CN115426025A - Low-orbit satellite network-oriented distributed computing data flow scheduling and routing method - Google Patents
Low-orbit satellite network-oriented distributed computing data flow scheduling and routing method Download PDFInfo
- Publication number
- CN115426025A CN115426025A CN202210793602.1A CN202210793602A CN115426025A CN 115426025 A CN115426025 A CN 115426025A CN 202210793602 A CN202210793602 A CN 202210793602A CN 115426025 A CN115426025 A CN 115426025A
- Authority
- CN
- China
- Prior art keywords
- flow
- satellite
- network
- distributed computing
- routing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/1851—Systems using a satellite or space-based relay
- H04B7/18519—Operations control, administration or maintenance
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
Abstract
The invention is suitable for the field of information communication technology improvement, and provides a distributed computing data flow scheduling and routing method for a low-orbit satellite network, which comprises the following steps: s1, constructing a satellite distributed computing frame based on an SDN (software defined network), simulating a giant constellation satellite orbit by using satellite simulation software, and acquiring a satellite network model; s2, establishing a virtual network simulating the satellite network topology by utilizing Mininet according to the acquired satellite network model; s3, configuring a routing algorithm by using the SDN controller; and S4, configuring a distributed computing framework Hadoop in a host node in the satellite network and deploying actual computing tasks to realize satellite distributed computing data packet level simulation. The method comprises the steps of obtaining a low orbit satellite network model through simulation of a giant constellation orbit, combining advantages of an OpenFlow flow table and topology perception and centralized control of an SDN controller, providing a flooding-free route establishment mechanism of a network topology, and improving forwarding efficiency of satellite network information.
Description
Technical Field
The invention belongs to the field of improvement of information communication technology, and particularly relates to a distributed computing data flow scheduling and routing method for a low-earth orbit satellite network.
Background
In recent years, with the wide use of Inter-Satellite Link (ISL) and the rapid development of On-board processing (OBP), it is possible to implement space-based multi-source data fusion and On-orbit intelligent computation by using networking exchange capability between a high-performance Satellite-borne processor and a Satellite. Due to the characteristics of large capacity, high speed, full coverage and the like, the low-orbit broadband constellation is becoming an important infrastructure for constructing a new generation ubiquitous communication and service integrated network, namely the ubiquitous network. The giant constellation satellite has wide coverage range and can provide communication service for traditional communication limited areas such as remote areas, unmanned zones, disaster areas and the like; on the other hand, the satellite processing capacity can be used as a supplement of a ground network, so that the overloaded ground flow is absorbed, and the quality of satellite communication service is improved. The satellite network integrating communication and service realizes inter-satellite cooperative computing by utilizing on-satellite processing capacity, and avoids the problem of great time delay brought to services due to serious broadband limitation between the ground and the satellite when the services are unloaded in a ground cloud computing center.
For distributed computing operation processed on the satellite, frequent data stream interaction needs to be performed among satellite nodes, and the network performance of a giant constellation directly influences the transmission of data streams, so that the performance of on-orbit processing services is influenced. In a distributed computing process of a service, concurrent request requirements are different. Taking the most popular distributed computing framework MapReduce as an example, the computing process is divided into three phases, namely, a mapping phase (Map), a shuffling phase (Shuffle), and a reduction phase (Reduce), as shown in fig. 1. The Map phase may generate a large amount of intermediate data, which may generate a large amount of traffic to be sent to the node where the Reduce phase is located. The Map and Reduce phases are mainly concurrent requests of local nodes; while the Shuffle phase is primarily a network concurrent request. The Shuffle phase generates a large number of concurrent data flows in a short time, called a flow (flow). Often co-streaming can cause the Shuffle phase time to occupy a significant portion, even more than 50%, of the overall distributed computing job runtime. To improve the execution performance of distributed computing jobs, it is important to reduce the completion time of the cooperative flow. Currently, some collaborative flow optimized scheduling methods have been proposed for collaborative flows of distributed computing in terrestrial data centers. The network model of a typical terrestrial data center is abstracted as a huge non-blocking switch connecting all physical hosts, so it can be assumed that network bottlenecks exist only at the last hop (the link going into the Reducer node) and the first hop (the link coming out of the Mapper node). The data flow scheduling schemes for distributed computing of the data center network lack the adaptive capacity to the satellite network conditions and the efficient scheduling strategy under the heterogeneous link state and the latticed multi-hop topological structure in the satellite network. However, the conventional routing strategy, such as a Shortest Path (SP) routing algorithm or an Equal Cost Multiple Path (Equal Cost Multiple Path) routing algorithm, is simply used in the satellite network, and cannot adapt to the problem caused by distributed computation of the cooperative flow.
Therefore, an optimized scheduling routing method for distributed computing data streams of a low-orbit satellite network is needed.
