CN106385641A - SDN-based live broadcast video streaming media distribution method - Google Patents

Abstract

The invention discloses a live broadcast streaming media routing optimization method based on a software defined network (SDN) concentrated central control unit. The method comprises the steps that S1, a video live broadcast service platform operation period is cut into multiple time segments; S2, current network topology information is collected through an SDN underlying switch at the beginning of each time segment and converged to an SDN central control unit through combination of live broadcast channel information and user information provided by a live broadcast platform; S3, the quality of service QoS is ensured by adopting a Lyapunov optimization method, the optimal solution is obtained by using a Pi-Page Rounding algorithm on the basis of ensuring the quality of service QoS, and the decision result includes routing selection of each video live broadcast channel; and S4, the routing selection decision is made for each video live broadcast channel based on the acquired decision result, decision information is transmitted to each SDN switch through an SDN southbound interface, and watching feedback information from users is also received in real time.

Description

A kind of live video stream media distribution method based on SDN

Technical field

The present invention relates to streaming media video, multi-media network, software defined network and cloud computing resources management domain, more Body ground, relate to a kind of live video stream media distribution method Ji Yu SDN (software defined network).

Background technology

In recent years, blowout growth in internet traffic, and especially multimedia video flow contribute to very big proportion.In advance Count 2019, multimedia video flow will account for the 80% of whole internet traffic.Meanwhile, wireless along with 3G/4G/WiFi The rapid deployment of network and the popularization of intelligent terminal, the userbase also rapid growth of mobile Internet, people can be more Conveniently and efficiently access in the Video service of magnanimity.

As the key application in Video service, net cast technology is widely used in every field, including but do not limit to In information service (as news and competitive sports), education services (online education and open class), entertainment service, (game is live And video display are live) also have commerce services (video conference and video monitoring).For example, the live platform Twitch that plays exists Abroad highly successful simultaneously, domestic live platform is also such as emerged rapidly in large numbersBamboo shoots after a spring rain；The maximum streaming media video provider of U.S.'s share The flow in the Internet peak time for the Netflix has accounted for the 37% of whole internet traffic.

In net cast platform architecture, video distribution system occupies very important status.In order to meet not same district The user in domain can quickly access the requirement of video, and how video content provider is setting up data center, using gateway service The dns server of device or ISP makes requests on the suitable data center of redirection selection to for user.When During one video of user's request, send this request to the self-built gateway server of live platform first, then according to self adaptation Algorithm selects optimal data center for this user, is directly communicated afterwards between data center and user, sends video straight Broadcast stream.

Although net cast development is swift and violent, builds high-quality net cast platform and still suffer from many difficult points.

First is the multiformity of user equipment.User can watch video using different equipment, from traditional IPTV, PC Panel computer till now, smart mobile phone, or even can watching video live broadcast in a lot of embedded devices.Each equipment Resolution, video processing capabilities are all not quite similar, and meanwhile, the access network of user is also not quite similar, and user can be reached by Wireless network, home network access Video service；This requires video provider to must provide for multiple Streaming Media versions, according to user Own situation adaptively goes to select video version.

Second is the multiformity of net cast channel.Net cast channel species is many, viewing number has greatest differences.Example As in 2014, exceeded million people and world cup being watched by net cast, accounted for the 40% of flow at that time；For example, Twitch Platform about fifty-five million spectators watch 1,000,000 different channels, and front 20% channel has attracted general 80% Spectators, assume typical long tail effect.

In order to tackle this huge challenge, fundamentally ensure the QoE of user, need to whole from user equipment to video source The transmission path of individual end-to-end (end-to-end) control effectively；When there is network congestion, can be by adjusting link, tune Section video code rate, or even the interim mode strengthening wireless network transmissions power, the QoS that safeguards system provides.Meanwhile, in order to carry out Coordination between direct broadcast band, it is to avoid be absorbed in local congestion in default of global information it should enter using central control unit Row centerized fusion.

The next generation network framework that appears as of software defined network (SDN) specifies direction.By the southbound interface of SDN (typical such as OpenFlow agreement), SDN controller can collect the network topological information of the overall situation, and passes through northbound interface and net The interaction of network application program obtains real-time routing algorithm, carries out route test to live video stream afterwards.

