CN113271213A

CN113271213A - Charging network outlet routing method based on SDN

Info

Publication number: CN113271213A
Application number: CN202110508057.2A
Authority: CN
Inventors: 顾慎凯; 郭索眸; 赵志宏; 白光伟
Original assignee: Nanjing Tech University
Current assignee: Nanjing Tech University
Priority date: 2021-05-10
Filing date: 2021-05-10
Publication date: 2021-08-17

Abstract

The invention discloses a method for selecting a charging network outlet route based on a software defined network and comprehensively considering network delay and network bandwidth. The method is used for selecting the egress route in an SDN network environment accessing a plurality of charging network exits, and comprises the following steps: the SDN controller presets a charging mode of each outlet line; when an access request is received, the SDN switch firstly installs a preset flow table for forwarding, if the flow table is not matched with the preset flow table, the service type of the request is judged, the request is sent to a controller, and the controller collects information such as time delay, bandwidth and the like of each network outlet; and calculating and comparing the weights of different outlets accessing different types of services, and issuing a proper flow table according to the service types. Meanwhile, the SDN controller can update the flow table entries at regular time, adapt to a dynamically changing network environment and prevent performance loss caused by overlong flow tables. The method reduces the load of the SDN controller, increases the utilization rate of network resources, and improves the network use experience of users.

Description

Charging network outlet routing method based on SDN

Technical Field

The invention relates to the technical field of networks, in particular to a charging network outlet routing method based on an SDN.

Background

Under the increasingly prosperous internet environment, as small as individual families and cells access a plurality of Network operators, as large as schools and enterprises access ISPs (internet service providers) and education networks, internal networks of companies and Virtual Private Networks (VPNs) simultaneously, the scene of accessing a plurality of lines to connect the internet in the same local area Network environment is very common, the problem of how to realize the maximum utilization of the lines is gradually highlighted, and the method is mainly embodied in two aspects: (1) different network services and different types of services are accessed, and different lines can bring different Quality of Service (QoS); (2) different lines have different costs, different traffic limits, different allocated bandwidths and different data amounts, and different egress line options generate distinct overheads.

In order to solve the above problems, there are two main solutions in the conventional network: load balancing and manual switching. The former specifies a line of a specific flow by setting a load balancing policy. The method can balance the use of each line to a certain extent, but is difficult to predict undefined and derived use scenes in use, and is difficult to adjust in time when the network topology changes. The latter is cumbersome, user unfriendly, requires the user to manually select the line to be used, and may cause unnecessary economic loss due to mishandling.

It is seen from the existing network communication management strategy that a method other than "0", that is, "1", is difficult to meet the needs of modern communication, and designing a routing method capable of adapting to the change of a network environment is one of key problems to be solved urgently in the modern internet, so that the network overhead of a large-scale data center and a large-scale enterprise network can be reduced, and the network use experience of an individual user can also be improved.

According to characteristics that Network flow in a media production Network can be predicted, the purposes of improving resource utilization rate and saving cost are achieved through a bandwidth reservation method in a document [ Barscan M, Moens H, Famaey J, et al. The research on data transmission delay problem in a literature [ Wanglauca, Uighur traffic university, 2014.] introduces a concept of 'virtual delay', proposes a delay-insensitive data flow scheduling algorithm, can monitor a port of a switching node, and switches delay-insensitive data flow on a corresponding link to a light-load link when a port flow load exceeds a set threshold; document [ li-penguil, zhuanlei, martin, etc.. routing mechanism based on service partitioning in SDN networks [ J ] computer science, 2017 (3): 118-. The above method proposes different routing algorithms from multiple dimensions such as network state, service type, etc., but has two defects: (1) the condition that an outlet routing part is not in the controllable SDN network range is not considered, and the method is difficult to apply in the scene that the outlet network is a black box; (2) the economic expenditure of asymmetric egress and charging lines is not considered.

Disclosure of Invention

In view of the above, the present invention provides a charging network egress routing method based on SDN. The method comprises the steps of firstly marking the service type of a stream through a DSCP field, determining different routing performance index weights, comprehensively evaluating under a specific index weight scheme by means of preset line cost and time information, and selecting an optimal route which meets the QoS and reduces the economic expenditure of a charging line according to the comprehensive index of each line.