Disclosure of Invention
The invention aims to provide a low-orbit satellite network-oriented distributed computing data flow scheduling and routing method, and aims to solve the technical problems.
The invention is realized in this way, and is a low orbit satellite network-oriented distributed computing data flow scheduling and routing method, which comprises the following steps:
s1, constructing a satellite distributed computing frame based on an SDN (software defined network), simulating a giant constellation satellite orbit by using satellite simulation Software (STK) and acquiring a satellite network model;
s2, establishing a virtual network simulating the satellite network topology by utilizing Mininet according to the acquired satellite network model;
s3, configuring a routing algorithm by using an SDN controller;
and S4, configuring a distributed computing frame Hadoop in a host node (satellite node) in the satellite network and deploying an actual computing task to realize the simulation of the satellite distributed computing data packet level.
The further technical scheme of the invention is as follows: and the SDN controller directly calculates the route through APP analysis to realize the non-flooding point-to-point forwarding.
The invention further adopts the technical scheme that: the method comprises the steps that ephemeris and connection relations of constellations are loaded in an SDN controller in advance, when an ARP packet reaches an OpenFlow switch, whether a corresponding matched flow entry exists or not is searched in a flow table of the switch, and if the corresponding matched flow entry does not exist, a complete flow table is established through a continuous packet _ in process to achieve network communication.
The further technical scheme of the invention is as follows: the method comprises the steps that a packet _ in process is sent to an SDN controller through a packet _ in message, the SDN controller analyzes an ARP packet to obtain a source address and a destination address of the ARP packet, then a shortest path between the source address and the destination address is calculated according to a satellite connection relation at the current moment, the shortest path is written into a flow table, the process is repeated continuously until a complete flow table is established, the SDN controller enters a flow scheduling stage after the complete flow table is established, and the flow table is updated continuously by the SDN controller according to a specific data flow scheduling strategy and a network state.
The further technical scheme of the invention is as follows: the routing scheduling algorithm for minimizing bandwidth competition in the data stream scheduling strategy characterizes the problem, analyzes the input, output and constraint conditions of the problem and the optimization problem, and provides a solving algorithm according to the optimization problem.
The further technical scheme of the invention is as follows: in the input of the analysis problem, the flow of the Shuffle stage is calculated in a distributed manner to form a set consisting of a plurality of parallel streams, the completion time of the set is determined by the completion time of the last stream in the set, and the set is defined as a collaborative flow set and is expressed as follows:
F＝{f i }＝{(s i ，d i ，v i ) 1 ≦ i ≦ N, where f i Expressed as a co-currentOf the ith stream, s i The source node representing the flow, d i Denoted as destination node of the flow, v i Represented as a data volume for the flow.
The invention further adopts the technical scheme that: from the input co-flows, in the output of the analytical problem, a routing plan for each flow can be derived, denoted as P = { P = 1 ，p 2 ，...，p N Wherein P is P i ，p i Is flow f i The selected routing path.
The further technical scheme of the invention is as follows: introducing an integer binary variable X in constraint conditions of analysis problems for ensuring path uniqueness i,p In P represents P i One path of which is gathered if flow f i Routing plan p i Using path p, then X i,p The value is 1, otherwise 0, and the constraint condition expression is obtained as
The further technical scheme of the invention is as follows: in the optimization problem of analysis problem, the route plan of flow is established by minimizing bandwidth competition of link, all route plans are determined according to introduced binary variable, and according to p i Can determine
If e uv On the path p i In the above, then
The value is 1, otherwise, the value is 0, and the binary variable is used
Is shown as
According to a binary variable
Get new per streamT of completion time i The formula:
and establishing an optimization problem of route scheduling for minimizing bandwidth competition.
The further technical scheme of the invention is as follows: obtaining the optimization problem by a collaborative flow routing greedy scheduling algorithm in the process of solving the optimization problem, calculating shortest paths of all flows in the collaborative flow by using a Dijkstra algorithm according to an input satellite network topological graph G and a collaborative flow information vector F by the collaborative flow routing greedy scheduling algorithm, and initializing a routing plan P according to the shortest paths; for each flow f i Searching its set of possible paths P i Finding out the optimal path, and initially creating a temporary routing scheme P according to the current P * Continuously using P i Middling path P to replace P * Path of the corresponding flow in accordance with P * And calculating the completion time corresponding to the whole Shuffle flow, wherein the path used by the obtained minimum Shuffle completion time is the optimal solution of the current flow.
The invention has the beneficial effects that: a low-orbit satellite network model is obtained through simulation of giant constellation orbits, and by combining the advantages of OpenFlow flow tables and topology sensing and centralized control of an SDN controller, a flooding-free route establishing mechanism of a network topology is provided, and the forwarding efficiency of satellite network information can be improved. The method is oriented to a low-orbit satellite network, analyzes a mathematical model for distributed computation of optimized scheduling of data streams, namely, the completion time of minimized cooperative streams, obtains a corresponding optimization problem, then provides a cooperative flow routing greedy scheduling (CoRGS) algorithm capable of minimizing bandwidth competition, and verifies the superiority of the scheduling algorithm.