In order to provide real time streaming video distribution service, net cast platform operation business needs to ISP (Internet service offer Business) buy corresponding service and bandwidth.For net cast platform operation business, bandwidth cost constitutes its main fortune Battalion's cost, in order to not only ensure the Consumer's Experience of spectators, can reduce the bandwidth cost that video distribution is brought, video simultaneously Live platform operation business needs a kind of optimization algorithm.

In sum, from the angle of net cast platform operation business, in order to minimize operation cost (bandwidth cost), Ensure the good viewing experience of user, net cast platform needs to design a kind of strategy and comes optimization bandwidth allocation and video simultaneously Stream distribution is so that net cast platform operation business can provide Consumer's Experience as well as possible with minimum expense.

Mukerjee,M.K.,Naylor,D.,Jiang,J.,Han,D.,Seshan,S.,and Zhang,H, “Practical,Real-time Centralized Control for CDN-based Live Video Delivery”, In Proceedings of the 2015ACM Conference on Special Interest Group on Data Communication. centerized fusion unit is introduced CDN platform by this technology, collects information and the overall situation of all distribution videos Network topological information, reduces algorithm complex using integer partitioning approximate data, for each video flowing distribution bandwidth and make Routing decision.Depend on existing CDN framework in this technology, consider not comprehensive in optimization aim, only user is received Code check as judging quota, and, this technology does not have and considers systematic function and expense from the cycle of operation of whole system Between balance.Finally, this technology provides approximate data rather than optimization algorithm in each time period.

F.Chen,C.Zhang,F.Wang,and J.Liu,“Crowdsourced Live Streaming Over The Cloud ", in INFOCOM, 2015. this technology are from video uploader angle it is considered to work as game main broadcaster to upload live video When, how to select an optimum server to receive the video flowing of its upload for it.When weighing server optimality, this skill Art is mainly using Regional Distribution and server load two indices as criterion.In the art, only discuss and how to be Video source selects an optimum video to upload point, not by other roles in net cast platform, such as：The user of spectators Experience, is added in the consideration of algorithm design.Merely the angle from video uploader enters line algorithm design it is impossible to ensure algorithm Actual optimization effect, and Consumer's Experience cannot be ensured.

Ganjam,A.,Siddiqui,F.,Zhan,J.,Liu,X.,Stoica,I.,Jiang,J.,...&Zhang,H, “C3:Internet-scale control plane for video quality optimization.In12th USENIX Symposium on Networked Systems Design and Implementation ", in NSDI, 2015. this technology Devise a kind of key-course-data Layer separation architecture being referred to as C3, to improve the Consumer's Experience in video playback.In order to solve The autgmentability of system, C3 reaches magnanimity extension by sacrificing certain model accuracy；Meanwhile, various in order to solve equipment Property, all of decision making algorithm has all been placed in a centerized fusion device C3, designs between key-course data layer simultaneously One intermediate layer is supplied to model in order to unified video playback information.Although this technology make use of the thought of SDN to be controlled The separation of preparative layer-data Layer, but its mainly solve be device diversity and the system expandability problem, do not carry out Routing optimality, and the method employing machine learning in its controller be each request make a policy not consider the overall situation Network topological information.

Egilmez H E,Tekalp A M.,“Distributed QoS architectures for multimedia streaming over software defined networks”,in IEEE Transactions on Multimedia, 2014. this technology utilize the global network topology information of software defined network and distributed controller frame, there is provided One streaming media distribution solution that can carry out horizontal extension.Meanwhile, this technology has good motility, adapts to not Same demand (such as low delay, high code check etc.).This technology does not have between consideration different video source in the process being optimized Cooperative cooperating, may result in the network congestion of local, meanwhile, do not account for yet the cost of net cast platform operation business because Element.Therefore, the video distribution strategy being obtained by this technology is not effective to ensure that Consumer's Experience, cannot guarantee that video is straight The cost broadcasting platform operation business can be controlled in a rational scope.

Content of the invention

Present in net cast platform, magnanimity channel, spectators' skewness weighing apparatus, equipment complexity are various, bandwidth disappears The problems such as consume huge, in order to, on the basis of meeting service quality, reduce the operation of net cast platform operation business as much as possible Cost, the present invention proposes a kind of live broadcast stream media distribution method Ji Yu SDN (software defined network).

