CN114826937A - Flow sensing method based on size time scale under heaven-earth fusion network - Google Patents

Abstract

The invention discloses a flow sensing method based on size and time scale under a world fusion network, which aims at the world fusion network environment and provides a world fusion architecture which can sense network resource nodes in a collection region and process flow information, can adapt to the characteristics of the world fusion network to predict the regional flow under the large time scale and can judge whether the resources in the region can meet the resource requirement corresponding to the predicted flow. By means of delay-sensitive cross-node traffic transfer on a small time scale, resource node overload conditions caused by traffic burst on the small time scale can be reduced, and resource utilization rate of idle nodes can be improved by performing queue caching and traffic transfer on delay-tolerant traffic.

Description

A traffic sensing method based on large and small time scales under the fusion network of heaven and earth

technical field

The invention belongs to the field of sky-earth fusion information network, and in particular relates to a flow perception method based on large and small time scales under the sky-earth fusion network.

Background technique

In recent years, as an emerging network architecture that integrates satellite and terrestrial communication systems, the Sky-Earth Convergence Network has received extensive attention from the academic and industrial circles. an important part of survivability. Thanks to its advantages of wide coverage, high throughput, and strong robustness, the integrated space-earth network can be applied to many practical fields such as earth observation and mapping, intelligent transportation systems, military missions, homeland security, and disaster rescue. Satellites with high throughput can provide seamless wireless access services globally, and densely deployed terrestrial network infrastructure supports high-speed data access. The integration of space-based and ground-based networks can bring immeasurable benefits to the future 6G wireless communication system and provide more applications and services.

For the world-ground fusion network, due to its own heterogeneity, self-organization and dynamic characteristics, the fusion network brings significant benefits to various services and applications, but also faces routing selection, resource allocation, power control, terminal end-to-end service quality requirements and many other challenges. The integration of space-based and ground-based networks will lead to an increase in the types and quantities of services, while traffic sensing in the past is carried out on a large time scale, such as tens of minutes, to predict network traffic in an area on a large time scale. Perception, through the prediction results to schedule resources to meet the resource requirements corresponding to the traffic, however, this traffic prediction perception only solves the traffic changes on a large time scale, and cannot cope with the fine-grained small time scale in a large number of business scenarios in the world-earth fusion network. Traffic changes, such as seconds or hundreds of milliseconds, are sudden on a small time scale, which will cause some nodes in the area to be overloaded, thereby affecting the quality of service in the entire area.

At present, the research on the traffic perception of the sky-earth fusion network is basically carried out on a large time scale, and the traffic changes on a small time scale are rarely considered. The previous traffic prediction algorithms are only for a single network, either satellite network or ground. There are few studies on the traffic forecasting of the sky-earth fusion network. There are two general types of traffic forecasting, namely linear and nonlinear. The linear one needs to manually set parameters to fit the data based on experience, and the application range is small. The actual network traffic is nonlinear, periodic, bursty, and self-similar. Therefore, the application range of nonlinear traffic forecasting is large. However, because the traditional traffic forecasting is carried out in a single network, the characteristics of space-based network and ground-based network are not considered jointly, so it is not suitable for traffic in the space-ground fusion network. predict.

Traditional traffic sensing processing is only for large time scales, and rarely considers traffic changes on small time scales, and it is generally difficult to predict such traffic changes. Therefore, traffic changes on small time scales are often ignored. Such sudden changes in traffic will lead to different results for nodes at different times. When the sudden change of traffic is at the peak, it will overload and cause congestion, which will cause the node to generate a large service delay, reduce the quality of service of the node, or the sudden change of traffic is at the bottom. When the node is idle, it will cause the waste of node resources. Therefore, how to solve the sudden change of traffic on a small time scale is very important for the balance of node traffic in the region and the rational utilization of resources.

SUMMARY OF THE INVENTION

The purpose of the present invention is to overcome the above deficiencies, and to provide a flow perception method based on large and small time scales in the world-earth fusion network. The Markov-based flow forecasting method under the scale, considering the characteristics of space-based nodes and ground-based nodes, predicts the flow in the area on a large time scale to judge whether the resources in the area can meet the flow demand, and then In a small time scale, according to the burstiness of the traffic and considering the characteristics of the traffic, the delay-sensitive type is transferred, and the delay-tolerant type is put into the queue buffer and traffic is transferred.

