According to the preset routing adjustment strategy, among the remaining non-congested transmission ports, it selects an idle transmission port (i.e., the data traffic is below a second preset threshold) as an alternative path and determines the data flow that needs to be transferred, thereby realizing dynamic reconfiguration of data transmission paths and adaptive adjustment of system load.

[0050] According to the above embodiments, by constructing a highly interconnected network structure of routing nodes and distributed computing nodes, combined with a dynamic routing adjustment mechanism based on traffic monitoring, efficient scheduling and transmission of data streams in large-scale computing systems are achieved. By configuring multiple traffic counters and routing decision units within each routing node, the computing system can detect the traffic status of each transmission port in real time. When local congestion occurs, it dynamically selects idle ports and target data streams according to a preset routing adjustment strategy to perform path adjustment and load balancing. This effectively reduces transmission latency and the frequency of congestion, improves the overall utilization of computing resources and the timeliness of data stream processing, and enhances the adaptability and scalability of the computing system to dynamic computing tasks.

[0051] In some embodiments, the routing decision unit includes:

[0052] The port selection module is used to determine that a transmission port is congested if the data traffic exceeds a first preset threshold. It queries the traffic status of the transmission ports in the corresponding routing nodes according to a first preset order, and determines that the transmission ports with data traffic less than a second preset threshold are in an idle state.

[0053] The data stream selection module is used to determine, based on the field information of the data packets in the data stream, at least one data packet in the data stream transmitted by the congested transmission port as the data stream to be transferred. The data stream to be transferred consists of multiple data packets to be transferred.

[0054] In this embodiment of the invention, in order to improve the transmission efficiency and link load balancing capability of the data flow computing system under high concurrency conditions, the routing decision unit can dynamically trigger routing adjustment strategies based on the real-time data traffic feedback from the traffic counter.

[0055] Each traffic counter periodically collects the data input and output traffic of its connected transmission port at preset time intervals and transmits the collected data traffic to the routing decision unit. This data traffic information may include the number of data bytes transmitted per unit time and the number of data packets.

[0056] After receiving the aforementioned data traffic, the port selection module performs traffic status analysis on the data traffic of each transmission port to determine whether each transmission port is currently in a congested or idle state. When it is detected that the data traffic of a certain transmission port continuously exceeds a first preset traffic threshold for multiple consecutive statistical periods, the port selection module can determine that the transmission port is in a congested state and trigger a routing adjustment strategy. The first preset traffic threshold can be set to more than 90% of the maximum carrying capacity of the transmission port, and can be dynamically configured by software according to the specific network environment and system performance requirements; this invention does not impose any limitations on this.

[0057] During route adjustment, the port selection module will select candidate transmission paths from the remaining output ports of the current routing node that are not congested, choosing those with current data traffic below a second preset threshold. These paths are used to divert some of the data traffic that was originally transmitted through the congested port, thus alleviating local link pressure. During port selection, the port selection module can prioritize checking the traffic status of transmission ports adjacent to the currently congested port in a first preset order (e.g., clockwise or counterclockwise). If the current data traffic of any transmission port is below 50% of its maximum capacity, that port is considered idle and selected as the preferred transmission port.

[0058] Furthermore, to ensure the correctness of the transferred data stream and its consistency with the computational tasks of downstream computing nodes, the data stream selection module will also filter the data streams transmitted through congested transmission ports based on field information in the transmitted data streams. This field information includes, but is not limited to, priority fields, timestamp fields, and dependency identifier fields. The data stream selection module can determine the scheduling priority, generation sequence, and existence of dependencies of data packets based on the above field information, thereby selecting data packets that can be forwarded independently without affecting the execution of subsequent computational tasks as the data streams to be transferred, and completing the routing and forwarding through the selected idle transmission port.

[0059] According to the above embodiments, by introducing a dynamic routing adjustment mechanism based on real-time data traffic feedback, the link scheduling flexibility and data transmission efficiency of the data stream computing system in a high-concurrency environment are effectively improved. By setting a traffic counter to periodically monitor the data input and output traffic of each transmission port, and having the routing decision unit analyze the collected traffic information in real time, congested transmission ports can be identified promptly. Based on this, the routing decision unit, combined with a preset traffic threshold, dynamically selects idle transmission ports as candidate paths and implements routing adjustments to effectively alleviate local link transmission pressure and improve the overall load balancing capability of the system. Furthermore, based on the field information contained in the data stream, the routing decision unit also identifies data packets that can be independently forwarded without affecting the correctness of subsequent computation tasks as data streams to be transferred, ensuring the consistency of transmission order and the accuracy of computation results during data path adjustment, thereby enhancing the operational stability and throughput performance of the computing system when handling large-scale concurrent tasks.

