Distributed computing cooperative flow routing method and system for satellite time-varying network topology and storage medium

Abstract

The application provides a distributed computing cooperative flow routing method and system for a satellite time-varying network topology and a storage medium, and the distributed computing cooperative flow routing method comprises the following steps: step one, constructing a cooperative flow transmission time-varying graph model under a satellite inter-satellite link dynamic network; step two, a heuristic routing algorithm based on the cooperative flow transmission time-varying graph; solving the available path set through path calculation on the cooperative flow time-varying graph model, setting a scheduling rule based on traffic priority, and constantly selecting the optimal path under the current situation through the evaluation of the paths in the feasible path set. The application has the beneficial effects that: 1. the time used in satellite distributed computing is effectively reduced; and 2. while effectively reducing the cooperative flow completion time, the efficiency of satellite distributed computing is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a distributed computing collaborative flow routing method, system, and storage medium for satellite time-varying network topologies. Background Technology

[0002] In recent years, the widespread use of Inter-Satellite Links (ISLs) and the rapid development of on-orbit processing capabilities have made it possible to network and exchange data between satellites using high-performance onboard processors. Simultaneously, high-speed ISL data exchange provides the transmission conditions for on-orbit distributed computing. For distributed computing operations on satellites, frequent data flow interactions are required between satellite nodes, and the network performance of the Giant Swarm constellation directly affects data flow transmission, thus impacting the performance of on-orbit processing services. The concurrent request requirements vary throughout the distributed computing process of a given service. Taking the most popular distributed computing framework, MapReduce, as an example, its computation process is divided into three stages: the Map stage, the Shuffle stage, and the Reduce stage. The Map stage generates a large amount of intermediate data, which is sent to the nodes in the Reduce stage, generating significant traffic. The Map and Reduce stages primarily involve concurrent requests from local nodes; while the Shuffle stage mainly involves network concurrent requests. The Shuffle stage generates a large number of concurrent data flows in a short period, known as coflow. Typically, collaborative flow can cause the Shuffle phase to account for a significant portion, even exceeding 50%, of the overall runtime of a distributed computing job. Therefore, reducing the completion time of collaborative flow is crucial for improving the performance of distributed computing jobs. Currently, several collaborative flow optimization scheduling methods have been proposed for distributed computing in terrestrial data centers.

[0003] However, unlike terrestrial data center networks, the network topology and link conditions in the low Earth orbit constellation network are in a dynamic process of change. Furthermore, the intermittent interruption of satellite motion links makes communication resources on the satellite network scarce and time-varying. These network conditions, which are very different from those of terrestrial data centers, mean that when collaborative computing is carried out in satellite distributed computing networks, the location of working nodes, the data flow transmission path, and the flow exchange mechanism will all affect the efficiency of collaborative computing.

[0004] The large-scale low-Earth orbit constellation generates a large number of intermediate value exchanges during on-orbit collaborative computing, which can lead to the formation of centralized concurrent data streams and cause congestion of inter-satellite links. The bandwidth capacity of inter-satellite links is limited and time-varying, and the connection and disconnection of inter-satellite links change intermittently over time. This makes it easy for bottleneck links to appear during collaborative stream transmission, which seriously weakens the efficiency of distributed on-orbit data processing of computing power networks.

[0005] Current collaborative flow scheduling is generally applied to terrestrial data center networks. Typically, the network model of a terrestrial data center is abstracted as a huge non-blocking switch connecting all physical hosts. Therefore, it can be assumed that network bottlenecks only exist in the last hop (the link entering the Reducer node) and the first hop (the link exiting the Mapper node). These distributed computing data flow scheduling schemes for data center networks lack adaptability to satellite network conditions. Furthermore, traditional satellite network routing algorithms struggle to meet the requirements of collaborative flow scheduling in distributed computing. Therefore, to meet the needs of deploying distributed computing on low-Earth orbit satellite networks and reduce the completion time of collaborative flows, a routing and scheduling algorithm adapted to dynamic constellation networks is needed. This is the motivation behind this invention. Summary of the Invention

[0006] To address the problems in the prior art, this invention provides a distributed computing collaborative flow routing method, system, and storage medium for satellite time-varying network topologies.

[0007] This invention provides a distributed computing cooperative flow routing method for satellite time-varying network topologies, comprising:

[0008] Step 1: Construct a time-varying graph model of cooperative stream transmission under a dynamic network of inter-satellite links;

[0009] Step 2: Heuristic routing algorithm based on time-varying graph of cooperative flow transmission; solve the available path set by path calculation on the time-varying graph model of cooperative flow, set scheduling rules based on traffic priority, and continuously select the optimal path under the current situation by evaluating the paths in the feasible path set.