Drawings
FIG. 1 is a schematic diagram of a MapReduce calculation process in the prior art.
Fig. 2 is a schematic diagram of an SDN-based satellite distributed computing framework according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of link distances (km) of a local satellite network topology according to an embodiment of the present invention.
Fig. 4 is a schematic diagram of a flooding-free route establishment mechanism of a mesh topology according to an embodiment of the present invention.
Fig. 5 is a comparison diagram of performance of the same total flow load and different numbers of operating nodes according to the embodiment of the present invention.
Fig. 6 is a schematic diagram illustrating comparison of performances under different total traffic loads and the same number of working nodes provided by the embodiment of the present invention.
Fig. 7 is a schematic diagram illustrating comparison of performances of the same amount of streaming data and different numbers of working nodes according to the embodiment of the present invention.
Detailed Description
To effectively manage satellite nodes and satellite networks, large-scale low-earth orbit satellite (LEO) networks may introduce a new network paradigm Software Defined Network (SDN). In SDN, a controller of the control plane can centrally manage data streaming, and a switch of the data plane is only responsible for data forwarding. For a low-orbit constellation network, the introduction of the SDN management paradigm can effectively solve the problems of network management and control and resource scheduling, and efficiently utilize network resources. A specific implementation of SDN may use the OpenFlow protocol. The OpenFlow protocol forwards packets using a Flow Table (Flow Table) composed of Flow entries (Flow Entry) for matching and processing packets.
Diversified broadband information services and marine quantitative earth observation application can directly perform distributed calculation on data by using a satellite cluster, and the load pressure of a satellite-earth communication link can be effectively relieved.
In-orbit cooperative computing of a low-orbit satellite network requires frequent data exchange interaction among satellite nodes, and the performance of the satellite network directly influences the transmission of data streams; and the link bandwidth capacity between satellites is limited due to satellite loading. A large number of concurrent data streams generated in a short time like a cooperative stream are prone to cause a problem of inter-satellite link congestion. The cooperative flow optimization routing scheduling strategy facing the satellite network can reduce the inter-satellite link congestion problem and improve the data flow transmission efficiency. Therefore, on the premise of considering a satellite network model, the optimization target of the designed distributed computing data flow scheduling strategy is similar to that of the scheduling strategy in the data center on the ground, and the completion time of computing operation is reduced by shortening the completion time of the cooperative flow, so that the service quality of the service is improved.
The scheduling strategy we propose may be implemented in the SDN controller. However, the existing focus research on SDSN is directed to the top-level design of the logic architecture, but the research on the technical details of SDN applications is relatively lacking, such as the research on satellite network-aware scheduling policies. Therefore, we will design some technical implementation details of the satellite SDN before the data flow scheduling policy is proposed to support the application implementation of the proposed scheduling policy.
Firstly, a satellite distributed computing framework based on the SDN is proposed, and a software model implementation process of the satellite distributed computing framework is introduced. Then, the giant constellation orbit is subjected to visual simulation, and a model of the low-orbit satellite network can be obtained. For a large-scale low-orbit constellation, the networking form of the inter-satellite links enables the satellite network to present a grid-shaped multi-hop topology with a ring. In order to solve the problem of information forwarding of a traditional network in a ring topology, a flooding-free route establishment mechanism of the network topology based on SDN is provided, the information forwarding efficiency among satellites can be improved by using the mechanism, a specific data flow scheduling strategy can be deployed at an SDN control layer, and the on-orbit collaborative computing efficiency in a satellite cluster is improved.
In the low-orbit satellite network, a large amount of intermediate value exchanges are generated in the orbit distributed computing process, centralized concurrent data streams can be formed, and inter-satellite link congestion is easily caused. We define these data streams as cooperative streams. Suppose a cooperative flow F has N flows (flows) F in total i For each flow, the source node s is known i Destination node d i And amount of data v i . For a satellite topological graph, a node set U and a link set E are known; for inter-satellite link e uv With link bandwidth capacity of B uv The link distance is l uv . The completion time of each stream is
In the formula
To representFlow f i Path p of (1) i And e uv The relation variable between, c represents the wave propagation velocity. The completion time of the cooperative flow is
After some constraint conditions of the cooperative flow are analyzed, an optimization problem is deduced, and a greedy scheduling algorithm of the cooperative flow routing is proposed to solve the optimization problem. The scheduling algorithm is divided into two stages: and in the stage I, initial routing planning is performed, and in the stage II, the initial solution is gradually optimized by utilizing a local search process.
