CN-122027568-A - Tunnel traffic dynamic scheduling system oriented to small particle slicing network

CN122027568ACN 122027568 ACN122027568 ACN 122027568ACN-122027568-A

Abstract

The invention discloses a tunnel traffic dynamic scheduling system oriented to a small particle slice network, and relates to the field of network slicing and traffic engineering. The invention aims to solve the problems that the utilization rate of small particle slice resources is low and the service experience is difficult to guarantee by passive response type scheduling in the prior art. The system comprises a small particle slice perception layer, a dynamic policy engine and a tunnel execution layer. The perception layer collects queue depth and time delay data of the tunnel in real time, the strategy engine pre-judges congestion based on a prediction model and identifies idle slices to initiate resource lending, and the execution layer dynamically adjusts paths and queue parameters to complete lending. The invention realizes active elastic scheduling based on prediction, remarkably improves the utilization rate of network resources and ensures the performance of high-priority service.

Inventors

ZHENG XUHUA
XU GANG

Assignees

杭州华思通信技术有限公司

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. The tunnel traffic dynamic scheduling system for the small particle slice network is characterized by comprising a small particle slice sensing layer, a dynamic strategy engine and a tunnel execution layer: The small particle slice sensing layer is used for being configured to collect running state data of a plurality of small particle slice tunnels in a network in real time, wherein the running state data comprises a queue depth change rate, a bidirectional delay difference and a current bandwidth utilization rate of each small particle slice tunnel; The dynamic policy engine is configured to be communicatively connected to the small particle slice aware layer, and configured to: based on the running state data, predicting the congestion probability of each small particle slice tunnel in a future preset time window by using a prediction model; When the congestion probability of the first small particle slice tunnel is predicted to exceed a first threshold value, identifying a second small particle slice tunnel with idle bandwidth resources currently existing; Generating a resource lending scheduling strategy, wherein the resource lending scheduling strategy indicates the first small particle slice tunnel to temporarily borrow part of bandwidth resources of the second small particle slice tunnel, and sets lending duration and return triggering conditions; The tunnel execution layer is configured to be deployed in a network node, is in communication connection with the dynamic policy engine, and is configured to analyze and execute the resource lending scheduling policy, dynamically adjust the flow forwarding paths and queue scheduling parameters of the first small particle slice tunnel and the second small particle slice tunnel, and complete temporary lending of bandwidth resources.
2. The small particle slice network oriented tunnel traffic dynamic scheduling system of claim 1, wherein the small particle slice aware layer is further configured to: collecting the running state data in a millisecond period, marking the collected data with a time stamp and a slice mark, and reporting the data to the dynamic policy engine in real time; The running state data also comprises packet loss rate, jitter and one-way time delay of each small particle slice tunnel.
3. The small particle slicing network-oriented tunnel traffic dynamic scheduling system according to claim 1, wherein the prediction model adopted in the dynamic policy engine is a cyclic neural network-based time sequence prediction model, the dynamic policy engine is configured to take a queue depth change rate, a bidirectional delay difference and a bandwidth utilization rate of a plurality of historical acquisition periods as input features, output a predicted value of congestion probability in a future preset time window through the prediction model, and the preset time window is 100 milliseconds to 1 second.
4. The system for dynamically scheduling tunnel traffic for a small particle slice network according to claim 1, wherein the dynamic policy engine identifies a second small particle slice tunnel in which idle bandwidth resources currently exist, specifically comprising: The dynamic policy engine is configured to calculate a difference value between a bandwidth utilization rate of each candidate small particle slice tunnel and a preset idle threshold value, and determine a candidate tunnel which has a difference value greater than zero and has the same service level protocol as the first small particle slice tunnel and belongs to the same resource pool as the second small particle slice tunnel.
5. The system of claim 1, wherein the dynamic policy engine, when generating the resource lending scheduling policy, is further configured to: And calculating a lending bandwidth limit according to the difference value between the predicted congestion probability value of the first small particle slice tunnel and the first threshold value and the idle bandwidth size of the second small particle slice tunnel, wherein the lending bandwidth limit is positively related to the difference value and does not exceed the idle bandwidth size.
6. The small particle slice network oriented tunnel traffic dynamic scheduling system of claim 1, wherein the tunnel execution layer is configured to dynamically adjust traffic forwarding paths and queue scheduling parameters, and specifically comprises: The tunnel execution layer adds a tunnel encapsulation label for indicating borrowing to the traffic meeting borrowing conditions at an entrance node of the first small particle slice tunnel, guides the traffic to the second small particle slice tunnel, and meanwhile, configures a traffic order-preserving buffer zone at an exit node to reorder the data packets of the same flow arriving through different tunnels.
7. The small particle slice network oriented tunnel traffic dynamic scheduling system of claim 6, wherein the return trigger condition comprises: reaching the lending duration or monitoring that the bandwidth utilization rate of the second small particle section tunnel rises back to a preset threshold value close to the rated capacity of the second small particle section tunnel; the dynamic policy engine is further configured to generate a return instruction to be issued to a tunnel execution layer when a return condition is triggered, and the tunnel execution layer responds to the return instruction to gradually switch the borrowed flow back to the first small particle slice tunnel and release borrowed resources on the second small particle slice tunnel.
8. The small particle slice network oriented tunnel traffic dynamic scheduling system of claim 1, wherein the tunnel execution layer is further configured to: And when the temporary lending is executed, the elastic non-real-time flow is preferentially scheduled to the second small particle slice tunnel, and the real-time interaction flow is reserved in the original tunnel.
9. The system of claim 1, wherein the dynamic policy engine is further configured to maintain a slice resource banking module that stores an initial bandwidth quota, a real-time borrowing, a lending record, and a credit score for each small particle slice tunnel, and wherein the dynamic policy engine determines whether the first small particle slice tunnel has lending qualification based on the credit score of the first small particle slice tunnel and determines lendable resources based on the credit score and the unused bandwidth of the second small particle slice tunnel when generating the resource lending scheduling policy.
10. The small particle slice network oriented tunnel traffic dynamic scheduling system of claim 1, wherein the tunnel execution layer is configured to dynamically adjust queue scheduling parameters, and specifically comprises: And adjusting a gating list, priority weights and shaping rates of the first small particle slice tunnel and the second small particle slice tunnel in an exit node queue so as to realize smooth occupation and release of borrowed bandwidth during borrowing and avoid flow jitter.