In order to achieve the above object, the present invention comprises the following steps:

S1, model the sky-earth fusion network to obtain a resource pool of satellite ground nodes;

S2, predict the arrival of traffic in the area;

S3, according to the prediction result, determine whether the satellite ground resources in the large time scale meet the resource demand of the flow in the area;

S4, determine the precondition parameters for small time scale flow transfer;

S5, transfer traffic according to traffic size, type and resource status;

S6, according to the result after the traffic transfer, the traffic balance of the node and the resource status evaluation are performed.

The specific method of S1 is as follows:

The satellite and ground nodes are virtualized to shield the dynamics of the satellite nodes. In the time slice, it is regarded as a geographical area covered by a set of immutable virtual satellite nodes and ground nodes, thus obtaining a resource pool of satellite and ground nodes.

In S2, the specific method for predicting the arrival of traffic in the area is as follows:

According to the historical data of satellite ground flow in the area and the periodicity of the flow, the prediction on a large time scale is carried out. The frequency transition matrix is obtained according to the state set and historical flow data, and the state transition matrix is obtained according to the frequency transition matrix, and then the Markov transition chain is established. , by analyzing the state transition to predict the traffic at the next moment.

In S3, according to the prediction result, the method for judging whether the satellite ground resources in the large time scale meet the resource demand of the flow in the area is as follows:

A _long =C _long ,S _long ,T _long }

Among them, A _long is the resource requirement corresponding to the flow H ^′ on a large time scale, which needs to be satisfied:

为大时间尺度内的总资源量，

为已使用的资源量。in,

is the total amount of resources in a large time scale,

is the amount of resources used.

In S4, the precondition parameters for small time scale flow transfer are as follows:

N ₀ =βN

Among them, N ₀ is the node that the containment node can accept traffic transfer, N is the total number of nodes in the area, and β is the number used to adjust the containment node.

In S5, traffic transfer includes delay-sensitive DS and delay-tolerant DT;

In the case of the same traffic, the node first meets the resource requirements of the delay-sensitive DS, the remaining resources change with the dynamic resource requirements of the delay-sensitive DS, and the resource requirements of the delay-tolerant DT are dynamically supplied by the remaining resources.

The traffic transfer of the delay-sensitive DS is determined according to the threshold T _{i ,} which is determined by the resource amount of the node:

T _i =γ·RES _t ,

Among them, F _i is the current flow of node i on a small time scale, and RES _t is the ground node resource pool;

When the traffic exceeds the threshold, the traffic will be transferred. The transferred traffic is F _ij , and j∈N ₀ is the accommodating node. After the transfer, the traffic of node j is:

Among them, ni=N-No, F _j is the original flow of the receiving node;

表示时隙t时节点i对种类为k的延迟容忍型DT分配的资源，未完成的延迟容忍型DT换存在队列Q(t)中，由于延迟容忍型DT也有时延要求，应控制队列的长度：For the delay-tolerant DT is related to the resource amount of the current node and the delay-sensitive DS, different types of delay-tolerant DT have different resource requirements, denoted as

Represents the resources allocated by node i to the delay-tolerant DT of type k at time slot t, and the unfinished delay-tolerant DT is exchanged in the queue Q(t). length:

为处理时隙t时队列中的已有流量，

为时隙t时到来的种类为k的延迟容忍型DT流量大小；Among them, s _i is the processing speed of node i 0≤s _i ≤1,

To process the existing traffic in the queue at time slot t,

is the delay-tolerant DT traffic size of type k that arrives at time slot t;

When the average processing speed of the node is less than the average traffic arrival rate, the traffic is transferred, and the traffic of node j after the transfer is:

i,j∈N为具有接受流量转移能力的节点。Among them, the transferred traffic is

i,j∈N is a node with the ability to accept traffic transfer.

Through the traffic transfer of the delay-sensitive DS and the queue cache and traffic transfer of the delay-tolerant DT, the number of traffic overloads of nodes in the area can be reduced, the resource utilization of idle nodes can be improved, and the overload ratio of nodes in the area can be reduced. The formula is as follows:

Among them, R _over is the overload ratio of nodes in the area, and N _over is the number of overloaded nodes;

For delay-tolerant DT to ensure the stability of the queue, the following formula is used:

Q _i (t)≤Q _T

Among them, Q _T is the threshold of the queue length, which together with the node processing speed and traffic arrival speed determines whether DT performs cross-node traffic transfer.