Since our solution is distributed computing oriented to low-earth orbit satellite networks, we first propose a SDN-based satellite distributed computing framework that can build a simulation with some software, as shown in fig. 2. The software simulation process comprises the steps of firstly simulating giant-constellation satellite orbits by satellite simulation Software (STK) to obtain a satellite network model, and then establishing a virtual network simulating the satellite network topology by utilizing Mininet according to the network model. Mininet is a process virtualization network simulation tool developed by stanford university based on Linux Container architecture, and can create a virtual network including a host, a switch, a controller and a link, wherein the switch supports OpenFlow and supports a highly flexible custom SDN. In a virtualized network established by Mininet, a satellite node includes a host node and an OpenFlow switch (OVS) node, the host node is used for simulating a server for satellite processing, and the OVS is used for simulating connection of inter-satellite links and realization of SDN. Then, we can configure our routing algorithm by using the SDN controller Ryu, and at the same time, SDN application functions such as topology awareness, traffic monitoring, routing scheduling, and the like can be implemented. Finally, a distributed computing framework Hadoop is configured in a host node (satellite node) in the network and some actual computing tasks are deployed, so that simulation of the satellite distributed computing data packet level is realized. Next, we will introduce the technical details of satellite network simulation, satellite SDN, and data flow scheduling policy, respectively.
The currently proposed LEO mega constellation designs are Telesat, amazon Kuiper, spaceX Starlink, and others. They have some common features: 1) The satellites are organized into a number of orbits; 2) Satellites in the same orbit are evenly distributed; 3) A plurality of tracks penetrate through the equator at uniform intervals to form a track shell; 4) One constellation may deploy one or more orbital shells. According to the characteristics, the corresponding constellation network can be simulated by using the STK, and the inter-satellite link distance of the network topology can be obtained by performing visual calculation through the generated orbit. If the microwave link is used by the inter-satellite link, the bandwidth capacity of the link can be calculated according to the link distance by using a free attenuation formula. Fig. 3 shows the link distances of the satellite network topology in the vicinity of 30 ° north latitude acquired after the first period of the simulation Starlink constellation. The longitudinal satellites are in the same orbit, and are uniformly distributed as the second point of the giant constellation, so that the inter-satellite distance is 659km; observing the inter-satellite distance of the transverse satellite, the regularity of the transverse satellite is found. Due to the dynamics and periodicity of the satellite network, the satellite network topology in other regions is similar. Moreover, the satellite operation period can be divided into a number of finite time intervals, within which the network can be assumed to be static; the regularity of the satellite motion is utilized to obtain a regular static network topology. These regional satellite network models are the basis for mathematical modeling of problems that we are studying the data flow scheduling strategy later.
By observing the obtained satellite network topological graph, the fact that in a large-scale low-orbit constellation network, the inter-satellite links connect satellites to form a grid topology can be known, and a large number of loops exist in the topology. In a physical layer flooding mechanism brought by an Address Resolution Protocol (ARP) in a conventional IP network, flooding in the topology of a ring network model may cause a network paralysis problem, and thus, a very expensive Spanning Tree Protocol (STP) is generally required to process the flooding. And with the SDN architecture, the route is directly calculated in the SDN controller through ARP analysis, and flooding-free point-to-point forwarding can be realized. The specific flow is shown in fig. 4.
Ephemeris and connection relations of the constellation are preloaded in the SDN controller. When an ARP packet arrives at an OpenFlow switch, first, a flow table of the switch is searched for whether there is a corresponding matching flow entry. If the source address and the destination address do not exist, the packet _ in message is sent to the SDN controller, the SDN controller analyzes the ARP packet to obtain the source address and the destination address, then the shortest path between the source address and the destination address is calculated according to the satellite connection relation at the current moment, and then the shortest path is written into the flow table. Through the continuous packet _ in process, the controller establishes a complete flow table to realize network communication. After the flow table is established, the SDN controller enters a flow scheduling stage, and the flow table is continuously updated by the SDN controller according to a specific data flow scheduling strategy and a network state.
Next we will propose a specific data flow scheduling strategy. The data flow scheduling strategy of the invention provides a corresponding routing scheduling algorithm capable of minimizing bandwidth competition aiming at the problem of inter-satellite link congestion caused by Shuffle flow in a distributed computing process in a low earth orbit satellite network. Firstly, the problem is characterized, the input, the output, the constraint condition and the optimization problem of the problem are analyzed, and then a solving algorithm of the optimization problem is given.
(1) Inputting:
and (4) Coflow: the distributed computation Shuffle stage flow is a set consisting of a plurality of parallel flows, the completion time of which is determined by the completion time of the last flow in the set, we define this set as a set of coordinated flows (Coflow), denoted as F = { F = { (F) i }＝{(s i ，d i ，v i )}，1≤i≤N(1)
In the formula f i The ith stream, s, denoted as the cooperative stream i Source node representing the flow, d i Denoted as destination node of the flow, v i Expressed as the volume of the stream, i.e., the total amount of data that the stream needs to transport. Here we assume that this information can be captured by upper layer applications or obtained with existing prediction techniques to support our subsequent work.