[0060] In some embodiments, the field information includes a priority field. Specifically, the data stream selection module is used to determine, if the priority field is of secondary priority, that the data packet corresponding to that priority field is the data packet of the data stream to be transferred.

[0061] In this embodiment of the invention, the data flow selection module can perform a judgment operation based on the priority field carried in the data packets when filtering data flows to be transferred. When a transmission port is detected to be congested, the data flow selection module identifies the priority of the data packets currently being transmitted on that transmission port. If the value of the priority field in the data packet indicates that the data packet has a secondary priority or is not the highest priority, then the data packet can be identified as a data packet to be transferred and rerouted through an idle transmission port. Conversely, if the value of the priority field indicates that the data packet has the highest priority, then the data packet is given priority to complete data transmission along the original transmission path to avoid transmission delays or calculation result deviations caused by port switching.

[0062] According to the above embodiments, by means of the above method, the computing system can comprehensively consider the priority and processing time requirements of data tasks when performing routing adjustment operations, and prioritize non-critical or delayable data streams for path redirection, thereby effectively alleviating local link congestion, ensuring the transmission stability and real-time performance of critical task data, and further improving the scheduling robustness of the computing system under high load environments.

[0063] In some embodiments, the field information further includes a timestamp field. Specifically, the data stream selection module queries the timestamp field of data packets in the data stream transmitted through a congested transmission port, sorts the data packets according to a second preset order and the timestamp field, and determines that data packets after the preset timestamp are the data packets of the data stream to be transferred.

[0064] In this embodiment of the invention, the field information also includes a timestamp field, used to indicate the generation time of the data packet or the time the data packet entered the transmission network. When the data flow selection module detects that a certain transmission port is in a congested state, it can sort the data packets currently being transmitted on that transmission port based on the timestamp field. The data flow selection module can compare each data packet in the data flow according to a second preset order (e.g., in descending order of timestamp value) and identify the data packet with the largest timestamp value as a newly generated data packet. If it is determined that the newly generated data packet has not yet participated in subsequent calculation dependencies or has not formed a strong transmission dependency chain, the data packet can be preferentially selected as the data packet to be transferred, and data rerouting can be completed through an idle transmission port.

[0065] According to the above embodiments, the timestamp field is used to realize the dynamic diversion of data streams in the transmission port that is in a congested state. This can not only effectively reduce the interference of key data packets that are already being transmitted in the transmission port, but also ensure the temporal integrity of high-priority and early-generated data packets, improve the flexibility and response efficiency of the computing system link scheduling, and help enhance the scheduling robustness and real-time processing capability of the computing system in large-scale and time-sensitive computing task scenarios.

[0066] In some embodiments, the field information further includes a dependency identifier field. Specifically, the data stream selection module queries the dependency identifier field of data packets in the data stream transmitted through a congested transmission port, and determines, based on the dependency identifier field, data packets that have no dependency relationship with other data packets as data packets in the data stream to be transferred.

[0067] In this embodiment of the invention, the field information also includes a Dependency Identifier (ID) field, which is used to indicate the dependencies between data packets. This Dependency Identifier field can be in the form of a unique identifier to indicate whether a data packet depends on the processing results of other data packets, or whether it serves as a dependent input for other data packets.

[0068] When a congestion is detected at a transmission port, the data flow selection module can perform dependency analysis on all data packets in the current data stream transmitted through that port. The data flow selection module extracts the dependency identifier field from each data packet and parses the dependency identifier field to determine whether the current data packet has a preorder or postorder dependency relationship with other data packets.

[0069] If the dependency identifier field of a data packet is empty, or does not identify any preceding or following dependencies, it is determined that the data packet has no dependencies on other data packets and is transferred from the congested transmission port to a selected idle transmission port. If the dependency identifier field of a data packet contains explicit dependency relationships, such as waiting for the preceding data packet to complete data processing or cooperating with other data packets in calculation, the data packet will continue to be transmitted along the original transmission port to avoid calculation misorder or abnormal results caused by transmission port adjustment.

[0070] According to the above embodiments, by means of the above method, when performing adaptive routing adjustment, a precise identification and separation scheduling mechanism for the dependencies between data is introduced, so that the transmission port in the congested state can prioritize the diversion of independent data streams without dependencies. This not only effectively alleviates the traffic pressure of local links, but also ensures the logical correctness of computing tasks and the consistency of execution order, thereby significantly improving the scheduling robustness and overall operational stability of the computing system in a high-concurrency environment.

[0071] In some embodiments, the routing decision unit is further configured to perform an adjustment count field increment operation after determining that the data packet is a data packet of the data stream to be transferred.

