CN102136998B - Traffic engineering and server selection joint optimization method, system and related equipment - Google Patents

Abstract

The embodiment of the invention provides a traffic engineering and server selection joint optimization method, a traffic engineering and server selection joint optimization system and related equipment. An optimal solution is calculated by utilizing a small amount of information in a relatively shorter convergence time on the premise of not disclosing key bottom-layer network information to a content provider (CP) by an instant service provider (ISP). The method comprises that: a router determines each outgoing link bearing traffic belonging to the same user node; the router acquires optimal link weights of each outgoing link; and the router acquires determines a division strategy of the traffic belonging to the same user node on any outgoing link (u,v) in the outgoing links according to the optimal link weights and the traffic belonging to the same user node. In the method, the system and the related equipment, calculation is not required to be performed for a plurality of turns, so the amount of transmitted information is greatly reduced, the convergence time of the system is shortened, time consumption is low generally, and the traffic division determining efficiency is higher.

Description

Combined optimization method, system and relevant device that traffic engineering and server are selected

Technical field

The present invention relates to the communications field, relate in particular to combined optimization method, system and relevant device that traffic engineering and server are selected.

Background technology

ISP (ISP, Internet Service Provider) relies on its physical network having to provide Internet to connect.Owing to there being Duo Tiao road warp between source node in network topology and destination node, therefore, in the time that ISP planning sends certain flow to certain destination node from certain source node, it need to determine how to be distributed in the flow on every paths, thereby (for example make the overall performance of network, the aspects such as delay, load balancing) optimum, the problems referred to above that ISP need to solve are called traffic engineering (TE, Traffic Engineering) problem.

And content supplier (CP, Content Provider) utilize the Internet connect to its user provide content aspect service for example file, Voice & Video share etc.Different from ISP, CP does not have authority to go to change the route of bottom-layer network.Corresponding each user's request, CP need to determine how the assignment of traffic of user's request is arrived to different servers, the problems referred to above that CP need to solve are called server and select (SS, Server Selection) problem.

Usually, ISP and CP solve respectively TE and SS independently, do not exist between the two can be mutually shared information.Because TE and SS can affect network condition, ISP and CP need constantly to adjust TE and SS, until system reaches stable state.As previously mentioned, because ISP does not generally have exchanging of information between the two with CP, therefore, under uncooperative scene, CP need to rely on the end-to-end detection that waits bottom-layer network information is solved to SS; Under partial cooperation scene, ISP provides bottom-layer network information accurately, such as topology information, link-state information and routing iinformation etc. to CP.But because ISP always exists inconsistently to the optimization aim of SS to TE and CP, this makes both sides all wish to stand in the target of optimization self in oneself position.Therefore, no matter be under uncooperative scene or under partial cooperation scene, this non-cooperative game mechanism of TE and SS can only guarantee that system converges to the balance of suboptimum, systematic function cannot reach Pareto (Pareto) optimum.

In order to guarantee that system converges to Pareto optimum, ISP and CP need to realize the combined optimization between TE and SS.In combined optimization, ISP and CP realize Complete Information and share, and ISP provides network topological information, link-state information and the routing iinformation etc. of oneself to CP, and CP provides oneself server info, user's request information etc. to ISP.Both realize Complete Information share time, TE and SS are no longer independently solved.Combined optimization carries out for TE and SS simultaneously, rather than takes a kind of mode (be that ISP carries out TE one time, CP carries out SS again a time, so circulation) of iteration.This combined optimization is to a lot of application, and such as demand (telecommunication) service etc., all can obtain good effect.

But the challenge that the combined optimization of above-mentioned TE and SS faces is how to design the agreement of a Fast Convergent, when guaranteeing that system obtains optimal performance, expose as few as possible ISP and CP information each other.For example, ISP exposes network topological information, link-state information and routing iinformation etc. as few as possible, and CP exposes server info, user's request information etc. as few as possible.

The problem facing for the combined optimization of above-mentioned TE and SS, industry has proposed associating SS and TE(COST, COoperative SS and TE) agreement.It is two sub-optimization problem and main optimization problems that are similar to TE and SS by initial combined optimization PROBLEM DECOMPOSITION that COST utilizes Duality Decomposition.Interrelated by common dual variable between two sub-optimization problems, main optimization problem is responsible for constantly updating associated dual variable so that system is approached optimal solution.In the time solving the combined optimization problem of TE and SS, the dual variable of given association, ISP solves the sub-optimization problem TE-NBS of similar TE, and CP solves the sub-optimization problem SS-NBS of similar SS.After each takes turns optimization, according to the solution of two sub-optimization problems, system is upgraded associated dual variable by solving main optimization problem, for the combined optimization of next round.Through after the abundant optimization of taking turns, COST guarantees to obtain optimum systematic function.

In the time of specific implementation, ISP can utilize existing technology to solve the sub-optimization problem TE-NBS of similar TE.Owing to requiring to obtain the optimal solution of sub-optimization problem, the technology that can adopt comprises the multiprotocol label switching (MPLS that has applied centralized algorithm, Multi-Protocol Label Switching) and the flow by index punishment of having applied distributed algorithm cut apart (PEFT, Penalizing Exponential Flow-spliTting).Meanwhile, CP need to obtain bottom-layer network information accurately and solve (comprising topology and time delay) the subproblem SS-NBS of similar SS.