A network topology diagram: to obtain the network topology of a constellation of giant satellites, we take the space x Starlink model as an example, each satellite can have 4 four inter-satellite links to connect with 4 surrounding satellites, and we divide the constellation from the giant constellation to form a network topology similar to a checkerboard, so as to obtain our network topology, which is denoted as G = { U, E, L, B } (2)
Wherein U is a set of satellite nodes; e is the link set (E) uv Representing a link of node u with node v); l is the set of inter-node distances (L) uv Represents the distance between node u and node v); b is the available bandwidth capacity of the link in E (B) uv Represents a link e uv Available bandwidth capacity).
(2) And (3) outputting:
and (3) routing planning: for an incoming collaborative flow, a routing plan, representation, may be obtained for each flow therein
Is P = { P 1 ，p 2 ，...，p N } (3)
In the formula p i Formula flow f i The selected routing path. In addition, we use integer variables
To determine each path P in P i ，
There are three possible values (1, 0 and-1) representing f i At link e uv How it is forwarded. Specifically, when data is on link e uv From node u to v, then
Equal to 1; when data is on link e uv From node v to u, then
Is equal to-1; when data is not sent between nodes u and v, then
Equal to 0. We can express the above as
Shuffle completion time: i am concerned withLet t be used to represent the completion time of the Shuffle phase, whose value depends on the stream that is the slowest to transmit, and let us use t i To represent flow f i The transmission completion time of (c) is obviously given by the formula of t
(3) Constraint conditions are as follows:
there are many paths that can be taken for each flow in the graph, we use P i To express flow f i Set of possible paths in the graph, flow f i The optimal path is in the set, and we only need to search the set for its optimal path. To ensure path uniqueness, we introduce an integer variable X i,p It is a binary variable (0 or 1) in which P represents P i One path of which is gathered if flow f i Routing plan p i Using path p, then X i,p The value is 1, otherwise 0. From this, a constraint expression is derived
In addition, to guarantee the certainty of the path, we give a flow conservation constraint formula. We first introduce a new variable N (u) representing the set of neighbor nodes of the satellite node u in the network graph. The flow conservation constraint formula is as follows
Constraint equation (7) ensures that data is sent from the source node over only one link; constraint equation (8) ensures that data is sent to the destination node over only one link; constraint equation (9) ensures that any flow and any intermediate node u that it passes through, the number of transit links sent to the u node is equal to the number of transit links from the u node.
(4) Optimizing the problem:
in the process of transmitting the traffic on the link, the congestion of the link is usually caused by that the traffic in the link is too large, and bandwidth contention occurs, so the optimal idea can be to minimize the bandwidth contention of the link to make a routing plan of the flow. Before we use integer variables
To determine each path P in P i For convenience and clarity, we introduce another binary variable
The value is 0 or 1. When all routing plans are determined, flow f i On link e uv The direction of (A) is naturally determined, so that p i Can determine
If e uv On the path p i In the above, then
The value is 1, otherwise 0. We can express the above as
Is provided with
After that, we can give a new per-stream completion time t i Is expressed as
Wherein the left side of the plus sign represents the transmission time, flow f i Bandwidth competition may occur between each link and other flows, each flow is in fair competition on the link, and the time is limited by the link with the largest competition; the right of the plus sign in the formula represents the propagation time of the stream. Next, an optimization problem of routing scheduling that minimizes bandwidth contention can be established, which is defined as an optimization problem P:
optimizing the target: minimizet (P)
And (3) constraint:
to solve the optimization problem, we propose a collaborative flow Routing Greedy Scheduling (surgs) algorithm. The CoRGS algorithm pseudo code is shown in Table 1. The CoRGS algorithm has two main stages: in the first stage, obtaining an initial routing plan based on the shortest path; and in the second stage, based on the thought of a greedy algorithm, further optimizing the obtained initial solution by utilizing a local search process. The details of the algorithm are described below.
TABLE 1 collaborative flow routing greedy scheduling algorithm pseudocode
Stage one: and calculating the shortest paths of all the streams in the cooperative streams by using a Dijkstra algorithm according to the input satellite network topological graph G and the cooperative stream information vector F, and initializing a routing plan P according to the shortest paths. Since our optimization goal is to minimize the one with the largest completion time in the cooperative flow, we can optimize the routing path of the flow with the largest completion time based on the greedy idea, which is also the reason for ordering F.