To achieve these goals, the technical scheme is that：

A kind of live broadcast stream media distribution method based on SDN, comprises the following steps：

S1. net cast service platform cycle of operation is cut into several time periods；

S2., when starting each time period, current network topology information is collected by SDN bottom switch, in conjunction with straight Direct broadcast band information, the user profile of platform offer are provided, collect to SDN central control unit；

S3. service quality QoS is guaranteed using Lyapunov optimization method, utilize Page Rounding to calculate on this basis Method obtains optimal solution, and its result of decision includes the Route Selection of each net cast channel；

S4. making routing decisions are made to each net cast channel based on the result of decision obtaining, by SDN south orientation Interface sends decision information to each SDN switch, and real-time reception is from the viewing feedback information of user simultaneously.

The mode solving acquisition decision-making in step S3 is specially：In a detailed embodiment, the present invention can be by Littleization operation cost, the optimization problem of optimization service quality are converted into Lyapunov optimization problem, by the shape of systematic collection State information, as the known conditions of this optimization problem, using service quality QoS index (time delay) as constraints, and is arranged simultaneously One tolerable lower bound that postpones, to ensure user experience quality, then solves optimal solution as the result of decision.In order that using Lyapunov Optimization Framework, the present invention will be converted into the bar based on string stability based on time averaging deferred constraint condition Part, in this optimization problem, defines a virtual queue θ for each AS_k(t).Define virtual queue renewal equation simultaneously：

${\mathrm{\θ}}_k\left(t\right)={\mathrm{\θ}}_k(t-1)+{\mathrm{\η}}_k-1_{\{D_k\left(t\right)\≤\mathrm{\ξ}\}}$

Wherein, D_kT video average delay that k-th AS of () expression is experienced in t time slot, η_kRepresent that k-th AS can connect The service quality being subject to,Represent whether current service quality meets expection, ξ represents default video average delay,.Empty Intend cumulative error that queue is used for weighing between the overall quality of service of the actual offer of system and the service quality of user's expectation away from.

According to Lyapunov Optimization Framework, define Lyapunov equation as follows：

$U\left(t\right)=\frac{1}{2}\underset{k=1}{\overset{K}{\Σ}}{\left({\[{\mathrm{\θ}}_k\left(t\right)\]}^{+}\right)}^2$

Δ U (t)=Ε [U (t)-U (t-1) | state at time (t-1)]

U (t) is used for weighing the size of queue, and Δ (U (t)) represents the variable quantity of two neighboring time period queue array.Root According to Lyapunov Optimization Framework, can calculate within each time period according to user power utilization solicited message and system status information Go out to meet following optimization equation：

$\min F\left(R\right)={\mathrm{min\Σ}}_{c_i\∈C}{\mathrm{\Σ}}_{l_j\∈L}{\mathrm{\Φ}}_j\left(j\right)R_i^j$

s.t.(a)(b)(c)

Wherein, C represents the live video set of current active, corresponding c_iRepresent i-th video；L represents currently all Link set, corresponding l_jRepresent j-th strip link；R represents the route matrix of live video stream,Represent that i-th video is No employ j-th strip link, 0 represents and does not use, and 1 represents and uses；Φ_iJ () is that i-th video is opened in the bandwidth of j-th strip link Pin, and F (R) is according to overhead used in current route matrix R transmission video, amount of restraint (a) (b) (c) is fixed for system The constraints of justice, specially：

a)All links, or transmission i video, or do not transmit；

b)All links, can only transmit less than the data limiting equal to its bandwidth；

c)The all links of j ' ∈ I (j), can only transmit the video flowing having received.I (j) represents jth The input link set of bar link.

Afterwards using Pi-Page Rounding algorithm by nondeterministic polynomial hardly possible problem (Non Deterministic Polynomial Hard Problem) NP-Hard problem is converted into solvable problem in polynomial time.