Compared with the prior art, the sky-earth fusion architecture proposed by the present invention for the network environment of sky-earth fusion can sense and collect network resource nodes in the area and process the flow information, and can adapt to the characteristics of the sky-earth fusion network and control the flow in the area on a large time scale. By making predictions, it can be judged whether the resources in the area can meet the resource requirements corresponding to the predicted traffic. Delay-sensitive cross-node traffic transfer on small time scales can reduce resource node overload caused by traffic bursts on small time scales, and queue buffering and traffic transfer for delay-tolerant traffic can improve the efficiency of idle traffic. The resource utilization of the node.

Description of drawings

1 is a network architecture diagram of the present invention;

Fig. 2 is the flow forecast flow chart in the area under the large time scale of the present invention;

Fig. 3 is the flow chart of small and medium time scale flow transfer of the present invention;

4 is a schematic diagram of peak shaving and valley filling of flow transfer in the present invention;

Fig. 5 is a node overload ratio and resource utilization diagram before and after traffic transfer in the present invention;

Detailed ways

The present invention will be further described below with reference to the accompanying drawings.

The SDN-based world-earth fusion network architecture of the present invention is shown in FIG. 1. The traffic processing module in this architecture performs traffic prediction on a large time scale. The flow of traffic prediction is shown in FIG. 2, and the flow of cross-node traffic transfer on a small time scale is shown in FIG. 3. , and finally obtain the traffic balance state of the nodes in the area after the traffic transfer. The peak shaving and valley filling of the traffic transfer is shown in Fig.

The present invention specifically includes the following steps:

Step 1: Build an integrated architecture of heaven and earth, and obtain a virtualized multi-dimensional resource pool. The constructed integrated network architecture of heaven and earth includes a control layer, a traffic processing layer, and a resource layer, as shown in Figure 1, where:

The control layer is mainly based on the SDN controller to obtain network information and control and manage the network.

The traffic processing layer is mainly responsible for large time scale traffic prediction and small time scale traffic transfer.

The resource layer virtualizes space-based and ground-based nodes, abstracts multi-dimensional resources, shields the dynamics of satellite nodes, and obtains a resource pool of satellite ground nodes.

In the time slice, it can be seen that the geographical area is covered by a set of immutable virtual satellite nodes and ground nodes, and the obtained satellite and ground node resource pool RES _t ={C _t ,S _t ,T _t },0<t≤Th , where C _t is the computing resource, S _t is the storage resource, T _t is the transmission resource, and the resource has a survival time, and 0<t≤Th indicates that the node is in the geographical area.

则认为预测结果可信，否则使用历史数据平均值

Step 2: Predict the traffic at the next moment. According to the periodic characteristics of the traffic in the sky-earth fusion network, predict the arrival of traffic in the area, and carry out large-scale time according to the historical data of satellite ground traffic in the area and the periodic characteristics of traffic. The prediction under the scale is shown in Figure 2. The historical flow data is H, and the corresponding state set S={1,2...s} is divided, and the frequency transition matrix f _ij is obtained according to the state set and historical flow data, and then the state is obtained. The transition matrix P(s)=p _ij (s), p _ij (s)=P{x _s+1 =j|x _s =i} represents the probability that time s is in state i after one step is in j, where p _ij (s)≥0, ∑ _j∈s p _ij (s)=1, and then establish a Markov transition chain {x _n ,n=1,2,...k}, which is predicted by analyzing the state transition The flow H ^′ at the next moment, if

The prediction result is considered credible, otherwise the average value of historical data is used

Step 3: Determine whether the resource demand is met according to the prediction result, determine whether the satellite ground resources in the large time scale meet the resource demand of the flow in the area, and determine whether the existing resource amount meets the resource demand corresponding to the flow according to the flow prediction result, as follows:

A _long ={C _long ,S _long ,T _long }

Among them, A _long is the resource requirement corresponding to the flow H ^′ on a large time scale, which needs to be met

为大时间尺度内的总资源量，

为已使用的资源量。in,

is the total amount of resources in a large time scale,

is the amount of resources used.