[0010] As a further improvement of the present invention, step one further includes:

[0011] Step S1: Obtain constellation orbit parameter information;

[0012] Step S2: Generate an orbital model in the equatorial-geocentric coordinate system and obtain the v of any satellite in the constellation. i Coordinates in the equatorial-geocentric coordinate system (x) i (t),y i (t),z i (t)), calculate any satellite pair v i ,v j Inter-satellite link distance d ij ;

[0013] Step S3: Construct a time-varying connectivity model for inter-satellite links on a constellation network;

[0014] Step S4: Generate a time-varying map G of the satellite networkt = (V, E, W), and then combine the characteristics of collaborative streaming to construct a time-varying graph model for collaborative streaming.

[0015] As a further improvement of the present invention, in step S3, the satellite pair v is calculated. i ,v j The geometric visible range d between m Link capacity C ij The analysis of satellite link and antenna visibility range yields v i ,v j Inter-satellite link time-varying on / off function e ij (t), and then construct the time-varying connectivity graph model of inter-satellite links on the constellation network.

[0016] As a further improvement of the present invention, step two further includes:

[0017] Step 1: In the time-varying graphical model G T In the computational flow f i The optimal path is determined, and the shortest completion time for each flow is calculated based on the optimal path. Flows are then sorted from highest to lowest time, and the longest-running flow is prioritized (f). i_pri Set it to the highest level, and decrease the rest in that order;

[0018] Step 2: In each iteration, find the set of conflicting paths with the longest execution time, and determine the flow f to be scheduled based on the flow priority within this set of conflicting paths. i ;

[0019] Step 3: Evaluate the fitness of the path at the current moment during each iteration;

[0020] Step 4: Adjust the flow scheduling priority;

[0021] Step 5: Determine whether a conflict-free path set has been found or the maximum number of iterations has been reached. If so, output the Cooperative Flow Completion Time (CCT) and the Cooperative Flow Transmission Path Set. Otherwise, return to step 2.

[0022] As a further improvement of the present invention, step 2 further includes:

[0023] Step A1: Calculate f for each flow in the cooperative flow in the cooperative flow time-varying graphical model. i Available path set

[0024] Step A2: Find the set of flows Ω in the cooperative flow time-varying graph model that has the longest link transmission time. c This schedules the lower-priority flows in the flow set.

[0025] As a further improvement of the present invention, in step 3, for all paths, the time increment C1 caused by selecting the path to other flows, the time increment C2 caused to the completion time of the cooperating flow, and the difference C3 between the current path and the optimal path are calculated. α1 and α2 (0≤α1, α2≤1) are defined as weight factors to measure the weight of different influencing factors in the path selection process. After selecting appropriate α1 and α2, the current optimal path p is selected according to the fitness function score=α1C1+α2C2+(1-α1-α2)C3 in the scheduling process. * .

[0026] As a further improvement of the present invention, in step 4, the longest flow set Ω is... c The priority f of the streams that have already been scheduled i_pri Set to the highest level.

[0027] The present invention also discloses a distributed computing collaborative flow routing system for satellite time-varying network topology, comprising: a memory, a processor, and a computer program stored in the memory, wherein the computer program is configured to implement the steps of the distributed computing collaborative flow routing method of the present invention when invoked by the processor.

[0028] The present invention also discloses a computer-readable storage medium storing a computer program configured to implement the steps of the distributed computing cooperative flow routing method of the present invention when invoked by a processor.

[0029] The beneficial effects of this invention are: 1. It effectively reduces the time required for satellite distributed computing; 2. While effectively reducing the completion time of collaborative flows, it effectively improves the efficiency of satellite distributed computing. Attached Figure Description

[0030] Figure 1 This is a flowchart of the distributed computing collaborative flow routing method for satellite time-varying network topology of the present invention;

[0031] Figure 2 This is a diagram showing the position of the satellite of this invention in the geocentric-equatorial coordinate system;

[0032] Figure 3 This is a diagram of inter-satellite links between adjacent orbits in the constellation of this invention;

[0033] Figure 4 This is a diagram showing the disconnection and reconstruction of the inter-orbit inter-satellite link in the polar region according to the present invention;

[0034] Figure 5 This is a diagram showing the network topology changes of the satellite at different times according to the present invention;

[0035] Figure 6 This is the time-varying diagram of collaborative stream transmission in this invention;

[0036] Figure 7 This invention provides a total load diagram of the cooperative flow with fixed working nodes.