Pi-Page Rounding has been divided into two sub-steps：

(1) convex function relaxes；Link bandwidth cost model due to definition is all convex function, is carried out using following probabilistic type Lax：

$r_i^j=P(R_i^j=1)=E\left(R_i^j\right)$

Wherein, r is referred to as probability matrix, r_i ^jRepresentProbability for 1,RepresentMathematic expectaion；

Then above-mentioned optimization problem：

$\min F\left(R\right)={\mathrm{min\Σ}}_{c_i\∈C}{\mathrm{\Σ}}_{l_j\∈L}{\mathrm{\Φ}}_i\left(j\right)R_i^j$

s.t.(a)(b)(c)

Can be converted into：

$\min F\left(r\right)={\mathrm{min\Σ}}_{c_i\∈C}{\mathrm{\Σ}}_{l_j\∈L}{\mathrm{\Φ}}_i\left(j\right)r_i^j$

s.t.(a)(b)(c)

(2) iterative step；Optimal solution r is solved first in polynomial time^*If, current r^*In there is not decimal, then Output route matrix R=r^*；If r^*In there is decimal, randomly choose a decimal, after converting thereof into 0 or 1, permissible So that overall operational overhead reduces.

Preferably, the delay D that k-th AS was experienced t-th time period_kT () must be in default tolerable scope Within, that is, meet following formula；

${\mathrm{\η}}_k\≤\underset{T\→+\mathrm{\∞}}{\lim }\frac{1}{T}\underset{t=1}{\overset{T}{\Σ}}E\[1_{\{D_k\left(t\right)\≤\mathrm{\ξ}\}}\]$

Bandwidth expense is defined as follows：

${\mathrm{\Γ}}_i\left(R\right(t\left)\right)=\underset{j=1}{\overset{n}{\Σ}}{\mathrm{\Φ}}_i(j,t)=\underset{j=1}{\overset{n}{\Σ}}\frac{{\mathrm{\ρ}}_i}{{\mathrm{\δ}}_j\left(t\right)}\·\mathrm{\φ}\left(\frac{{\mathrm{\δ}}_j\left(t\right)}{{\mathrm{\ψ}}_j\left(t\right)}\right)$

Wherein, δ_jT () represents the transmission data total amount of t j-th strip link, ψ_j(t) represent t j-th strip link can With bandwidth, ρ_iRepresent the priority level of video flowing, andThen reflect the congestion level of current ink, and function phi definition Unit bandwidth price, typically one incremental convex function.

The present invention has taken into full account different live types (the simultaneously online direct broadcast band of magnanimity, such as world cup；There is long-tail The user of effect generates live content, such as YouTube direct broadcast band), the global network information being provided using SDN and Lyapunov Optimization Framework, in QoS of customer with operator cost, using proposed by the present invention based on Pi-Page Rounding routing algorithm, finds an optimum equilibrium point, selects an optimal routed path for each video flowing While it is ensured that operation cost minimize.

Brief description

The flow chart of live video flow routing algorithm in Fig. 1 present invention.

Specific embodiment

The present invention will be further described below in conjunction with the accompanying drawings, but embodiments of the present invention are not limited to this.

Net cast platform

Net cast platform (Live Video Streaming Platform) is with television channel, webcast website, user Uploaded videos are video source, provide the video platform of real-time video services for spectators.Net cast platform is mainly responsible for collection and is regarded Frequency source, and video source is distributed to each spectators.

Net cast service provider

Net cast service provider (Live Video Streaming Service Provider) is that net cast is put down The operator of platform.By technology such as software defined network, cloud computing, cloud storage, network flow optimization, at live video stream Reason, and the video flowing after processing is distributed to spectators.

SDN centerized fusion route

Software defined network (Software Defined Networking) is the follow-on network architecture, and it is by tradition Control in the network equipment and separate with forwarding capability, global network information is obtained by the controller of a center type, using south orientation Interface OpenFlow agreement is controlled to infrastructure network, can be obviously improved the performance of network system.

Lyapunov Optimization Framework

Lyapunov Optimization Framework is used for the on-line decision algorithm that design is directed to dynamical system, and makes decision-making reach optimum While keep system mode stability.The sharpest edges of this Optimization Framework are the states not needing to know that system is following in advance Information.

Pi-Page Rounding routing algorithm

Pi-Page Rounding routing algorithm is used for solving the problems, such as to calculate video flowing transmission route in real time.Due to SDN Provide the network information of the overall situation, bring great while optimize space, also because network topology structure complicated so that in real time Calculating routing variable obtains extremely difficult.Pi-Page Rounding routing algorithm can provide polynomial time in solution party Case.