And a second stage: for each flow f i We search its set of possible paths P i And find the optimal path therefrom. In detail, initially we create a temporary routing plan P according to the current P * Then continuously using P i Middling path P to replace P * Path of the corresponding flow in accordance with P * The completion time corresponding to the entire Shuffle stream is calculated. And the path used by the finally obtained minimum Shuffle completion time is the optimal solution of the current flow. In a word, a local search process is used for finding an optimal path of flow, namely a local optimal solution, and finally a global optimal solution is obtained.
Through simulation, performance comparison graphs of the proposed data flow scheduling algorithm and three comparative classical routing algorithms are obtained under different working nodes and different total traffic loads. The three comparison algorithms are Shortest Path with Minimum hop (SPMH) algorithm, shortest Path with Minimum Consumption (SPMC) algorithm, and Equal Cost Multiple Path (ECMP) algorithm, respectively. The network topology map of the experimental input is obtained from fig. 2. The cooperative flow input in the experiment is obtained by randomly generating n working nodes, the number of the corresponding flow is n x (n-1), 1000 samples are randomly generated in each group, the working nodes are randomly selected from the topological graph each time, and then the average value of the 1000 sample values is taken as the final result of the experiment. We performed three sets of experiments: 1) Fixing the total load (5 GBytes) of the Shuffle flow, and increasing the number of working nodes (the number of flows); 2) Fixing the number of the working nodes (5), and increasing the total load of the Shuffle flow; 3) The size of each flow data amount is fixed (128 MBytes), and the number of working nodes (the number of flows) is increased. Fig. 5 shows a performance comparison of the experimental group (1), fig. 6 shows a performance comparison of the experimental group (2), and fig. 7 shows a performance comparison of the experimental group (3).
It can be obtained that for different working nodes and different total traffic loads, the coordinated flow completion time indexes of the proposed scheduling algorithm CoRGS are all smaller than those of the other three routing algorithms. In addition, the performance optimization percentage of the algorithm is irrelevant to the data volume of the flow or the flow load, and is mainly influenced by the number of the working nodes, and when the number of the working nodes occupies more total nodes of the topological graph, the performance is gradually reduced. This is because the more flows in the network graph, the less paths that can be optimized, which can be solved by scaling up the network topology.
Under the framework of a software-defined satellite network, the scheduling algorithm can be deployed in an SDN controller, the data stream transmission efficiency of satellite cluster distributed computation can be effectively improved, and therefore the performance of on-orbit processing services is improved.
The method is characterized in that a topological model of a low earth orbit satellite network is analyzed, data flow in a distributed Shuffle calculation process is researched, not only annular topology in the satellite network is considered, but also link congestion generated by Shuffle flow in links among satellites is considered, and on a software-defined satellite network architecture, according to a mathematical model for minimizing the completion time of cooperative flow, a routing scheduling algorithm is provided:
a low-orbit satellite network model is obtained through simulation of a giant constellation orbit, advantages of an OpenFlow flow table and topology perception and centralized control of an SDN controller are combined, a flooding-free route establishing mechanism of a network topology is provided, and forwarding efficiency of satellite network information can be improved.
For a low-orbit satellite network, a mathematical model of distributed computation data flow optimization scheduling, namely the completion time of a minimized cooperative flow is analyzed, a corresponding optimization problem is obtained, then a cooperative flow routing greedy scheduling (CoRGS) algorithm capable of minimizing bandwidth competition is provided, and the superiority of the scheduling algorithm is verified.
The above description is intended to be illustrative of the preferred embodiment of the present invention and should not be taken as limiting the invention, but rather, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.
Claims (10)
1. The low-earth-orbit satellite network-oriented distributed computing data flow scheduling and routing method is characterized by comprising the following steps of:
s1, constructing a satellite distributed computing framework based on an SDN, simulating a giant constellation satellite orbit by using satellite simulation Software (STK) and acquiring a satellite network model;
s2, establishing a virtual network simulating the satellite network topology by utilizing Mininet according to the acquired satellite network model;
s3, configuring a routing algorithm by using the SDN controller;
and S4, configuring a distributed computing frame Hadoop in a host node (satellite node) in the satellite network and deploying an actual computing task to realize the simulation of the satellite distributed computing data packet level.
2. The low-earth-orbit satellite network-oriented distributed computing data flow scheduling and routing method according to claim 1, wherein the point-to-point forwarding without flooding is realized in the SDN controller by analyzing direct computing routes through APP.
3. The low-orbit satellite network-oriented distributed computing data flow scheduling and routing method of claim 2, wherein ephemeris and connection relations of constellations are preloaded in the SDN controller, when an ARP packet reaches an OpenFlow switch, whether a corresponding matched flow entry exists is searched in a flow table of the switch, and if not, a complete flow table is established through a continuous packet _ in process to realize network communication.