[0037] Figure 8 This is a diagram showing the increase in the number of working nodes with a fixed total workload according to the present invention;

[0038] Figure 9 This invention changes the time variation of collaborative stream transmission under different satellite link establishment conditions. Detailed Implementation

[0039] like Figure 1 As shown, this invention discloses a distributed computing cooperative flow routing method for satellite time-varying network topologies, comprising:

[0040] Step 1: Construct a time-varying graph model of cooperative stream transmission under a dynamic network of inter-satellite links;

[0041] To overcome the negative impact of the highly dynamic topology and intermittent link connections of LEO satellite networks on cooperative flow routing, a time-varying graphical model is proposed to describe the time-varying network topology of satellites. By constructing an on / off model of inter-satellite links and a model of available transmission rates for inter-satellite links, the dynamic changes in satellite topology and transmission resources are accurately described. Furthermore, by combining satellite dynamic topology modeling with a cooperative flow transmission model on the satellite constellation, the transmission status of cooperative flows at different times can be accurately described as the satellite network topology dynamically changes.

[0042] Step 2: Heuristic routing algorithm based on time-varying graph of cooperative flow transmission; solve the available path set by path calculation on the time-varying graph model of cooperative flow, set scheduling rules based on traffic priority, and continuously select the optimal path under the current situation by evaluating the paths in the feasible path set.

[0043] Step one also includes:

[0044] Step S1: Obtain constellation orbit parameter information;

[0045] Step S2: Generate an orbital model in the equatorial-geocentric coordinate system and obtain the v of any satellite in the constellation. i Coordinates in the equatorial-geocentric coordinate system (x) i (t),y i (t),z i (t)), calculate v i ,v j Inter-satellite link distance

[0046] Step S3: Construct a time-varying connectivity model for inter-satellite links on a constellation network;

[0047] Step S4: Generate a time-varying map G of the satellite network t = (V, E, W), and then combine the characteristics of collaborative streaming to construct a time-varying graph model for collaborative streaming.

[0048] In step S2, the inter-satellite link connectivity calculation process is as follows:

[0049] First, the low-Earth orbit satellite orbits are described in the equatorial-geocentric coordinate system, such as... Figure 2 As shown, the satellite's geocentric coordinate system has the Earth's center as the origin, and the positive half-axis of the three-dimensional coordinate system pointing from the Earth's center to the vernal equinox lies in the equatorial plane; the positive half-axis of the Z-axis points from the Earth's center to the North Pole; the Y-axis conforms to the right-handed Cartesian coordinate system; and the planes of the X-axis and Z-axis contain the Prime Meridian.

[0050] In this coordinate system, the ascending node of a satellite's orbit is defined as the intersection of its orbital plane with the equatorial plane when the satellite orbits the Earth from south to north, and the descending node is the intersection of its orbital plane with the equatorial plane when the satellite orbits the Earth from north to south. The line of intersection is the line connecting the two intersection points of the orbital plane and the equatorial plane and passing through the Earth's center. In the figure, α is the angle between the satellite's orbital plane and the equatorial plane, φ is the angle between the ascending node and the Y-axis, φ is the initial phase angle of the satellite, and ω is the initial phase angle of the satellite. s Let be the satellite's rotational angular velocity, and t be the satellite's orbital time. Based on these parameters, the satellite's position coordinates on the coordinate axes at time t can be calculated as follows:

[0051]

[0052] in Given the satellite phase after time t, based on the satellite's position coordinates in the coordinate system, the phase of any satellite in the constellation can be calculated. i With v j The distance between them can be calculated as follows:

[0053]

[0054] The Walker Delta constellation is a common constellation configuration containing multiple satellites with the same orbital altitude and inclination, typically between 40° and 60°. Inter-satellite links within the same orbit in a Walker Delta constellation are generally considered stable, while links between satellites in different orbits are mostly temporary. Factors such as the relative orbital direction of the satellites, their attitude, and the angle of their antennas all affect the connectivity of these links. The visibility of inter-satellite links can be measured by three metrics: geometric visibility, antenna visibility, and inter-satellite link capacity. The following section first calculates the maximum link distance required to maintain geometric visibility, i.e., line-of-sight visibility.

[0055] Due to Earth's obstruction, establishing an inter-satellite link (ISL) requires that the satellites be visible at line of sight. Simultaneously, the ISL must also avoid the ionosphere, the height of which is denoted by h. χ This means that h usually χ It is estimated to be around 80km. Let R be... E h is the Earth's radius. s Given the satellite's orbital altitude, the maximum distance at which the line-of-sight (ISL) remains visible is:

[0056]

[0057] The connection or disconnection of satellite links is also related to the change in azimuth angle between satellite antennas. Satellites in the Walkerdelta constellation, which are close to high latitudes, will undergo a transition process from latitude increase to decrease. During this process, the inter-satellite links between adjacent orbital satellites may be shut down. The present invention will explain this by specifically calculating the rate of change of azimuth angle between satellites.