Present in net cast platform, magnanimity channel, spectators' skewness weighing apparatus, equipment complexity are various, bandwidth disappears The problems such as consume huge, in order to, on the basis of meeting service quality, reduce the operation of net cast platform operation business as much as possible Cost, the present invention proposes a kind of live broadcast stream media distribution method based on software defined network.

The present invention has taken into full account different live types (the simultaneously online direct broadcast band of magnanimity, such as world cup；There is long-tail The user of effect generates live content, such as YouTube direct broadcast band), the global network information being provided using SDN and Lyapunov Optimization Framework, in QoS of customer with operator cost, the present invention is based on Pi-Page Rounding Routing algorithm, finds an optimum equilibrium point, while selecting an optimal routed path for each video flowing, protects Card operation cost minimizes.

The basic fundamental of the present invention includes：Service quality QoS model, net cast service provider cost model and live Video flowing routing algorithm.

Service quality QoS model

In net cast platform, service quality QoS reflects the viewing experience of user indirectly.Service quality QoS is weighing apparatus One of important indicator of amount net cast platform property.

First, in order to support adaptive coding technology and differential service, save band while optimizing the experience of user Width, the present invention is the Video priority that each live video stream defines objective encoder bit rate and subjectivity.Due to video Live operator can not control AS to reach the path of user equipment, and the present invention is using the time delay between video source and AS as system QoS index.

As a rule, the delay that user experiences mainly is made up of three parts：

Video source server generates video, carries out the time delay of Video coding；

Time delay in network transmission；

In user equipment client decoding video, the time delay that carries out Video Rendering.

In the present invention, primary concern is that network delay.Network delay as the index of system service quality Q oS, such as Shown in lower formula：

${\mathrm{\η}}_k\≤\underset{T\→+\mathrm{\∞}}{\lim }\frac{1}{T}\underset{t=1}{\overset{T}{\Σ}}E\[1_{\{D_k\left(t\right)\≤\mathrm{\ξ}\}}\]$

Wherein, η_kRepresent service quality lower limit predefined in k-th AS,Represent in current time sequence In row, active service quality in k-th AS, and be defined as being received in current AS in the delay of t in k-th AS All videos average delay, shown in equation below：

$D_k\left(t\right)=\frac{1}{H_k\left(t\right)}\underset{i=1}{\overset{m}{\Σ}}{d}_k^i\left(t\right)\·{\mathrm{\ω}}_i$

Wherein ω_iDefine the priority of each video,Define the time delay in k-th AS of i-th video flowing, and H_kT () defines the number of videos that k-th AS currently receives.This formula ensure that the mistake run in whole net cast platform Cheng Zhong, the service quality that each AS receives is not less than certain default lower limit.

Net cast service provider cost model

The present invention, on the basis of ensureing service quality QoS, reduces net cast service provider further as far as possible Operation cost.In net cast platform, the operation cost of net cast service provider is exactly bandwidth cost cost.

For the different data center of location distribution, because the geographical position of distribution is different, unit bandwidth valency Lattice are also not quite similar, and unit bandwidth price fluctuates over time.Simultaneously it is contemplated that in the case of link congestion network bag Transmission wait situation occurs, indirectly increase bandwidth cost.So, in the present invention, the consumption of bandwidth depends on two Factor：Send the size of data and the jam situation of link.

In the present invention, the bandwidth cost of each video flowing is defined as follows：

${\mathrm{\Γ}}_i\left(R\right(t\left)\right)=\underset{j=1}{\overset{n}{\Σ}}{\mathrm{\Φ}}_i(j,t)=\underset{j=1}{\overset{n}{\Σ}}\frac{{\mathrm{\ρ}}_i}{{\mathrm{\δ}}_j\left(t\right)}\·\mathrm{\φ}\left(\frac{{\mathrm{\δ}}_j\left(t\right)}{{\mathrm{\ψ}}_j\left(t\right)}\right)$

Wherein, δ_jT () represents the transmission data total amount of t j-th strip link, ψ_j(t) represent t j-th strip link can With bandwidth, ρ_iRepresent the priority level of video flowing, andThen reflect the congestion level of current ink, and function phi definition Unit bandwidth price, typically one incremental convex function.