[0072] In this embodiment of the invention, to avoid network resource waste and routing oscillations caused by frequent hops of data packets between different routing nodes, an adaptive rerouting constraint mechanism based on an adjustment count field is further introduced. When performing route adjustment operations, the routing decision unit introduces an adjustment count field (Adjust Number, AN) to record the number of data packet rerouting attempts, serving as a constraint factor in the route adjustment process. This adjustment count field represents the cumulative number of times each data packet's transmission port has been changed due to congestion along its transmission path.

[0073] Initially, the AN field is set to 0. Whenever the data packet undergoes a transmission port adjustment in the routing node due to a route adjustment policy, the AN field value is incremented by 1.

[0074] In some embodiments, the routing decision unit is further configured to query the field information of data packets in the data stream, and if the adjustment count field in the field information is equal to a third preset threshold, select the transmission port that is currently in a congested state for data transmission.

[0075] In embodiments of the present invention, such as Figure 3 and Figure 4 As shown, the data flow path indicated by the red dashed line means that the data is first sent by computing node 2, enters the network through the routing node 1 connected to it, and is routed and forwarded through the transmission port 6 of routing node 1 according to the preset routing path, and is finally transmitted to the target computing node 6.

[0076] like Figure 3 and Figure 4 As shown, when the Adjust Number (AN) field in the data packet reaches the third preset threshold k, even if the current transmission port is congested, the routing adjustment strategy will no longer be triggered to perform a rerouting operation; instead, the original transmission port will continue to transmit data. The third preset threshold k can be configured by the system software, and its specific value is not limited.

[0077] For example, such as Figure 3 As shown, suppose that the traffic counter corresponding to transmission port 6 of routing node 1 detects that the data traffic on this port exceeds 90% of its maximum capacity, and thus determines that transmission port 6 is in a congested state. At this time, Figure 3 The data stream shown by the blue dashed line originally continued to be transmitted through transmission port 6 according to the preset transmission path. Based on the routing adjustment strategy, the routing decision unit selected transmission port 5 as an idle transmission port from other output ports that were not congested, and rerouted the data stream to transmission port 5 for transmission. At the same time, the AN field value in the data packet incremented from its initial value of 0 to 1, recording this path adjustment behavior.

[0078] like Figure 4As shown, the blue data stream is transmitted to routing node 4 via transmission port 5, and is prepared to be further sent to the target computing node 6 via transmission port 8 (labeled in the figure) of routing node 4. If, during this process, the real-time traffic of transmission port 8 exceeds 90% of its maximum capacity, and the AN field value of the current data packet is equal to the third preset threshold k (e.g., k=1), then even if the routing adjustment trigger condition is met, path switching will not be performed. Instead, the current transmission path will be maintained, that is, the data packet will continue to be transmitted to the target computing node 6 via transmission port 8.

[0079] According to the above embodiments, the above method effectively limits the excessive hopping behavior of data packets in the network, reduces the routing table update overhead and link instability risk caused by frequent rerouting, and significantly improves the transmission predictability and control accuracy of the system under high concurrency load scenarios, thereby ensuring the stability and efficiency of the data flow computing architecture in complex network environments.

[0080] This application provides a data stream computing method applied to the aforementioned data stream computing system. This data stream computing method and the data stream computing system in one embodiment of this application are based on the same inventive concept and have similar problem-solving principles. Therefore, the implementation of the data stream computing method is the same as the implementation of the data stream computing system in one embodiment of this application, and repeated details will not be described again. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the system described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0081] This data stream computing method, applied to the aforementioned data stream computing system, includes:

[0082] Routing nodes transmit data streams between different computing nodes via data transmission links;

[0083] The traffic counter acquires the data traffic of the transmission port at preset time intervals and transmits the data traffic to the routing decision unit;

[0084] The routing decision unit performs traffic status analysis on the transmission ports based on data traffic. If it is determined that there are transmission ports in a congested state, it selects an idle transmission port and the data stream to be transferred based on a preset routing adjustment strategy, and transmits the data stream to be transferred through the selected transmission port.

[0085] In some embodiments, if the data traffic exceeds a first preset threshold, the port selection module determines that the transmission port is in a congested state; it queries the traffic status of the transmission ports in the routing nodes according to a first preset order, and determines that the transmission ports with data traffic less than a second preset threshold are in an idle state.

[0086] The data stream selection module determines, based on the field information of the data packets in the data stream, at least one data packet in the data stream transmitted by the congested transmission port as the data stream to be transferred.

[0087] In some embodiments, the field information includes a priority field; if the priority field is a secondary priority, the data stream selection module determines that the data packet corresponding to the priority field is the data packet of the data stream to be transferred.