[0058] In the Walker Delta constellation, any two adjacent orbits will form two intersection points on the north and south sides of the Earth, such as... Figure 3 As shown, in the geocentric-equatorial coordinate system, p i With p j PN and PS are the orbits of two adjacent satellites in the constellation, located at the northern and southern ends of the Earth, respectively. Where φ... i With φ j Let ω be the initial phase angle of the satellite. s φ is the angular velocity of the satellite in its orbit. N and These are the phase angles from the ascending nodes of the two satellite orbits to their north nodes. According to the spherical cosine theorem, the angle Φ between the two satellite orbital planes can be expressed as the angle Δβ = β between the two ascending nodes and the Earth's center. i -β j Calculations show that:

[0059]

[0060] This allows us to calculate the central angle r between two satellites in adjacent orbits within a constellation. ij :

[0061]

[0062] in Given the phase angle of the satellite at the current moment, substituting equation (5) into (6) yields the following:

[0063]

[0064] The establishment and maintenance of satellite links require the alignment and tracking of the satellite's onboard antennas. When the rate of change in distance and azimuth of the inter-satellite link between satellites on opposite orbits is too rapid, the satellite link cannot be maintained. The change in the relative position of the satellites can be measured by calculating the azimuth angle of their relative positions. The azimuth angle of the satellite link direction can be defined as the angle from the north-pointing line at one end of the link, clockwise, to the link direction.

[0065] Due to the high-speed motion of satellites, the azimuth angles between the two satellites in an inter-plane satellite link undergo complex changes. (Satellite v) j to v i The azimuth angle can be calculated as follows:

[0066]

[0067] In the formula φ i With φ j Let r be the initial phase angle of the satellite. ij φ is the angular distance between satellites. N and These are the phase angles from the ascending node of the two satellite orbits to the north node of the two orbits, respectively.

[0068] To obtain The rate of change can be calculated. For r ij Second-order partial derivatives:

[0069]

[0070] Where F(r) ij The limit of ) with r ij It approaches zero and nears infinity. This means The rate of change with r ij It increases as it decreases. For example, Figure 4 As shown, when r ij When it reaches its minimum near the intersection of the orbits The rate of change is the largest, which makes target tracking and aiming very difficult.

[0071] Furthermore, due to the intersecting orbits of the Walker constellation, when two satellites linked by interplane inter-satellite links pass through the intersection point P... N or P S At this time, the satellite's onboard antenna must rotate 180° to maintain the previous connection. When the two satellites pass point P... NTo maintain the inter-plane ISL connection, the satellite antenna needs to rotate 180°, meaning the two antennas must swap positions, which is generally considered difficult to achieve. Therefore, this invention temporarily disables the inter-plane ISL before crossing the intersection point and re-establishes it after leaving the intersection point.

[0072] To ensure the stability of the inter-satellite link, when the angular distance between the satellites is less than the minimum angular distance r... min When the satellite link is closed, the Euclidean distance between the satellites is d. min Furthermore, the number of transmitting and receiving devices that a satellite can carry and maintain is limited. Following common assumptions, it is generally stipulated that a satellite can maintain one transmitting and receiving antenna between adjacent orbits.

[0073] Let L ij Here, FSPL (Free Space Path Loss) is the base frequency, and f is the carrier frequency. All antennas have fixed transmit power and gain in the main lobe direction; the satellite antenna transmit power is P. t The transmit gain is G. t The receiver gain is represented by G. r Therefore, the signal strength received from point i to point j is:

[0074]

[0075] Where c is the speed of light and f is the carrier frequency. This invention assumes that all satellites have sufficiently narrow antenna beams and precise beam alignment capabilities. Therefore, satellites can communicate in an interference-free environment. In an interference-free environment, the gain of all established ISLs is G. t G r At any given time, from v j to v i The signal-to-noise ratio (SNR) of continuous transmission can be expressed as:

[0076]

[0077] Where k B τ is the Boltzmann constant, τ is the thermal noise in Kelvin, and B is the channel bandwidth in Hertz.

[0078] Satellite v can be calculated using Shannon's formula. i With v j Link capacity C between ij for:

[0079]

[0080] Assuming the minimum required capacity to maintain inter-satellite links is C0, then if C0 < C ij It is believed that v j With vi The link between them is possible; otherwise, v i With v j The links between them are not connected.