Live video flow routing algorithm

With reference to flow process Fig. 1 and embodiment, the present invention is described further.

Fig. 1 is the flow chart of the present invention, comprises the following steps that：

(S101) net cast service platform cycle of operation is cut into several time periods；

(S102) when starting each time period, current network topology information is collected by SDN bottom switch, in conjunction with The direct broadcast band information of live platform offer, user profile, collect to SDN central control unit；

(S103) service quality QoS is guaranteed using Lyapunov optimization method, utilize Page Rounding on this basis Algorithm obtains optimal solution, and there is selection on the road that its result of decision includes each net cast channel；

(S104) making routing decisions are made to each net cast channel based on the result of decision obtaining, by SDN Southbound interface sends decision information to each SDN switch, and real-time reception is from the viewing feedback information of user simultaneously.

In a detailed embodiment, the status information of systematic collection includes：The code rate information of video flowing, user's request Information, bandwidth capability information, unit bandwidth pricing information etc..

In a detailed embodiment, the present invention can by minimize operation cost, optimization service quality excellent Change problem is converted into Lyapunov optimization problem, using the status information of systematic collection as the known conditions of this optimization problem, incites somebody to action Service quality QoS index (time delay) is as constraints, and arranges a tolerable delay lower bound to ensure user's body simultaneously The amount of checking the quality, then solves optimal solution as the result of decision.In order that using Lyapunov Optimization Framework, the present invention will be put down based on the time Equal deferred constraint condition is converted into the condition based on string stability, in this optimization problem, defines one for each AS Individual virtual queue θ_k(t).Define virtual queue renewal equation simultaneously：

${\mathrm{\θ}}_k\left(t\right)={\mathrm{\θ}}_k(1-1)+{\mathrm{\η}}_k-1_{\{D_k\left(t\right)\≤\mathrm{\ξ}\}}$

Wherein, D_kT video average delay that k-th AS of () expression is experienced in t time slot, η_kRepresent that k-th AS can connect The service quality being subject to,Represent whether current service quality meets expection.Virtual queue is used for weighing that system is actual to be carried For overall quality of service and user expect service quality between cumulative error away from.

According to Lyapunov Optimization Framework, define Lyapunov equation as follows：

$U\left(t\right)=\frac{1}{2}\underset{k=1}{\overset{K}{\Σ}}{\left({\[{\mathrm{\θ}}_k\left(t\right)\]}^{+}\right)}^2$

Δ U (t)=Ε [U (t)-U (t-1) | state at time (t-1)]

U (t) is used for weighing the size of queue, and Δ (U (t)) represents the variable quantity of two neighboring time period queue array.Root According to Lyapunov Optimization Framework, can calculate within each time period according to user power utilization solicited message and system status information Go out to meet following optimization equation：

$\min F\left(R\right)={\mathrm{min\Σ}}_{c_i\∈C}{\mathrm{\Σ}}_{l_j\∈L}{\mathrm{\Φ}}_j\left(j\right)R_i^j$

s.t.(a)(b)(c)

Wherein, C represents the live video set of current active, corresponding c_iRepresent i-th video；L represents currently all Link set, corresponding l_jRepresent j-th strip link；R represents the route matrix of live video stream,Represent that i-th video is No employ j-th strip link, 0 represents and does not use, and 1 represents and uses；Φ_iJ () is that i-th video is opened in the bandwidth of j-th strip link Pin, and F (R) is according to overhead used in current route matrix R transmission video, amount of restraint (a) (b) (c) is fixed for system The constraints of justice, specially：

a)All links, or transmission i video, or do not transmit；

b)All links, can only transmit less than the data limiting equal to its bandwidth；

c)The all links of j ' ∈ I (j), can only transmit the video flowing having received.I (j) represents jth The input link set of bar link.

The false code of this algorithmic procedure is as follows.

For step 5 required majorization of solutions target in above-mentioned algorithm, can be further by execution proposed by the present invention In the Pi-Page Rounding routing algorithm of each time slot, solve optimal solution in polynomial time, as follows.