4. The low-earth-orbit-satellite-network-oriented distributed computing data flow scheduling and routing method of claim 3, wherein the packet _ in process is sent to an SDN controller through a packet _ in message, the SDN controller analyzes an ARP packet to obtain a source address and a destination address of the ARP packet, then calculates a shortest path between the source address and the destination address according to a satellite connection relationship at the current moment, writes the shortest path into a flow table, repeats the process continuously until a complete flow table is established, enters a flow scheduling stage after the complete flow table is established, and continuously updates the flow table according to a specific data flow scheduling policy and a network state.
5. The low-earth-orbit satellite network-oriented distributed computing data flow scheduling and routing method of claim 4, wherein the routing scheduling algorithm that minimizes bandwidth contention in the data flow scheduling policy characterizes the problem, analyzes the input, output, constraint conditions and optimization problem of the problem, and provides a solving algorithm according to the optimization problem.
6. The low-earth orbit satellite network-oriented distributed computing data flow scheduling and routing method according to claim 5, wherein in the input of the analysis problem, the distributed computing Shuffle phase traffic is a set composed of a plurality of parallel flows, the completion time of which is determined by the completion time of the last flow in the set, and the set is defined as a set of cooperative flows, and is expressed as: f = { F i }＝{(s i ，d i ，v i ) 1 ≦ i ≦ N, where f i The ith stream, s, denoted as the cooperative stream i Source node representing the flow, d i Denoted as destination node of the flow, v i Represented as a data volume for the flow.
7. The method of claim 6, wherein the routing plan of each flow is obtained from the inputted collaborative flows in the output of the analysis problem, and is expressed as P = { P = } 1 ，p 2 ，...，p N Wherein P is P i ，p i Is flow f i The selected routing path.
8. The low-earth-orbit-satellite-network-oriented distributed computing data flow scheduling and routing method as claimed in claim 7, wherein an integer binary variable X is introduced for ensuring path uniqueness in constraint conditions of analysis problems i,p In P represents P i One path of which is gathered if flow f i Routing plan p i Using path p, then X i,p The value is 1, otherwise 0, and the constraint condition expression is obtained as
9. The low-earth-orbit-satellite-network-oriented distributed computing data flow scheduling and routing method of claim 8, wherein the flow routing plan is formulated by minimizing bandwidth competition of links in an optimization problem of an analysis problem, all routing plans are determined by introducing binary variables, and p is the basis of p i Can determine
If e uv On the path p i In the above, then
The value is 1, otherwise, the value is 0, and the variable is binary
Is shown as
According to a binary variable
Get new per flow when doneT of (2) i The formula:
and establishing an optimization problem of route scheduling for minimizing bandwidth competition.
10. The low-earth-orbit satellite network-oriented distributed computing data flow scheduling and routing method as claimed in claim 9, wherein the optimization problem is solved by obtaining the collaborative flow path routing greedy scheduling algorithm, the collaborative flow path routing greedy scheduling algorithm calculates the shortest paths of all flows in the collaborative flow by using Dijkstra algorithm according to the input satellite network topological graph G and the collaborative flow information vector F, and initializes the routing plan P according to the shortest paths; for each flow f i Searching its set of possible paths P i Finding out the optimal path, and initially creating a temporary routing scheme P according to the current P * Continuously using P i Middling path P to replace P * Path of the corresponding flow in accordance with P * And calculating the completion time corresponding to the whole Shuffle flow, wherein the path used by the obtained minimum Shuffle completion time is the optimal solution of the current flow.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210793602.1A CN115426025B (en) | 2022-07-06 | 2022-07-06 | Distributed computing data flow scheduling and routing method for low orbit satellite network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210793602.1A CN115426025B (en) | 2022-07-06 | 2022-07-06 | Distributed computing data flow scheduling and routing method for low orbit satellite network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115426025A true CN115426025A (en) | 2022-12-02 |
| CN115426025B CN115426025B (en) | 2023-10-20 |
Family
ID=84196083
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210793602.1A Active CN115426025B (en) | 2022-07-06 | 2022-07-06 | Distributed computing data flow scheduling and routing method for low orbit satellite network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115426025B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6480495B1 (en) * | 1997-04-18 | 2002-11-12 | Nortel Networks Limited | Connectionless communications network |