Step 4: Determine the precondition parameters for small time scale traffic transfer. Considering the complexity of small time scale traffic transfer, it is necessary to set some nodes to have the ability to accept traffic transfer. The parameter N ₀ =βN, where N ₀ is The node that can accept traffic transfer by the accommodating node, N is the total number of nodes in the area, and the parameter β is used to adjust the number of accommodating nodes. For delay-tolerant traffic, because this type can tolerate a certain degree of delay, the adjacent idle node is selected as the accommodating node. ;

Step 5: The traffic transfer is implemented, and the traffic is transferred according to the traffic size, type and resource status. As shown in Figure 3, the cross-node traffic transfer on a small time scale considers two different scenarios: delay-sensitive DS and delay-tolerant DT. In the case of the same traffic, a small service delay will lead to a large resource demand, so the node should ensure a low level of overload, so the resource demand of the delay-sensitive DS should be satisfied first, and the remaining resources will increase with the delay-sensitive DS. The dynamic resource requirements of the DS vary, and the resource requirements of the delay-tolerant DT are dynamically supplied by the remaining resources.

The traffic transfer to the delay-sensitive DS is determined according to a certain threshold T _{i ,} which is determined by the resource amount of the node.

T _i =γ·RES _t ,

Among them, F _i is the current flow of node i on a small time scale,

When the traffic exceeds the threshold, the traffic will be transferred. The transferred traffic is F _ij , and j∈N ₀ is the receiving node. After the transfer, the traffic of node j is

Among them, ni=N-No, F _j is the original flow of the receiving node;

表示时隙t时节点i对种类为k的延迟容忍型DT分配的资源，未完成的延迟容忍型DT可以换存在队列Q(t)中，由于延迟容忍型DT也有一定的时延要求，应控制队列的长度：For the delay-tolerant DT is related to the resource amount of the current node and the delay-sensitive DS, different types of delay-tolerant DT have different resource requirements, denoted as

Represents the resources allocated by node i to the delay-tolerant DT of type k at time slot t. The incomplete delay-tolerant DT can be replaced in the queue Q(t). Since the delay-tolerant DT also has certain delay requirements, it should be Control the length of the queue:

为处理时隙t时队列中的已有流量,

为时隙t时到来的种类为k的延迟容忍型DT流量大小,Among them, s _i is the processing speed of node i 0≤s _i ≤1,

In order to process the existing traffic in the queue at time slot t,

is the delay-tolerant DT traffic size of type k that arrives at time slot t,

When the average processing speed of the node is less than the average traffic arrival rate, the queue will be too long and unstable to meet the delay requirement, and the traffic will be transferred. After the transfer, the traffic of node j is:

i,j∈N为具有接受流量转移能力的节点。Among them, the transferred traffic is

i,j∈N is a node with the ability to accept traffic transfer.

Step 6: Evaluate the effect of the traffic transfer, and perform the node traffic balance and resource status evaluation according to the result of the traffic transfer. Through the traffic transfer of the delay-sensitive DS and the queue buffering and traffic transfer of the delay-tolerant DT, the reduction can be achieved. The number of traffic overloads of nodes in the area, improve the resource utilization of idle nodes, and reduce the overload ratio of nodes in the area, as follows:

subjecttoN(1-β)≤ni<N,

i,j=1,...,N;

Among them, R _over is the overload ratio of nodes in the area, N _over is the number of overloaded nodes,

For delay-tolerant DT to ensure the stability of the queue, the following formula

Q _i (t)≤Q _T ,

i,j=1,...,N;

Among them, Q _T is the threshold of the queue length, which together with the node processing speed and traffic arrival speed determines whether the delay tolerance DT performs cross-node traffic transfer.

Figure 5 depicts the node overload ratio and resource utilization diagram before and after the traffic transfer. It can be seen that before the traffic is not transferred, the node's overload ratio is high and the resource utilization is very low. After the traffic transfer, the node overload ratio is greatly reduced to 8.4 %, and the resource utilization rate is increased to 92.1%, indicating that the present invention can effectively reduce the overload situation of resource nodes caused by the burst of traffic on a small time scale, and also improve the resource utilization rate.

Claims (8)

1. A flow sensing method based on size time scale under a world fusion network is characterized by comprising the following steps:

s1, modeling the heaven and earth fusion network to obtain a satellite ground node resource pool;

s2, predicting arrival of the regional flow rate;

s3, judging whether the satellite ground resources meet the resource requirements of the flow in the region in a large time scale according to the prediction result;

s4, determining precondition parameters for small-time scale flow transfer;

s5, carrying out traffic transfer according to the traffic size, type and resource state;

and S6, carrying out traffic balance and resource state evaluation of the nodes according to the result after traffic transfer.