[0081] In step S3, the specific process of constructing the time-varying connectivity model of inter-satellite links on the constellation network is as follows:

[0082] Inter-satellite links exhibit time-varying characteristics. When factors such as link signal-to-noise ratio and antenna tracking pointing do not meet the basic conditions for link establishment and data transmission, inter-satellite links will exhibit periodic changes of linking-disconnecting-connecting. Therefore, LEO satellite networks are time-varying networks with highly dynamic topologies, and static graph models of terrestrial networks are not applicable to satellite networks. This invention aims to use time-varying graph models to characterize the time-varying topology and resource distribution of LEO satellite networks, providing a reliable model basis for routing design.

[0083] Time-varying models of satellite dynamic networks can be modeled as a set of time-varying graphs:

[0084] G t =(V,E,W) (13)

[0085] Where V is the set of vertices in the graph, representing satellite nodes. E is the set of edges in the graph, i.e., inter-satellite links (ISLs), where e uv ∈E represents the link between node u and node v, and E=V×V={e uv}, u, v ∈ V. W is the set of edge weights of the satellite graph, ω uv The physical meaning of ∈W is satellite link e uv The available transmission rate on the satellite link is worth noting. uv Available inter-satellite transmission rate ω uv Both are functions of time t. That is, they exist:

[0086]

[0087] In a satellite network topology diagram, the types of ISLs between any two adjacent satellites include in-plane ISLs (where the two satellites are in the same orbital plane) and inter-plane ISLs (where the satellites are in adjacent orbital planes). Each node typically has two in-plane ISLs and two inter-plane ISLs. In-plane ISLs are used to connect adjacent nodes between its preceding and following satellites in the same orbit. Inter-plane ISLs are used to connect adjacent satellites in two adjacent orbits. Based on this network connection method, the LEO satellite network structure at a given time slice can be viewed as a two-dimensional Manhattan network.

[0088] During data transmission, the satellite network topology may change due to the connection or disconnection of inter-satellite links. To describe the transmission of traffic across time slots on the satellite network, snapshots of different time slots can be used to describe the satellite topology. If a satellite link is disconnected during traffic transmission, the traffic cannot continue to be transmitted. Figure 5 As shown, the on / off relationship of the link changes in different time slots, which in turn causes topology changes.

[0089] To accurately describe the transmission process of coflows on time-varying networks, this invention first analyzes and models the transmission characteristics of coflows. In the distributed computing framework MapReduce, the process of transferring intermediate data generated by the Mapper to the Reducer is called Shuffle, and the Reducer task cannot be started until the Shuffle is complete. This Shuffle data flow is defined as a coflow. The flow in the Shuffle phase of distributed computing is a set of multiple parallel flows, and its completion time is determined by the completion time of the last flow in the set. This invention defines this set as a coflow set, denoted as...

[0090] F = {f i}={(s i ,d i ,v i )},1≤i≤N (15)

[0091] In the formula, F represents the input Coflow information vector, f i Let s be the i-th flow in a cooperative flow. i This indicates the source node of the stream (usually a Mapper node), d i This is represented as the destination node of the stream (usually a Reducer node), v i This is represented as the data volume of the stream, which is the total amount of data that the stream needs to transmit. Here, the present invention assumes that this information can be captured by upper-layer applications or obtained using existing prediction techniques, thereby supporting the subsequent work of the present invention.

[0092] For each flow f in the cooperative flow i For each ∈F, there is a corresponding path from the Mapper to the Reducer, represented as...

[0093] P = {P1, P2, ..., P} i ,...,P N} (16)

[0094]

[0095] In the formula P i Indicates flow f i Mi Available paths and P represents i One path. Once the transmission paths of all flows are determined, the path vector for cooperative flow routing is...

[0096]

[0097] The bandwidth allocation scheme for the collaborative flow of this invention is as follows:

[0098]

[0099] In the formula, b * (t) represents the set of bandwidth allocation plans for all flows. Indicates flow f i Bandwidth allocation at different times.