In order to achieve the purpose, the invention provides the following technical scheme:

a charging network outlet routing method based on SDN comprises the following steps:

(1) firstly, initializing an SDN controller, setting information such as bandwidth, flow limitation, flow rate and the like of each line, determining a weight calculation function, and creating a list for a flow table updating queue;

(2) the terminal network sends a data packet to the SDN switch;

(3) if the data packet is matched with the existing flow table, forwarding the data packet according to the flow table rule; if not, the SDN switch marks the service type of the flow through the DSCP field, sends the service type to the SDN controller, and executes the step four;

(4) the SDN controller detects the network state of each line access target service; determining calculation weights according to the service types, respectively calculating the weights, determining an exit route, and generating and issuing a flow table item to the SDN switch;

(5) when the SDN controller issues the flow table, recording the service type and the issuing address of the flow table item, and inserting a new table item into the tail of a queue to be updated; performing round-robin according to the busy degree of the CPU, and checking whether the exit route of the head-of-queue flow table needs to be updated;

(6) and creating a linear table on the SDN switch, and setting a length threshold of the flow table entry. And inserting a new flow table entry into the table tail, taking out the flow table entry from the linear table and re-inserting the table tail when the flow table entry is successfully matched, deleting 20% of the flow table entries when the length of the flow table entry exceeds a threshold value, sending the flow table entries to the SDN controller, and deleting the entries from a queue to be updated.

For the step (1), the setting method of the weight function is as follows:

the optional line set is P, and for any target service in the line P, the network state information of the line accessing the target service comprises a triple group of bandwidth, time delay and packet loss rate; calculating to obtain the link bandwidth utilization rate of the line;

in addition to the traditional network state, a variable rate c (p) is introduced to represent the unit rate of line p. If the line charging mode is charging according to the using amount, C (p) is the flow unit price of the line p, namely the using amount/price; if the line is a quota type line, C (p) is the residual flow/time of the line p and represents the abundant degree of quota; when the line is not counted, c (p) is 0;

after a network state measuring module is called to obtain a network state, Min-Max standardization is carried out on a calculation result, and the standardized available bandwidth is A (p)^*Delay of D (p)^*Packet loss rate L (p)^*Link bandwidth utilization of R (p)^*Unit expenditure is C (p)^*So that each measurement index can be used for uniform consideration;

the weight calculation method comprises the following steps: parameter matrix α (p) ═ a, -d, -l, -r, -c]Line state matrix s (p) ═ a (p)^*，D(p)^*，L(p)^*，R(p)^*，C(p)^*]^TThen, the weight w (p) ═ α (p) s (p).

For the step (3), the flow table matching adopts a three-layer message header and a four-layer port number; it is considered that the traffic type of the same destination address does not change in one routing cycle.

For step (3), the service type is determined by using a DSCP field with 6 bits in the IP header of the data packet, for distinguishing the service code:

according to the 3GPP standard, dividing the network service into a session service, a streaming media service, an interactive service and a background service;

the DSCP can be divided into a Class Selector, Expedited Forwarding, Assured Forwarding and Default defaults, so as to determine the service type;

for example, CS4 is streaming video, which is classified as streaming media service, and EF is interactive voice, which is classified as interactive service.

For the step (4), the bandwidth, the time delay and the packet loss rate of each line access service need to be measured on the SDN switch;

in step (4), when parameters are set on the SDN controller, the sizes of a, d, l, and r are dynamically adjusted according to different service types, and the following requirements are required to be met:

the conversation type service comprises voice service, video conference and the like, the time delay is sensitive, the requirements of packet loss and streaming media service on the time delay can be allowed to be lower than that of the conversation type service, the same service can tolerate a certain packet loss rate, the interaction type service comprises Web service and the like, the sensitivity on the time delay is lower, the sensitivity on the packet loss rate is improved, and the background type service is not sensitive to the time delay but has more strict requirements on the packet loss rate;

according to the above requirements, the relationship of the weights that different types of services should satisfy is as follows, the expenditure weight c ∈ [0, 1] is set during initialization and the reserved interface is customized by the user, reflecting the sensitivity of the user to the economic expenditure:

for step (5), using additional storage space in the SDN controller for maintenance of queues to be updated:

the storage mode records an associated array taking the binary group of the exchanger and the number of the flow table as a key value and taking the service type and the forwarding line as numerical values;

setting a maximum polling period tau, wherein the polling period is the product of tau and the current CPU occupancy rate: t ═ τ × r (CPU), a timer with a duration of T is set, and flow table updating is avoided when the CPU occupancy is high;

when a new flow table is issued, generating an associated array to be inserted into the tail of the queue;

and when the timer is triggered, taking out the array at the head of the queue, recalculating the outlet line, if the outlet line is changed, executing the step four, otherwise, inserting the array at the tail of the queue.