[0088] In some embodiments, the field information further includes a timestamp field; the data stream selection module queries the timestamp field of data packets in the data stream transmitted by the transmission port in a congested state, and sorts the data packets according to the second preset order and the timestamp field to determine that the data packets after the preset timestamp are the data packets of the data stream to be transferred.

[0089] In some embodiments, the field information further includes a dependency identifier field; the data stream selection module queries the dependency identifier field of data packets in the data stream transmitted by the congested transmission port, and determines, based on the dependency identifier field, data packets that have no dependency relationship with other data packets as data packets of the data stream to be transferred.

[0090] In some embodiments, the routing decision unit queries the field information of data packets in the data stream. If the adjustment count field in the field information is equal to a third preset threshold, the unit selects the transmission port that is currently in a congested state for data transmission.

[0091] In some embodiments, when it is determined that the data packet is a data packet of the data stream to be transferred, the routing decision unit performs the operation of incrementing the adjustment count field by one.

[0092] The present invention also provides a chip comprising the above-described data stream calculation method and / or the above-described data stream calculation system.

[0093] The present invention also provides a board card including the above-mentioned chip.

[0094] The present invention also provides an electronic device including the above-described board.

[0095] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0096] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0097] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0098] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0099] In the description of this specification, the references to terms such as "an embodiment," "a specific embodiment," "some embodiments," "for example," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0100] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A data stream computing system, characterized in that, include: Multiple routing nodes, multiple computing nodes, and data transmission links; Adjacent routing nodes are connected via the data transmission link; The computing node is connected to at least one of the routing nodes via the data transmission link; The routing node is used to transmit data streams between different computing nodes through the data transmission link; The routing node includes: multiple traffic counters and a routing decision unit; one end of the traffic counter is connected to the routing decision unit, and the other end is connected to the data transmission link through a transmission port, for acquiring the data traffic of the transmission port at preset time intervals and transmitting the data traffic to the routing decision unit; The routing decision unit is used to perform traffic status analysis on the transmission port based on the data traffic; if it is determined that there is a transmission port in a congested state, it selects an idle transmission port and the data stream to be transferred based on a preset routing adjustment strategy, and transmits the data stream to be transferred through the selected transmission port.

2. The system according to claim 1, characterized in that, The routing decision unit includes: The port selection module is used to determine that the transmission port is in a congested state if the data traffic exceeds a first preset threshold; and to query the traffic status of the transmission ports in the routing nodes according to a first preset order to determine that the transmission ports with data traffic less than a second preset threshold are in an idle state. The data stream selection module is used to determine, based on the field information of the data packets in the data stream, at least one data packet in the data stream transmitted by the congested transmission port as the data stream to be transferred.

3. The system according to claim 2, characterized in that, The field information includes a priority field; the data stream selection module is specifically used to determine that if the priority field is a secondary priority, the data packet corresponding to that priority field is the data packet of the data stream to be transferred.

4. The system according to claim 2, characterized in that, The field information also includes a timestamp field; the data stream selection module is specifically used to query the timestamp field of data packets in the data stream transmitted by the transmission port in a congested state, and sort the data packets according to the second preset order and the timestamp field to determine the data packets after the preset timestamp as the data packets of the data stream to be transferred.

5. The system according to claim 2, characterized in that, The field information also includes a dependency identifier field; the data stream selection module is specifically used to query the dependency identifier field of data packets in the data stream transmitted by the transmission port in a congested state, and determine the data packets that have no dependency relationship with other data packets as data packets of the data stream to be transferred based on the dependency identifier field.

6. The system according to claim 1, characterized in that, The routing decision unit is also used to query the field information of data packets in the data stream. If the adjustment count field in the field information is equal to the third preset threshold, the transmission port that is currently in a congested state is selected for data transmission.

7. The system according to claim 6, characterized in that, The routing decision unit is also used to perform the increment operation of the adjustment count field after determining that the data packet is a data packet of the data stream to be transferred.

8. A data stream computation method, characterized in that, The data stream computing system described in any one of claims 1 to 7 includes: The routing node transmits data streams between different computing nodes through the data transmission link; The traffic counter acquires the data traffic of the transmission port at preset time intervals and transmits the data traffic to the routing decision unit; The routing decision unit performs traffic status analysis on the transmission port based on the data traffic; if it determines that there is a transmission port in a congested state, it selects an idle transmission port and the data stream to be transferred based on a preset routing adjustment strategy, and transmits the data stream to be transferred through the selected transmission port.

9. A chip, characterized in that, Includes the data stream computing system as described in any one of claims 1 to 7.

10. A circuit board, characterized in that, Includes the chip described in claim 9.

11. An electronic device, characterized in that, Includes the board as described in claim 10.

Images