2. The traffic sensing method based on size time scale under the heaven-earth converged network according to claim 1, wherein the specific method of S1 is as follows:

the method comprises the steps of virtualizing a satellite and ground nodes, shielding the dynamic property of the satellite nodes, and covering a group of unchanged virtual satellite nodes and ground nodes in a geographic region within a time slice to obtain a satellite and ground node resource pool.

3. The method for sensing traffic based on size and time scale in the world-wide converged network according to claim 1, wherein in S2, the specific method for predicting the arrival of regional traffic is as follows:

according to historical satellite ground flow data and periodicity of flow in the region, prediction under a large time scale is carried out, a frequency transfer matrix is obtained according to a state set and the historical flow data, a state transfer matrix is obtained according to the frequency transfer matrix, a Markov transfer chain is further established, and the flow at the next moment is predicted by analyzing the state transfer.

4. The method for sensing traffic based on the size and time scale in the space-ground converged network according to claim 1, wherein in S3, the method for determining whether the satellite ground resources in the large time scale meet the resource demand of the traffic in the region according to the prediction result is as follows:

A _long ＝{C _long ，S _long ，T _long }

wherein ,A_long The resource requirements corresponding to the flow H' under the large time scale need to be met:

wherein ,

is the total amount of resources in a large time scale,

is the amount of resources used.

5. The method for sensing traffic based on the size-time scale in the space-ground converged network according to claim 1, wherein the precondition parameters for the small-time scale traffic migration in S4 are as follows:

N ₀ ＝βN

wherein ,N₀ For the nodes that can accept traffic transfer, N is the total number of nodes in the area, and β is the number used to adjust the number of the receiving nodes.

6. The method for sensing traffic based on size-time scale in a space-ground converged network of claim 1, wherein in S5, traffic shifting comprises delay-sensitive DS and delay-tolerant DT;

under the condition of the same traffic, the node firstly meets the resource requirement of the delay-sensitive DS, the rest of the resources change along with the dynamic resource requirement of the delay-sensitive DS, and the resource requirement of the delay-tolerant DT is dynamically supplied by the rest of the resources.

7. The method of claim 6, wherein the delay-sensitive DS traffic diversion is based on a threshold T _i To determine the threshold value T _i Determined by the amount of resources of the node:

T _i ＝γ·RES _t ，

wherein ,F_i Is the current flow, RES, of node i at a small time scale _t A ground node resource pool;

when the flow exceeds the threshold value, the flow is transferred, and the transferred flow is F _ij ，j∈N ₀ To accommodate a node, the traffic at node j after the transition is:

wherein ni is N-No, F _j The original flow of the node is accommodated;

for the delay tolerant type DT related to the resource amount of the current node and the delay sensitive type DS, different kinds of delay tolerant type DT having different resource amount requirement are marked as

Indicating the resources allocated by node i to the delay tolerant type DT of type k at time slot t,the outstanding delay tolerant DT is stored in the queue Q (t), and the length of the queue should be controlled because the delay tolerant DT also has delay requirement:

wherein ,s_i Is the processing speed of the node i is not less than 0 and not more than s _i ≤1，

To handle the existing traffic in the queue at time slot t,

the delay tolerant type DT flow size with the type of k coming at the time slot t;

and when the average processing speed of the node is less than the average flow arrival rate, carrying out flow transfer, wherein the flow of the transferred node j is as follows:

wherein the diverted flow rate is

Is a node with the capability of accepting traffic transfer.

8. The traffic sensing method based on the size and time scale in the space-ground converged network according to claim 6, wherein the traffic overload number of the nodes in the area is reduced, the resource utilization of the idle nodes is improved, and the overload ratio of the nodes in the area is reduced by the traffic transfer of the delay sensitive DS and the queue caching and traffic transfer of the delay tolerant DT, according to the following formula:

wherein ,R_over For node overload ratio in the area, N _over The number of overloaded nodes;

the stability of the queue to be guaranteed for the delay tolerant type DT has the following formula:

Q _i (t)≤Q _T

wherein ,Q_T The threshold value of the queue length, together with the node processing speed and the traffic arrival speed, determines whether the DT transfers traffic across the nodes.

Images