[0100] In the time-varying satellite network model, for each flow, f i Completion time T i Transmission time ξ transmit and propagation time ξ propaga And waiting time ξ wait Composition, formula is

[0101] T i =ξ transmit +ξ wait +ξ propaga (20)

[0102] The first term on the right-hand side of the equation represents the transmission time; the second term represents the waiting time of the stream's transmission path in the event of a disruption in the satellite network topology; and the third term is the propagation time of multi-hop ISLs, which is the sum of the microwave or laser propagation times of multiple ISLs. For the transmission delay ξ of the stream... transmit As mentioned earlier, the available transmission rate allocated to a stream changes at different times due to the time-varying satellite topology. Assume stream f... i Transmitting over L time slices results in:

[0103]

[0104] Where {τ1,τ2,…τ L} represents the lifetime of the satellite topology, therefore we have The transmission traffic in each time slice is equal to the time slice length multiplied by the transmission rate allocated to that stream during that time period. Therefore, the transmission delay ξ... transmit It can be represented as:

[0105]

[0106] To illustrate the above process, the existing flow f i Reach the destination via the path {S-S10-D}, such as Figure 6 As shown, flow f i The transmission process occurs through three time slots: τ1, τ2, and τ3. The available transmission rate of the stream varies with time in each of the three time slots, and can be expressed as follows: Due to the link in time slot τ2 Unable to meet streaming requirements, streaming is interrupted. Adding a waiting link between node replicas created between time slots τ2 and τ3 indicates that traffic is now in a waiting phase until the inter-satellite link recovers. It is easy to see from the diagram that flow f... i Waiting time ξ wait =τ2. The rate of cooperative stream transmission at each moment is equal to the available transmission rate of the most congested link at that moment. The co-track links can be calculated using equations (1) and (2). The inter-satellite link distance is 5800km, from which its usable link capacity can be calculated to be 400Mbps. Similarly, the link capacity can be calculated. The maximum available rate is 365 Mbps. Due to the bottleneck effect of streaming transmission, the current in time slot f at τ1 is limited. j The maximum rate is 365 Mbps, while the transmission rate in time slot τ2 is 0 due to the link interruption. In time slot τ3, there is flow f. j Similarly, it passes through the link. Assume there is Due to contention for transmission of the stream, stream f j The maximum rate on time slot τ3 is 200 Mbps.

[0107] Step two also includes:

[0108] Step 1: In the time-varying graphical model G T In the computational flow f i The optimal path is determined, and the shortest completion time for each flow is calculated based on the optimal path. Flows are then sorted from highest to lowest time, and the longest-running flow is prioritized (f). i_pri Set it to the highest level, and decrease the rest in that order;

[0109] Step 2: In each iteration, find the set of conflicting paths with the longest execution time, and determine the flow f to be scheduled based on the flow priority within this set of conflicting paths. i The goal is to minimize competition between high-priority flows and other flows.

[0110] Specifically, it includes:

[0111] Step A1: Calculate f for each flow in the cooperative flow in the cooperative flow time-varying graphical model. i Available path set

[0112] Step A2: Find the set of flows Ω in the cooperative flow time-varying graph model that has the longest link transmission time. c This schedules the lower-priority flows in the flow set.

[0113] Step 3: In each iteration, evaluate the fitness of the path at the current time. For all paths, calculate the time increment C1 caused by selecting this path to other flows, the time increment C2 caused to the completion time of the cooperating flow, and the difference C3 between the current path and the optimal path. Based on the fitness function score = α1C1 + α2C2 + (1 - α1 - α2)C3 in the scheduling process, select the current optimal path p. * ;

[0114] Step 4: Adjust flow scheduling priority; select the longest flow set Ω. c The priority f of the streams that have already been scheduled i_pri Set to the highest level to avoid potential circular scheduling caused by only scheduling lower-priority flows on the longest conflicting path each time;

[0115] Step 5: Determine whether a conflict-free path set has been found or the maximum number of iterations has been reached. If so, output the Cooperative Flow Completion Time (CCT) and the Cooperative Flow Transmission Path Set. Otherwise, return to step 2.

[0116] Heuristic routing algorithms based on time-varying graphs of cooperative stream transmission:

[0117] This invention uses CCT to represent the completion time of the Shuffle stage, i.e., the Coflow Completion Time. Its value depends on the slowest stream. This invention uses T... i To represent flow f i The transmission completion time can obviously be used to derive the formula for CCT.

[0118]

[0119] Here, the present invention uses two binary variables. and The former variable Can determine ISL Does it include link e? uv .when Includes e uv This invention sets otherwise The latter variable Indicate path Whether it was flowed i The selected option. If The flow f iChoose, then otherwise The present invention can be represented as follows:

[0120]

[0121]

[0122] Meanwhile, the available bandwidth of the link is limited, so b i The settings are naturally subject to constraints. First, the bandwidth of each stream must be non-negative, i.e., b i ≥0; then on any link e uv (t) The total bandwidth allocated to each flow in the cooperative flow at any given time cannot exceed the total available transmission rate C of the link at that time. uv (t), the formula can be expressed as:

[0123]

[0124] Each flow has many traversable paths in the network model graph. This invention uses P... i To represent flow f i In the set of possible paths in the graph, flow f i The optimal path lies within this set; this invention only needs to search for its optimal path within this set. To ensure the uniqueness of the path, this invention can derive a constraint expression.