Description

Tunnel traffic dynamic scheduling system oriented to small particle slicing network Technical Field The invention discloses a tunnel traffic dynamic scheduling system oriented to a small particle slice network, and relates to the field of network slicing and traffic engineering. Background Currently, a tunnel scheduling technology based on Segment Routing traffic engineering (SR-TE) mainly adopts an SR Policy system, and realizes automatic traffic flow drainage and on-demand next hop generation through color and endpoint identification of an intended driving path. However, the conventional SR-TE scheme still uses a conventional tunnel interface or a candidate path model, and has obvious disadvantages in coping with a small-particle slice scenario, for example, CN108965024a discloses a virtual network function scheduling method based on prediction for 5G network slices, which performs VNF scheduling through a queue model and prediction, but the scheme focuses on resource allocation at a virtual network function level, does not involve dynamic adjustment of a traffic forwarding path at a tunnel level, and cannot solve the problem of real-time congestion inside a small-particle slice tunnel. In the aspect of special protection technology for small particle slices, for example, CN116743584A adopts a PW 1:1 protection and SR-TP tunnel superposition mode to realize service escape under faults. The scheme belongs to a passive response type protection mechanism, and is characterized in that rerouting is triggered only after a fault occurs, impending congestion cannot be predictively avoided, meanwhile, a hard isolation mode of a main redundant tunnel and a standby redundant tunnel is adopted, the resource utilization rate is low, idle slice resources cannot be temporarily borrowed, and the elastic scheduling requirement under the sudden flow is difficult to meet. In addition, CN116743584a discloses a dynamic RAN slicing method based on information sensing and joint calculation buffering, which senses the traffic density distribution of the vehicle through ConvLSTM model to realize dynamic configuration of the radio access network slices. The scheme is applied to a wireless access network scene, mainly relates to resource allocation of a base station and a roadside unit, belongs to different technical fields with tunnel traffic scheduling in a core network or a bearing network, and has scheduling granularity of resource adjustment at a slice level, and cannot go deep into a queue depth perception inside a small-particle slice tunnel and a cross-slice resource lending mechanism. Disclosure of Invention The invention aims to provide a tunnel traffic dynamic scheduling system for a small particle slice network, which comprises a small particle slice sensing layer, a dynamic strategy engine and a tunnel execution layer, wherein the small particle slice sensing layer is used for collecting running state data of a plurality of small particle slice tunnels in the network in real time, including a queue depth change rate, a bidirectional delay difference and a current bandwidth utilization rate, the dynamic strategy engine is used for predicting congestion probability of each tunnel in a preset time window in the future by using a prediction model based on the running state data, identifying a second tunnel with idle bandwidth resources when the congestion probability of a first tunnel is predicted to exceed a first threshold value, generating a resource lending scheduling strategy for indicating the first tunnel to temporarily borrow part of bandwidth resources of the second tunnel, setting lending duration and return triggering conditions, and the tunnel execution layer is deployed in a network node and used for analyzing and executing the resource lending scheduling strategy, dynamically adjusting traffic forwarding paths and queue scheduling parameters of the first tunnel and the second tunnel and completing temporary lending of bandwidth resources. The aim of the invention can be achieved by the following technical scheme: the tunnel traffic dynamic scheduling system facing the small particle slice network comprises a small particle slice sensing layer, a dynamic strategy engine and a tunnel execution layer: The small particle slice sensing layer is used for being configured to collect running state data of a plurality of small particle slice tunnels in a network in real time, wherein the running state data comprises a queue depth change rate, a bidirectional delay difference and a current bandwidth utilization rate of each small particle slice tunnel; The dynamic policy engine is configured to be communicatively connected to the small particle slice aware layer, and configured to: based on the running state data, predicting the congestion probability of each small particle slice tunnel in a future preset time window by using a prediction model; When the congestion probability of the first small particle slice tunnel i