For step (6), a linear table is created in the SDN switch for control of the number of flow tables:

the storage mode records the number of the flow table as a key value and stores the key value in the linear table, and the table entry is taken out from the linear table and inserted into the tail of the queue after each successful flow table matching;

when the number of the flow tables exceeds a threshold value, sequentially taking out table entries with the total length of 20% from the head of the queue, and deleting the flow tables;

and reporting the deleted flow table numbers to an SDN controller, and removing the flow table numbers from the queue to be updated, so that repeated calculation is avoided, and the overhead is reduced. The method and the system not only ensure that the flow table dynamically changes along with the network environment, but also prevent old data from occupying the flow table space for a long time.

Compared with the prior art, the invention has the following advantages:

the method can be used for routing selection under the condition that part of network environment is a black box scene, and can be dynamically adjusted according to the change of the network state in the black box, compared with the prior art, the method reduces the network modification cost, and widens the application scene of the routing method;

secondly, the invention provides a judgment basis for taking the economic expenditure of the network as the routing selection, when the user accesses the limited flow sum or the charging network, the economic expenditure of the user can be effectively reduced, and compared with the prior art, the use cost of the network is reduced.

Drawings

FIG. 1 is a schematic diagram of a route forwarding model;

figure 2 is a diagram of an SDN network architecture according to the present invention;

FIG. 3 is a flow chart of a main implementation of the routing method according to the present invention;

fig. 4 is a flowchart of an implementation of the method for maintaining a flow table on an SDN controller according to the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

The following further describes the embodiments of the present invention with reference to the attached drawings:

charging network outlet routing method based on SDN

Fig. 1 is a schematic diagram of a route forwarding model according to the present invention. Fig. 1 illustrates a scenario in which a user terminal accesses an SDN network, connects to the internet via an SDN switch, and accesses various network services.

Fig. 2 is a diagram illustrating an SDN network architecture according to the present invention, where the network architecture of the charging network egress routing method according to the present invention includes an SDN controller, an SDN switch, and an end user.

The terminal user generates various types of service data, including session data, streaming media data, interactive data and background data; an SDN switch is a data forwarding device in the entire fabric. The flow table matching forwarding module is a basic function of the SDN switch and forwards data according to the flow table; the service type judging module judges the service type on the SDN switch through data plane programming and provides the service type for the SDN controller, and the flow table item management module manages the flow table to avoid overlong flow table. The SDN controller is the core of the overall architecture. The SDN controller defines a routing selection module, a network state measurement module, a cost management module, a flow table updating monitoring module and a flow table issuing module. The network state measurement module is communicated with the SDN switch through a southbound interface to obtain time delay, available bandwidth and packet loss rate information of an outgoing line; the expense management module is responsible for adjusting the weight of the line expense and recording the flow use conditions of different lines; the flow table updating and monitoring module is communicated with the SDN switch through a southbound interface to ensure the consistency of the SDN controller and the SDN switch in maintaining the content of the flow table, and the routing module calculates the optimal exit route according to the information of the network state and the cost; and the flow table issuing module encapsulates the optimal exit route into a flow table and issues the flow table to the SDN switch.

Fig. 3 is a flowchart illustrating a main implementation of the routing method of the present invention, which specifically includes the following steps:

initializing an SDN controller, setting the set of outlet lines as P, setting the bandwidth B (P) of each outlet line P belonging to P and the sensitivity c (c belonging to [0, 1]) of the charge, and setting the line rate C (P) according to the following mode.

Let the line charging period be a charging _ cycle, elapsed Time _ past, line traffic Quota be quote, used traffic in the period be quote _ used, unit Price of per-volume charging line unit data volume in the period be Price, thus:

elapsed time ratio

Consumed flow rate ratio

Line tariff

For example, line p₁For conventional unlimited lines, C (p)₁) Line p ═ 0₂The charging period is 30 days, the available flow is 1000GB, the charge is 100 yuan, the whole 15 days are passed, the used flow is 400GB, at this time

Line p₃The traffic cost is 10 yuan per 1GB used, at this time C (p)₃)＝10。

And (2) the terminal network sends the data packet to the SDN switch.

And (3) if the data packet is matched with the existing flow table, forwarding the data packet according to the flow table rule, wherein the flow table items issued by the invention adopt three-layer message headers and four-layer port numbers for matching. Meanwhile, the service type of the same destination address is not considered to change in a routing period.

And if not, modifying the data forwarding pipeline logic through data plane programming, reading a DSCP field with 6 bits in an IP packet header of the data packet, distinguishing the service types, and sending the service types to the SDN controller to execute the step four.

DSCP refers to Differentiated Services Code Point (Differentiated Services Code Point). It prioritizes by encoding values using used 6 bits and unused 2 bits in the class of service ToS identification byte of each packet IP header.

DSCPs fall into four categories in total: class Selector (CS) AAA000, Expedited Forwarding (EF)101110, Assured Forwarding (AF) AAABB0, default Default (BE) 000000.