[0125]

[0126] Furthermore, to ensure path determinism, this invention provides a flow conservation constraint formula. This invention first introduces a new variable N(u), representing the set of neighboring nodes of satellite node u∈U in the network graph. The flow conservation constraint formula is as follows:

[0127]

[0128]

[0129]

[0130] Constraint formula (3-13) ensures that data originates from the source node through only one link; constraint formula (3-14) ensures that data is sent to the destination node through only one link; constraint formula (3-15) ensures that for any flow and any intermediate node u it passes through, the number of links sent to node u is equal to the number of links originating from node u.

[0131] In summary, the shortest completion time optimization problem for cooperative flow in time-varying graphs can be formulated as follows:

[0132] Optimization goal:

[0133] Constraints: (26)(27)(28)(29)(30)

[0134] Standard algorithms consume a significant amount of time to obtain the optimal solution to NLP problems. As the number of network nodes and worker nodes increases, the complexity of the solution can increase exponentially. Therefore, traversing all path combinations to solve the optimization problem is impractical. To address this, a heuristic algorithm based on path scoring and dynamic flow priority scheduling is proposed. The algorithm flow is shown below.

[0135] Algorithm 3: Heuristic Routing Algorithm Based on Time-Varying Graph of Cooperative Flow Transmission

[0136] Input: Set of cooperative flows F,

[0137] Output: Cooperative flow final path vector P * Collaborative Flow Completion Time (CCT)

[0138]

[0139]

[0140] Simulation experiments verify:

[0141] Experimental simulations were conducted in a Walker-delta constellation with a configuration of (72 / 9 / 1). The simulation parameters are shown in the table below. To verify the effectiveness of the proposed algorithm, this algorithm will be compared with the Dijkstra shortest path algorithm based on time-varying graphs (TD), the Software Defined Routing Algorithm in LEO Satellite Networks (SDRA) algorithm using SDN, and the Equal-Cost Multipath Routing (ECMP) algorithm commonly used for ground cooperative flow routing.

[0142] Table 1 Simulation parameter settings

[0143]

[0144]

[0145] Experiment 1: Fixed number of working nodes, increase total traffic load

[0146] like Figure 7As shown in the results of Experiment 1, we can see that:

[0147] 1. For all routing algorithms, with the number of flows remaining constant, the required completion time increases as the total load increases. This is because the increased flow load requires the network to transmit data for longer periods.

[0148] 2. Compared to the Equivalent Multi-Path Routing (ECMP) algorithm commonly used for terrestrial cooperative flows, the heuristic routing algorithm based on time-varying graphs of cooperative flow transmission (CFTHR) proposed in this invention performs better. This is because ECMP does not consider network outages, resulting in poor load balancing performance. While the SDN-based Low-Earth Orbit Constellation Routing (SDRA) algorithm considers link performance and load balancing in dynamic networks, it cannot guarantee a reasonable scheduling order for cooperative flows, leading to weaker load balancing performance compared to the algorithm proposed in this invention.

[0149] like Figure 8 As shown, Experiment 2: With a fixed total load, the number of working nodes is increased.

[0150] The results of Experiment 2 show that:

[0151] 1. For all routing algorithms, the total load of the flow remains constant. When the number of flows is reduced, the required completion time increases because the increased load of the flow requires the network to transmit for a longer period of time.

[0152] 2. When the total workload is fixed and the number of working nodes is increased, the proposed algorithm shows a significant performance improvement over SDRA due to its more reasonable flow routing selection order.

[0153] Experiment 3: Fix the working node and load, and change the link on / off time.

[0154] like Figure 9 As shown in the results of Experiment 3, we can see that:

[0155] 1. After changing the antenna tracking capability, the available link establishment time of the satellite link changes within a cycle. It is not difficult to find that all routing algorithms will cause the time for cooperative flow to increase as the number of available links in the network topology decreases.

[0156] 2. Under different topology conditions, the algorithm proposed in this invention can achieve performance improvements compared to routing algorithms applied to satellite networks and other commonly used terrestrial data center load balancing algorithms. This proves the effectiveness of CFTHR in different constellation topologies.