The present invention proposes a kind of live broadcast stream media distribution method based on software defined network, and the present invention describes in detail Live video stream distributing method.The present invention can have multiple methods in the specific implementation, including but not limited to：

1st, it is directed to the characteristic of net cast platform, obtain network topological information using software defined network, and collect video Stream information, user request information, in order to support adaptive coding technology and differential service, define preferentially for each video flowing Level；

2nd, it is directed to different live types, adaptive change can be carried out by adjusting threshold value, and can be for difference The user in area carries out personal settings；

In the present invention, the structure of each module and connected mode all can be varied from, in technical solution of the present invention On the basis of, all improvement structure of indivedual algoritic modules being carried out according to the principle of the invention and equivalents, all should not exclude Outside protection scope of the present invention.

Crucial mathematical model in working-flow proposed by the present invention and method have：Service quality QoS model, video Direct broadcast service operator cost model and live video flow routing algorithm.The service quality QoS model tormulation globality of system Can, and net cast platform operation business's cost model have expressed system expense in running, both bases of the present invention Plinth.Proposed by the present invention based on live video flow routing algorithm can according to limited system status information, AS postpone etc. information, move Make to state video flowing routing decision, on the basis of ensureing service quality QoS, optimization of video live platform operation business always opens Pin.Live video flow routing algorithm proposed by the present invention is the core content of the present invention.

The embodiment of invention described above, does not constitute limiting the scope of the present invention.Any at this Done modification, equivalent and improvement etc. within bright spiritual principles, should be included in the claim protection of the present invention Within the scope of.

Claims (4)

1. a kind of live video stream media distribution method based on SDN is it is characterised in that comprise the following steps：

S1. net cast service platform cycle of operation is cut into several time periods；

S2. when starting each time period, current network topology information is collected by SDN bottom switch, in conjunction with live flat The direct broadcast band information of platform offer, user profile, collect to SDN central control unit；

S3. service quality QoS is guaranteed using Lyapunov optimization method, utilize Pi-Page Rounding algorithm on this basis Obtain the result of decision, its result of decision includes the Route Selection of each net cast channel；

S4. making routing decisions are made to each net cast channel based on the result of decision obtaining, by SDN southbound interface Decision information is sent to each SDN switch, real-time reception is from the viewing feedback letter of user simultaneously；To realize live video stream Distribution of media.

2. method according to claim 1 is it is characterised in that the mode solving the acquisition result of decision in step S3 is concrete For：The mode form of minimum operation cost, the target mathematics of maximization system service quality Q oS is turned to one solve Optimization problem；Current information collected by SDN central control unit, using network bandwidth information as constraints, using Lyapunov Optimum Theory provides system service quality level；Then go out the result of decision using Pi-Page Rounding Algorithm for Solving；

First, a virtual queue θ defined in each user access point AS (k)_k(t)；Define virtual queue renewal side simultaneously Journey：

${\mathrm{\θ}}_k\left(t\right)={\mathrm{\θ}}_k(t-1)+{\mathrm{\η}}_k-1_{\{D_k\left(t\right)\≤\mathrm{\ξ}\}}$

Wherein, θ_kT k-th AS of () expression is in the virtual queue of t time slot, θ_k(t-1) k-th AS of expression is in the void of t-1 time slot Intend queue, η_kRepresent the receptible service quality of k-th AS,Represent whether current service quality meets expection, D_k T video average delay that k-th AS of () expression is experienced in t time slot, ξ represents default video average delay, virtual team Arrange cumulative error between the service quality that overall quality of service for weighing the actual offer of system and user expect away from；

According to Lyapunov Optimization Framework, define U (t) and Δ U (t) is as follows：

$U\left(t\right)=\frac{1}{2}\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{\left({\[{\mathrm{\θ}}_k\left(t\right)\]}^{+}\right)}^2$

Δ U (t)=Ε [U (t)-U (t-1) | state at time (t-1)]

U (t) is used for weighing the size of queue, and Δ (U (t)) represents the variable quantity of two neighboring time period queue array；According to Lyapunov Optimization Framework, calculates satisfaction according to user request information and system status information as follows within each time period Optimization equation：

$\begin{array}{c}\min F\left(R\right)={\mathrm{min\Σ}}_{c_i\∈C}{\mathrm{\Σ}}_{l_j\∈L}{\mathrm{\Φ}}_i\left(j\right)R_i^j\\ s.t.\left(a\right)\left(b\right)\left(c\right)\end{array}$