And (4) detecting the network state of each line access target service by the SDN controller. The detection method can be customized according to the actual scene, such as the sflow of the third-party tool.

The available bandwidth A (p), the time delay D (p) and the packet loss rate L (p) are measured for the line p, and the bandwidth utilization rate can be calculated

Reading C (p) from the expense management module, then carrying out Min-Max standardization,

the standardized available bandwidth is obtained as A (p)^*Delay of D (p)^*Packet loss rate L (p)^*Link bandwidth utilization of R (p)^*Unit expenditure is C (p)^*So that each measurement index can be used for uniform consideration;

then, the weight is set according to the service type, and the relationship of the weight which different types of services should satisfy is as follows

For example, in the step one, the tariff weight c is set to 0.2, for the session-like service, the delay weight d is set to 0.3, the link bandwidth utilization ratio weight r is set to 0.25, the available bandwidth weight a is set to 0.15, and the packet loss ratio weight l is set to 0.1.

Fig. 4 is a flowchart illustrating an implementation of the method for maintaining a flow table on an SDN controller according to the present invention, where the specific flow includes the following steps:

step (5) is to maintain a flow table to be updated in the SDN controller:

and creating a queue, and storing an associated array taking the binary group of the switch and the flow table number as a key value and the service type and the forwarding line as numerical values in the queue.

And setting a timer with the duration of T, and maintaining the flow table queue each time the timer is triggered. In order to avoid the performance deficiency caused by frequently updating the queue, the T is reset according to the following method after the timer is triggered each time:

by setting a maximum polling period tau, the current CPU occupancy rate is multiplied by tau to realize dynamic regulation: t τ × r (cpu).

When the SDN controller issues a flow table, inserting an associated array into the tail of a queue;

and when the timer is triggered, taking out the array at the head of the queue, recalculating the outlet line, if the outlet line is changed, executing the step four, otherwise, inserting the array at the tail.

Step (6) is the maintenance of a flow table in the SDN switch:

creating a linear table, and storing the address of the flow table entry in use, namely the pointer of the flow table entry in the linear table;

when the flow table is successfully matched each time, the pointer of the table entry is taken out and reinserted into the tail of the queue, so that the matching times of the table entries closer to the head of the queue are lower;

when the length of the flow table reaches a threshold value, the table entry with the length of 20% of the head of the queue is deleted, so that the situation that when the flow table is full, the SDN switch repeatedly switches in and out the flow is avoided, the load of the controller is greatly increased, and meanwhile, the time delay loss caused by overlong flow table is also reduced.

Claims

1. A charging network outlet routing method based on SDN is characterized by comprising the following steps:

(2) the terminal network sends a data packet to the SDN switch;

(6) the method comprises the steps that a linear table is established on an SDN switch, a length threshold value of flow table entries is set, new flow table entries are inserted into a table tail, when the flow table entries are successfully matched, the flow table entries are taken out of the linear table and inserted into the table tail again, when the length of the flow table entries exceeds the threshold value, 20% of the flow table entries are deleted and sent to an SDN controller, and the table entries are deleted from a queue to be updated.

2. The SDN-based charging network egress routing method according to claim 1, wherein the weight function is set in step (1):

the weight calculation method comprises the following steps: parameter matrix a (p) is [ a, -d, -l, -r, -c]Line state matrix s (p) ═ a (p)^*，D(p)^*，L(p)^*，R(p)^*，C(p)^*]^TThe weight w (p) is a (p) s (p).

3. The SDN-based charging network egress routing method according to claim 1, wherein in step (3), the flow table matching uses a three-layer header and a four-layer port number; it is considered that the traffic type of the same destination address does not change in one routing cycle.

4. The SDN-based charging network egress routing method of claim 1, wherein the service type determination in step (3) adopts a DSCP field with 6 bits in an IP header of the data packet for distinguishing service codes:

5. The SDN-based charging network egress routing method according to claim 1, wherein in step (4), the SDN controller invokes an SDN switch to measure bandwidth, delay, and packet loss rate of each line access service.

6. The SDN-based charging network egress routing method according to claim 1, wherein when the SDN controller sets the parameters in step (4), the sizes of a, d, l, and r are dynamically adjusted according to different service types, and the following requirements are satisfied:

7. the SDN-based charging network egress routing method according to claim 1, wherein in step (5), extra storage space is used in the SDN controller for maintaining queues to be updated:

8. The SDN-based charging network egress routing method of claim 1, wherein in step (6), a linear table is created in the SDN switch for controlling the number of flow tables:

and reporting the deleted flow table numbers to an SDN controller, removing the flow table numbers from a queue to be updated, avoiding repeated calculation, and reducing the overhead.