[0157] This invention proposes a distributed computing collaborative flow routing method for satellite time-varying network topologies, which effectively reduces the time required for satellite distributed computing. The innovations include the following parts:

[0158] 1. A time-varying graph model for cooperative flow transmission is proposed to describe changes in satellite topology. This model can accurately describe the dynamic changes of satellite communication resources over time. A transmission model for cooperative flow on a constellation network is established, and the proposed time-varying graph accurately describes the transmission status of cooperative flow at different times as the satellite network topology changes dynamically.

[0159] 2. A heuristic routing algorithm based on a time-varying graph for collaborative flow transmission is proposed. Building upon the proposed time-varying graph for collaborative flow transmission in low-Earth orbit satellite networks, the algorithm calculates the available path set on the time-varying graph. A scheduling rule based on traffic priority is established, and the optimal path for the current situation is continuously selected by evaluating paths in the feasible path set. This algorithm can adapt well to dynamic network changes within the constellation, effectively reducing the collaborative flow completion time. Furthermore, compared with satellite routing algorithms proposed in recent years, it is more suitable for collaborative flow transmission mechanisms, effectively improving the efficiency of satellite distributed computing.

[0160] The beneficial effects of this invention are: 1. It effectively reduces the time required for satellite distributed computing; 2. While effectively reducing the completion time of collaborative flows, it effectively improves the efficiency of satellite distributed computing.

[0161] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A distributed computing cooperative flow routing method for satellite time-varying network topology, characterized in that... ,include: Step 1: Construct a time-varying graph model of cooperative stream transmission under a dynamic network of inter-satellite links; Step 2: Heuristic routing algorithm based on time-varying graph of cooperative flow transmission; solve the available path set by path calculation on the time-varying graph model of cooperative flow, set scheduling rules based on traffic priority, and continuously select the optimal path under the current situation by evaluating the paths in the feasible path set; Step two also includes: Step 1: In the time-varying graphical model Middle Computation Flow The optimal path is determined, and the shortest completion time for each flow is calculated based on the optimal path. Flows are then sorted by time from highest to lowest, and the longest-running flow is prioritized. Set it to the highest level, and decrease the rest in sequence; Step 2: In each iteration, find the set of conflicting paths with the longest execution time, and determine the flows to be scheduled based on the flow priority within this set of conflicting paths. ; Step 3: Evaluate the fitness of the path at the current moment during each iteration; Step 4: Adjust the flow scheduling priority; Step 5: Determine if a conflict-free path has been found. If the solution set or the maximum number of iterations is reached, output the Cooperative Flow Completion Time (CCT) and the Cooperative Flow Transmission Path Set; otherwise, return to step 2. Step 2 also includes: Step A1: Calculate the value of each flow in the cooperative flow in the cooperative flow time-varying graphical model. Available path set ; Step A2: Find the set of flows with the longest link transmission time in the cooperative flow time-varying graph model. Schedule the lower-priority flows in this flow set; In step 3, the time increment caused by selecting a path to other flows is calculated for each path. The incremental time caused by collaborative flow completion The difference between the current path and the optimal path ,definition and ,in These are weighting factors used to measure the weight of different influencing factors in the path selection process, in order to select an appropriate path. and Then, based on the fitness function during the scheduling process Select the current optimal path ; In step 4, the longest set of flows is... Priority of flows that have been scheduled Set to the highest level.

2. The distributed computing collaborative flow routing method according to claim 1, characterized in that, Step one also includes: Step S1: Obtain constellation orbit parameter information; Step S2: Generate an orbital model in the equatorial-geocentric coordinate system and obtain the orbital model of any satellite in the constellation. Coordinates in the equatorial-geocentric coordinate system Calculate any satellite pair Inter-satellite link distance ; Step S3: Construct a time-varying connectivity model for inter-satellite links on a constellation network; Step S4: Generate a time-varying map of the satellite network Then, a time-varying graph model for collaborative streaming is constructed by combining the characteristics of collaborative streaming.

3. The distributed computing collaborative flow routing method according to claim 2, characterized in that, In step S3, satellite pairs are calculated. Geometric visible range Link capacity The analysis of satellite links and antenna visibility range yielded the following results. Inter-satellite link time-varying on / off function This leads to the construction of a time-varying connectivity graph model for inter-satellite links on a constellation network.

4. A distributed computing collaborative flow routing system for satellite time-varying network topologies, characterized in that, include: A memory, a processor, and a computer program stored on the memory, the computer program being configured to implement the steps of the distributed computing cooperative flow routing method of any one of claims 1-3 when invoked by the processor.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program configured to implement the steps of the distributed computing cooperative flow routing method of any one of claims 1-3 when invoked by a processor.

Images