| CN106656302A (en) * | 2016-09-22 | 2017-05-10 | 南京理工大学 | Distributed node self-adaptive routing algorithm for LEO satellite network |
| CN108540211A (en) * | 2018-04-23 | 2018-09-14 | 大连大学 | A kind of satellite network framework based on SDN Yu ICN technologies |
| CN110034817A (en) * | 2019-04-29 | 2019-07-19 | 北京邮电大学 | Low-orbit satellite network route method and device based on software defined network |
| CN113099506A (en) * | 2021-03-31 | 2021-07-09 | 上海交通大学 | Routing method and system based on link aging in software defined satellite network |
| CN113852409A (en) * | 2021-09-27 | 2021-12-28 | 中国电子科技集团公司第五十四研究所 | Inter-satellite label routing method suitable for low-earth-orbit satellite network |
| CN114422423A (en) * | 2021-12-24 | 2022-04-29 | 大连大学 | Satellite network multi-constraint routing method based on SDN and NDN |
-
2022
- 2022-07-06 CN CN202210793602.1A patent/CN115426025B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6480495B1 (en) * | 1997-04-18 | 2002-11-12 | Nortel Networks Limited | Connectionless communications network |
| CN106656302A (en) * | 2016-09-22 | 2017-05-10 | 南京理工大学 | Distributed node self-adaptive routing algorithm for LEO satellite network |
| CN108540211A (en) * | 2018-04-23 | 2018-09-14 | 大连大学 | A kind of satellite network framework based on SDN Yu ICN technologies |
| CN110034817A (en) * | 2019-04-29 | 2019-07-19 | 北京邮电大学 | Low-orbit satellite network route method and device based on software defined network |
| CN113099506A (en) * | 2021-03-31 | 2021-07-09 | 上海交通大学 | Routing method and system based on link aging in software defined satellite network |
| CN113852409A (en) * | 2021-09-27 | 2021-12-28 | 中国电子科技集团公司第五十四研究所 | Inter-satellite label routing method suitable for low-earth-orbit satellite network |
| CN114422423A (en) * | 2021-12-24 | 2022-04-29 | 大连大学 | Satellite network multi-constraint routing method based on SDN and NDN |
Non-Patent Citations (4)
| Title |
|---|
| JIAN LI等: "Temporal Netgrid Model-Based Dynamic Routing in Large-Scale Small Satellite Networks", 《IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY》, vol. 68, no. 6, pages 6009 - 6021, XP011730494, DOI: 10.1109/TVT.2019.2910570 * |
| LINLIN ZHANG等: "A Routing Scheme for Software-Defined Satellite Network", 《2017 IEEE INTERNATIONAL SYMPOSIUM ON PARALLEL AND DISTRIBUTED PROCESSING WITH APPLICATIONS AND 2017 IEEE INTERNATIONAL CONFERENCE ON UBIQUITOUS COMPUTING AND COMMUNICATIONS (ISPA/IUCC)》, pages 24 - 31 * |
| 王丽冲等: "一种基于OpenFlow的软件定义卫星网络架构设计方案", 《电子设计工程》, vol. 24, no. 17, pages 85 - 89 * |
| 费长江等: "兼顾控制流量的软件定义卫星网络路由策略", 《北京航空航天大学学报》, vol. 44, no. 12, pages 2575 - 2585 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115426025B (en) | 2023-10-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chen et al. | Hierarchical domain-based multicontroller deployment strategy in SDN-enabled space–air–ground integrated network | |
| Wang et al. | An adaptive routing algorithm for integrated information networks | |
| CN109714219B (en) | Virtual network function rapid mapping method based on satellite network | |
| CN111970044B (en) | Satellite network time slot allocation and routing planning method based on Lagrange relaxation | |
| Gao et al. | Virtual network function placement in satellite edge computing with a potential game approach | |
| Lai et al. | Starperf: Characterizing network performance for emerging mega-constellations | |
| CN113315569B (en) | Satellite reliability routing method and system with weighted link survival time | |
| CN109005108B (en) | Space network route planning method and system | |
| CN109586785B (en) | Low-orbit satellite network routing strategy based on K shortest path algorithm | |
| Huang et al. | Reinforcement learning based dynamic distributed routing scheme for mega LEO satellite networks | |
| Liu et al. | Routing for predictable multi-layered satellite networks | |
| Secinti et al. | FOCUS: Fog computing in UAS software-defined mesh networks | |
| Qi et al. | SDN-based dynamic multi-path routing strategy for satellite networks | |
| Chen et al. | Optimal gateway placement for minimizing intersatellite link usage in LEO megaconstellation networks | |
| CN111884703A (en) | Service request distribution method based on cooperative computing among communication satellites | |
| Gao et al. | Dynamic resource management for neighbor-based VNF placement in decentralized satellite networks | |
| Zhu et al. | Efficient hybrid multicast approach in wireless data center network | |
| CN110290069A (en) | A kind of reliable multicast method for routing under SDN scene | |
| CN115426025B (en) | Distributed computing data flow scheduling and routing method for low orbit satellite network | |
| Han et al. | Space edge cloud enabling service migration for on-orbit service | |
| Zhang et al. | Research on routing control with delay constraint based on contact plan for integrated satellite terrestrial network | |
| Li et al. | LEO Mega-Constellations Routing Algorithm Based on Area Segmentation | |
| CN115021793B (en) | Network coding-based satellite network inter-satellite link planning and power distribution method | |
| CN114024894B (en) | Dynamic calculation method and system in software-defined heaven-earth integrated network | |
| Wang et al. | Optimal download of dynamically generated data by using ISL offloading in LEO networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |