CN106789660B - A QoS-aware approach to traffic management in software-defined networking - Google Patents

Abstract

The invention discloses a QoS (quality of service) perceptible flow management method in a software defined network, which comprises the following steps: sending the QoS requirement to the network through the network upper layer application of the software defined network; the network checks the service quality requirement of the application and divides the application flow into a plurality of categories with different priorities; identifying the application traffic, and performing different routes on the application traffic of different categories; monitoring the utilization rates of all links, and performing flow scheduling on the congested link; and carrying out flow limitation on the congestion link after flow scheduling. Different QoS strategies are adopted for data streams with different requirements, data with high priority are processed preferentially, flow in the network is treated differently, and smooth passing of key service flow through the network is guaranteed.

Description

A QoS-aware approach to traffic management in software-defined networking

technical field

The invention relates to a traffic management method in a software-defined network, in particular to a QoS-aware traffic management method in the software-defined network. Different traffic scheduling and flow rate control methods are used for different types of business traffic, and differentiated QoS policies are implemented.

Background technique

Guaranteeing the quality of service (QoS) of the network is a complex problem. In general, traditional networks treat data streams indiscriminately, perform services according to the principle of best-effort, and pursue maximization of throughput. Different applications have different requirements for network service quality. For example, in an enterprise private network, the data volume of customers' daily operations (access, storage, etc.) is small, and the network delay will directly affect the user experience. The data used for program data synchronization in the enterprise server, although the amount of data is huge, can bring a lot of pressure to the network, but this part of the data is relatively less sensitive to time. In the face of today's huge network scale and massive network data, the single processing method of the traditional network is not enough.

Software Defined Network (SDN) separates the control layer from the data layer. The control layer can uniformly allocate network resources from a global perspective. Network managers can formulate various strategies at the control layer to ensure the stable and efficient operation of the network. One of the most popular standards in software-defined networks is OpenFlow. The functions of the control layer are implemented by the controller, and the functions of the data layer are implemented by the main switches. The controller controls the traffic transmission path by delivering flow entries to each switch.

The software-defined network has centralized management and software programmable features. It can formulate appropriate QoS policies according to the needs of upper-layer services, and more efficiently serve network services with different needs. The present invention also makes use of the advantages of software-defined networks to formulate suitable traffic management strategies.

SUMMARY OF THE INVENTION

In view of the above-mentioned technical problems, the purpose of the present invention is to: provide a QoS-aware traffic management method in a software-defined network, adopt different QoS policies for data streams with different requirements, and prioritize data with high priority. Differentially treat traffic in the network to ensure that critical business traffic flows smoothly through the network.

The technical scheme of the present invention is:

A QoS-aware traffic management method in a software-defined network, comprising the following steps:

S01: Send QoS requirements to the network through the upper-layer network application of the software-defined network;

S02: The network checks the application service quality requirements, and divides the application traffic into several categories with different priorities;

S03: Identify application traffic, and perform different routes for different types of application traffic;

S04: Monitor the usage rates of all links, and perform traffic scheduling on congested links;

S05: Perform traffic restriction on the congested link after traffic scheduling.

Preferably, the traffic classification in the step S02 is based on the QoS requirements of the traffic and the importance of the traffic, including minimum required bandwidth, delay tolerance range, and service type.

Preferably, the method for identifying different types of traffic in the step S03 is: the network device identifies by checking the matching field corresponding to the packet header of the application, and the traffic marked with the application is regarded as a priority flow, otherwise it is regarded as no priority flow level flow.

Preferably, the routing method in the step S03 is:

For the flow with priority, the controller calculates the shortest path of the traffic in the network. If each link on the shortest path meets its bandwidth requirements, it will route according to this path; For the flow with low priority, and the required bandwidth of the low-priority flow exceeds the required bandwidth of the current priority, it will be routed according to the shortest path, and the flow of the low-priority flow on the path will be scheduled at the same time; If there is no low-priority flow or traffic scheduling cannot be performed, the next matching path is searched for routing in turn. If no path that meets the requirements is found, routing is performed according to the shortest path, and traffic restriction is enabled at the same time;

For non-priority flows, the path with the lowest network bandwidth utilization is chosen for routing.

Preferably, the traffic scheduling method is as follows: the controller detects the network link, when the link utilization rate exceeds a threshold, schedules low-priority traffic on the link, and schedules a part of the traffic on the link with a large amount of data. Guided to other links, the controller filters out suitable data flows, notifies the corresponding switch to modify the flow table entry, and change the traffic forwarding path; if no other idle links can be found or the link bandwidth utilization is still high after traffic scheduling Enable traffic throttling.

Preferably, the flow restriction method is:

Set the rate thresholds for different types of traffic. The low-priority traffic rate threshold is equal to the difference between the maximum rate of the link and the total rate of high-priority traffic. Set the same number of queues as the traffic priority, and each queue only serves a certain type of traffic. , and set the minimum transmission rate for it, the minimum transmission rate value is equal to the sum of all traffic demand rates of the queue;

The controller monitors the link usage, counts the traffic of each priority level, and dynamically limits the low-priority traffic rate threshold. First, the lowest-priority traffic rate is reduced to the lowest level, and then the sub-priority traffic rate is limited.

Compared with the prior art, the advantages of the present invention are:

1. Increase the network's perception of upper-layer application services, and avoid the problem of unreasonable network resource allocation caused by indiscriminate treatment of traffic in traditional networks.

2. Differentiated routing strategies and traffic scheduling strategies can arrange suitable links for them according to business requirements, ensure transmission quality, and relieve network congestion to a certain extent.

3. The traffic control mechanism combining Queue and Meter is adopted to limit the rate of the flow set according to the priority to ensure the smooth transmission of important business traffic.

Description of drawings

Below in conjunction with accompanying drawing and embodiment, the present invention is further described:

1 is a block diagram of steps of a traffic management method of the present invention;

Fig. 2 is the flow chart of routing method;

Figure 3 is a flow control flow chart.

Detailed ways

The above scheme will be further described below in conjunction with specific embodiments. It should be understood that these examples are intended to illustrate the invention and not to limit the scope of the invention. The implementation conditions used in the examples can be further adjusted according to the conditions of specific manufacturers, and the implementation conditions not specified are usually the conditions in routine experiments.

Example:

As shown in FIG. 1 , the QoS-aware traffic management method in the software-defined network (SDN) of the present invention mainly includes the steps of traffic identification, dynamic routing, traffic scheduling and flow rate regulation. Specific steps are as follows:

(1) In the software-defined network (SDN) environment, the upper-layer application on the network sends the quality of service (QoS) requirements to the network;

(2) The network checks the application QoS requirements, and divides the application traffic into several categories with different priorities;

(3) The network identifies the traffic in step (2), and arranges routes for different types of traffic respectively;

(4) The network monitors the utilization rate of all links and schedules traffic on congested links;

(5) rate limiting the link that is still congested after the traffic scheduling in step (4);

Traffic identification and classification:

The upper-layer application on the network sends its QoS requirement (bandwidth, delay) information to the control layer, and the application can modify its requirement information at any time;

The controller checks the QoS requirements and classifies the traffic in the network: critical (data is extremely important, or very time-sensitive), high-priority data, general-priority data, and no-priority data.

The upper-layer application needs to mark the data packets with priority, and the data packets without marking are data without priority, that is, the priority is the lowest. The matching field in the packet header, such as the ToS (Type of Service) field in the IP protocol, can be used for marking, and switches and controllers identify the priority of the corresponding traffic by checking the matching field.

As shown in Figure 2, the routing method is: after the new data flow enters the network, the edge switch encapsulates the packet-in message and sends it to the controller, the controller parses the packet-in message, the controller checks the header of the data packet, and executes it according to the priority. routing;

If the data packet does not have a priority, the controller finds all its traversable paths through an algorithm, and selects the most idle (lowest usage) path for routing.

If the data packet has priority, it needs to check the link usage before routing: the controller calculates the shortest path (default path) of the data packet through the network through an algorithm. If each link on the shortest path meets its bandwidth requirements, Then route according to this path. If the bandwidth of a link on the shortest path does not meet its requirements, the route is performed as follows:

For a flow with a certain priority, if there is a flow with a lower priority than the flow on the link on the shortest path, and the required bandwidth of the low-priority flow exceeds the required bandwidth of the current priority, it will still be routed according to the shortest path , and perform traffic scheduling on the low-priority flow on the path at the same time; if there is no low-priority flow in the link, or the traffic scheduling cannot be performed, then follow this method to find the next matching path for routing. If no suitable path is found that meets the requirements, the default shortest path is used for transmission, and traffic restriction is enabled. In this way, the highest priority flow is always routed according to the shortest path first, and the lower priority flow is always preferentially rerouted or limited.

As shown in Figure 3, the specific method of traffic scheduling is:

Traffic scheduling is initiated when a new data flow needs to use the link or when the link usage is too high. In the case of traffic scheduling caused by the arrival of a new data flow to the network, the scheduled flow is a flow with a lower priority than the new flow; in the case of traffic scheduling caused by excessive link bandwidth usage, all flows can be scheduled; select scheduling The flow method is: select the elephant flow in the lowest priority flow in sequence according to the priority, and guide its flow to other idle links, and the low priority flow will be guided first; only the elephant flow will be scheduled, and the mouse flow will be scheduled. Will not be scheduled; if all the elephant flows are guided and still do not meet the bandwidth requirements of the new data flow, the flow of the low-priority flow will be limited; after the flow scheduling, the link usage is still too high to meet the demand. Flow control.

The specific methods of flow control are as follows:

Use Meter to set the rate threshold of each type of traffic. A certain type of flow is associated with the rate threshold set in the Meter Band, which is equal to the difference between the link's assignable rate and the rate of high-priority traffic; set the same number of queues as the number of traffic priorities. The queue only serves a certain type of flow, is used to guarantee the minimum transmission rate of the data flow, and sets the minimum transmission rate for it. The initial minimum rate of the queue is a certain control value R (R>0). According to the bandwidth value required by the traffic associated with the queue, the minimum rate of the queue is dynamically increased or decreased; if the set value of the minimum rate exceeds the value in the Meter table for this type of traffic When the meter threshold is raised, the minimum transmission rate value of the queue will increase to the rate required by the corresponding flow along with the threshold. The low-priority flow rate threshold is equal to the difference between the maximum rate of the link and the total rate of high-priority flows, that is, for a flow of a certain priority, the flow rate is maximized when the rate of higher priority is maximized; If the rate of the flow exceeds the threshold, the rate sum of all low-priority flows is limited; the minimum value of the rate threshold is not lower than the queue rate control value R, and the maximum value of the threshold does not exceed the maximum rate of the link.

The controller monitors the link usage, counts the traffic of each priority level, and dynamically limits the rate threshold of low-priority traffic so that it does not affect the rate of traffic with a higher priority than it; It is to first reduce the lowest priority rate to the lowest, and then limit the priority rate, which can ensure the maximization of the high-priority flow rate.

Specific examples are given below:

(1) Assume that there are four types of applications A ₁ , A ₂ , A ₃ , and A ₄ in the network, and the priorities decrease in turn. The data of A ₁ , A ₂ , and A ₃ have priority, and the traffic of A ₄ has no priority.

(2) Routing arrangement:

_The traffic of the A1 application has the highest priority. The controller calculates the path of its traffic through the network, and checks the unallocated bandwidth of each link on the path. If it meets the requirements _of the A1 application, it will route according to this path; if If the link bandwidth is not satisfied, schedule or limit the traffic applied by A ₂ , A ₃ , and A ₄ on the link;

A ₂ applies the same traffic routing principle as A ₁ , and if and only if the congested link low-priority flows (A ₃ , A ₄ ) cannot be scheduled and the remaining bandwidth of the link does not meet the requirements of A ₂ , it loops to find the next optimal flow path. If no satisfying path is found, select the original optimal path for transmission, and restrict the traffic of A ₃ and A ₄ ;

_The traffic routing method of A3 is the same as that of A2 _; the application traffic of A4 is a flow without priority _. During routing, the controller finds the most idle path in the network for transmission, and can transmit in multiple paths according to its bandwidth requirements.

(3) Traffic scheduling: There are two situations for starting traffic scheduling: 1) The link bandwidth utilization rate is too high, then the link scheduling reduces the link utilization rate to a certain threshold (set to 80%); 2) Priority If a flow with a higher priority needs to occupy the bandwidth of the link B, the flow with a lower priority is scheduled to satisfy its bandwidth.

For the scheduled link L, the above-mentioned four kinds of traffic exist on the link, and it is necessary to schedule the flow with the priority lower than P. The controller filters out flows with a priority lower than P, checks the counters in each flow entry or meter band, sorts the flows according to the count, and selects the flows with the count value greater than a certain value X and the lowest priority for scheduling. Calculate whether there are other schedulable paths for this part of the flow. If so, the controller modifies the flow entry on the corresponding switch to forward the traffic through other paths. After the flow is scheduled, the minimum transmission rate R _Q set by the corresponding queue will decrease, and the minimum transmission rate value of the traffic-directed link queue will increase. If the demand is still not met after scheduling, the traffic will be restricted on the link.

(4) The method of traffic limitation is: several Meter Bands are set in the Meter Table, M ₁ , M ₂ , M ₃ , and M ₄ record the traffic of A ₁ , A ₂ , A ₃ , and A ₄ respectively, and M _{1 *} records For traffic with _a lower priority than A1, M _2* , M _3* are the same, and the corresponding rate thresholds are recorded as R _M1 , R _M2 , R _M3 , R _M4 , R _M1* , R _M2* , R _M3* . R _Max is the maximum rate that the link can allocate, and the actual rates of A ₁ to A ₄ on the link are denoted as R _{1 to} R ₄ .

i为队列中流的数量，本例中设置：

某类流量的最小阈值不小于服务于该优先队列的最小传输速率值：R_Q1min≤R_Q1≤R_M1，R_Q2min≤R_Q2≤R_M2，R_Q3min≤R_Q3≤R_M3。Queues Q ₁ , Q ₂ , and Q ₃ are set in the switch to ensure the traffic of A ₁ , A ₂ , and A ₃ respectively. The minimum transmission rates set by the queues are R _Q1 , R _Q2 , and R _Q3 respectively, where R _Q1 >R _Q1min , R ₂ >R _Q2min , R ₃ >R _Q3min . Let R _i be the required rate of flow,

i is the number of streams in the queue, set in this example:

The minimum threshold of a certain type of traffic is not less than the minimum transmission rate value serving the priority queue: R _{Q1min ≤R Q1} ≤R _M1 , R _{Q2min ≤R Q2} ≤R _M2 , R _{Q3min ≤R Q3} ≤R _M3 .

Then the above rates satisfy the following inequalities:

The calculation formula of R _M1 ~ R _M4 is:

in,

To sum up, it can be concluded that the method of restricting the flow rate is to make the maximum value of R _M1 to R _M4 in order under the condition of satisfying (1) and (2) inequalities.

The above examples are only to illustrate the technical concept and characteristics of the present invention, and the purpose is to enable those who are familiar with the technology to understand the content of the present invention and implement it accordingly, and cannot limit the protection scope of the present invention. All equivalent transformations or modifications made according to the spirit of the present invention should be covered within the protection scope of the present invention.

Claims (4)

1. A QoS aware traffic management method in a software defined network, comprising the steps of:

s01: sending, by a network upper layer application of the software defined network, a quality of service (QoS) requirement to the network;

s02: the network checks the QoS requirement of the application and divides the application flow into a plurality of categories with different priorities;

s03: identifying application traffic, and performing different routes on different classes of application traffic, wherein the classes of the application traffic comprise priority flows and non-priority flows; the routing method comprises the following steps:

for the flow with priority, the controller calculates the shortest path of the flow in the network, if each link on the shortest path meets the bandwidth requirement, the flow is routed according to the path; otherwise, if the link on the shortest path has a flow with a priority lower than that of the flow, and the required bandwidth of the low-priority flow exceeds the required bandwidth of the current priority, routing according to the shortest path, and simultaneously carrying out flow scheduling on the flow with the low priority on the path; if no flow with low priority exists in the link or the flow scheduling cannot be carried out, the next path which meets the requirement is sequentially searched for routing, if no path which meets the requirement is found, the routing is carried out according to the shortest path, and meanwhile, the flow limitation is started;

for the non-priority flow, selecting a path with the lowest network bandwidth utilization rate for routing;

the method for traffic scheduling comprises the following steps:

selecting the elephant flow in the lowest priority flow according to the priority in turn, and guiding the flow to other idle links; if all the elephant flows are guided and still do not meet the bandwidth requirement of the new data flow, carrying out flow limitation on the low-priority flow;

s04: monitoring the utilization rates of all links, and performing flow scheduling on the congested link;

s05: and carrying out flow limitation on the congestion link after flow scheduling.

2. The method of claim 1, wherein the classification of the traffic in the step S02 is based on the QoS requirement of the traffic and the importance of the traffic, including minimum required bandwidth, delay tolerance range, and traffic class.

3. The QoS aware traffic management method in a software defined network according to claim 1, wherein the method of identifying different classes of traffic in step S03 is: the network equipment identifies by checking a matching field corresponding to the header of the data packet of the application, and the traffic with the application mark is regarded as a priority flow, otherwise, the traffic is regarded as a non-priority flow.

4. The QoS aware traffic management method in a software defined network of claim 1, wherein the traffic limiting method is:

setting speed threshold values of different types of flow, wherein the low-priority flow speed threshold value is equal to the difference between the maximum speed of a link and the total speed of a high-priority flow; setting queues with the same number as the flow priority levels, wherein each queue only serves flows of a certain category and sets a minimum transmission rate for the queues, and the minimum transmission rate value is equal to the sum of all flow demand rates of the queues;

the controller monitors the link service condition, counts the flow of each priority level, dynamically limits the threshold value of the low-priority flow rate, firstly reduces the lowest-priority flow rate to be lowest, and then limits the second-best-priority flow rate.

Images