Wherein, C represents the live video set of current active, corresponding c_iRepresent i-th video；L represents currently all of link Set, corresponding l_jRepresent j-th strip link；R represents the route matrix of live video stream,Represent whether i-th video uses J-th strip link, 0 represents and does not use, and 1 represents and uses；Φ_iJ () is i-th video bandwidth cost in j-th strip link, and F (R) it is according to overhead, the pact that amount of restraint (a) (b) (c) defines for system used in current route matrix R transmission video Bundle condition, specially：

A) for all links,Represent this link transmission i video or do not transmit；

B) for all links,Expression can only transmit less than the data limiting equal to its bandwidth；Its Middle Ψ (j, t) represents j-th strip link in the available bandwidth of t time slot, ρ_iThe code check size of i-th video respectively, m represents video Sum；

C) for all links,Expression can only transmit the video flowing having received, and I (j) represents The input link set of j-th strip link, j ' represents a certain bar input link of j-th strip link；

Using Pi-Page Rounding algorithm, nondeterministic polynomial hardly possible problem is converted in polynomial time solving afterwards Problem；

Pi-Page Rounding has been divided into two sub-steps：

(1) convex function relaxes；Link bandwidth cost model due to definition is all convex function, carries out pine using following probabilistic type Relax：

$r_i^j=P(R_i^j=1)=E\left(R_i^j\right)$

Wherein, r is referred to as probability matrix,RepresentProbability for 1,RepresentMathematic expectaion；

Then above-mentioned optimization problem：

$\begin{array}{c}\min F\left(R\right)={\mathrm{min\Σ}}_{c_i\∈C}{\mathrm{\Σ}}_{l_j\∈L}{\mathrm{\Φ}}_i\left(j\right)R_i^j\\ s.t.\left(a\right)\left(b\right)\left(c\right)\end{array}$

Can be converted into：

$\begin{array}{c}\min F\left(r\right)={\mathrm{min\Σ}}_{c_i\∈C}{\mathrm{\Σ}}_{l_j\∈L}{\mathrm{\Φ}}_i\left(j\right)r_i^j\\ s.t.\left(a\right)\left(b\right)\left(c\right)\end{array}$

(2) iterative step；Optimal solution r is solved first in polynomial time^*If, current r^*In there is not decimal, then export Route matrix R=r^*；If r^*In there is decimal, randomly choose a decimal, after converting thereof into 0 or 1, enable to Overall operational overhead reduces.

3. method according to claim 1 is it is characterised in that by the video receiving time delay as passing judgment on service quality Standard, the video average delay D that k-th AS was experienced t-th time period_k(t) must default tolerable scope it Interior, that is, meet following formula；

${\mathrm{\η}}_k\≤\underset{T\→+\mathrm{\∞}}{\lim }\frac{1}{T}\underset{t=1}{\overset{T}{\mathrm{\Σ}}}E\[1_{\{D_k\left(t\right)\≤\mathrm{\ξ}\}}\]$

Wherein, T represents the total duration of net cast platform operation, levels off to infinity, η_kRepresent default minimum in k-th AS Service quality, ξ is default delay threshold, in this invention, Video service time delay is less than the service of ξ, clothes referred to as preferably Business,Illustrate and currently have that how many service in k-th AS has reached standard ξ.

4. method according to claim 1 is it is characterised in that in order to ensure while providing certain service quality, to the greatest extent Possibly reduce bandwidth operational overhead, bandwidth expense Γ_i(R (t)) is defined as follows：

${\mathrm{\Γ}}_i\left(R\right(t\left)\right)=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\Φ}}_i(j,t)=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}\frac{{\mathrm{\ρ}}_i}{{\mathrm{\δ}}_j\left(t\right)}\·\mathrm{\φ}\left(\frac{{\mathrm{\δ}}_j\left(t\right)}{{\mathrm{\ψ}}_j\left(t\right)}\right)$

Wherein, δ_jT () represents the transmission data total amount of t j-th strip link, ψ_jT () represents the available band of t j-th strip link Width, ρ_iRepresent the code check size of i-th video, andThen reflect the congestion level of current ink, and function phi definition Unit bandwidth price, is an incremental convex function.