Because COST need to be through the optimization of too much wheel, and take turns ISP in optimizing process at each and need to solve the similar sub-optimization problem TE-NBS with TE, CP need to solve the similar sub-optimization problem SS-NBS with SS, and be to realize communication by the price of each link in renewal network between the two, compared to existing TE agreement, the time consumption of COST be doubled and redoubled (the wheel number that is proportional to optimization), further, because not only comprising each, the required information transmission of COST takes turns the information transmission that solves two sub-optimization problems in optimization, while also comprising the main optimization problem of solution, upgrade the required information transmission of associated dual variable.In other words, COST needs the convergence time grown and a large amount of information transmission in the process that reaches optimal solution.

In addition, COST, in the time solving SS optimization problem, requires ISP to provide crucial bottom-layer network information (such as network topological information, link-state information etc.) to CP.But many times, such requirement is for ISP unreasonable, especially in the time that ISP occupies leading position in cooperation; In the time comprising multiple CP in network, ISP is also unwilling to leak crucial bottom-layer network information to each CP.

Summary of the invention

Combined optimization method, system and relevant device that the embodiment of the present invention provides traffic engineering and server to select, leaking to CP without ISP under the prerequisite of crucial bottom-layer network information, tries to achieve optimal solution with shorter convergence time and a small amount of information transmission.

The combined optimization method that the embodiment of the present invention provides a kind of traffic engineering and server to select, comprising: router determines that carrying belongs to each outside link of the flow of same user node, described router obtains the optimum link weight of described each outside link, and described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node, described router calculates described router through any outside link (u, v) to the shortest path length of described user node and the extremely difference of the shortest path length of described user node of described router, according to described difference and the extremely equivalent number of described user node of described router, described in determining, belong to flow any outside link (u in each outside link of same user node, v) the ration of division going up, described in calculating, belong to the flow of same user node and the product of the described ration of division, described in obtaining, belong to flow any outside link (u in described each outside link of same user node, v) the flow of cutting apart of going up, wherein, u represents the node at described router place, v represents to form described outside link (u, v) another node.

The combined optimization method that the embodiment of the present invention provides a kind of traffic engineering and server to select, comprising: server obtains shortest path length and the described router extremely equivalent number of described user node of described router to user node from connected router; Described server obtains the optimum link weight of each virtual link, described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node, and described virtual link is that the described server of the described dummy node that increases in network topology forms; Described server calculates described dummy node through any virtual link (Ns, a s _i) to the shortest path length of described user node and the extremely difference of the shortest path length of described user node of described dummy node, according to described equivalent number and described difference, determine described total flow demand any virtual link (Ns, s in each virtual link _i) on the ration of division, calculate the product of described total flow demand and the described ration of division, obtain described total flow demand with described any virtual link (Ns, a s _i) distribute to the traffic demand of described user node on corresponding server, wherein, Ns represents described dummy node, s _irepresent to form described virtual link (Ns, s _i) another node.

The combined optimization method that the embodiment of the present invention provides a kind of traffic engineering and server to select, comprise: by increase dummy node in network topology, each server node in described network topology to the traffic demand of user node is converted into the virtual flow of described dummy node to described user node, described virtual flow is constantly equal to the total flow demand of described user node, between each server node in described dummy node and described network topology, forms virtual link;

Take the constant of described virtual flow user's request bound term in many article network flow MCF problem, solve described MCF problem and obtain the optimal flux on each link of described network topology, each link of described network topology comprises physical link and the described virtual link in described network topology, and in described MCF problem, user's request bound term is:

$\underset{(s,v)\∈E}{\mathrm{\Σ}}{f}_{s,v}^t-\underset{(u,s)\∈E}{\mathrm{\Σ}}{f}_{u,s}^t=D(s,t),\∀s\∈V,t\∈V,$ Wherein, (s, v) ∈ E represents network topological diagram G (V, E) the oriented link from server node s to node v in, (u, s) ∈ E represents the oriented link from node u to server node s, V and E represent respectively node set and link set

represent oriented link (s, belongs to the traffic demand of user node t on v),

represent to belong on oriented link (u, s) traffic demand of user node t, D (s, t) represents the traffic demand from server node s to user node t, and described D (s, t) is the constant of described user's request bound term;

Solve network entropy maximization NEMR problem according to the optimal flux on described link, obtain the optimum link weight on described virtual link and physical link; Described optimum link weight is distributed to the router node in described network topology.

The embodiment of the present invention provides a kind of router, comprising: outwards link determination module, for determining that carrying belongs to each outside link of the flow of same user node; Acquisition module, for obtaining the optimum link weight of each definite outside link of described outside link determination module, described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node; The first difference computational unit, be used for calculating described router through any outside link (u, v) to the shortest path length of described user node and the extremely difference of the shortest path length of described user node of described router, wherein, u represents the node at described router place, v represents to form described outside link (u, another node v); The first ration of division determining unit, be used for according to described router to the equivalent number of described user node and the difference that described the first difference computational unit is tried to achieve, described in determining, belong to flow any an outside link (u, the ration of division v) going up in described each outside link of same user node; Cut apart flow rate calculation unit, described in calculating, belong to the flow of same user node and the product of the described ration of division, described in obtaining, belong to flow any outside link (u, flow of cutting apart of v) going up in described each outside link of same user node.

The embodiment of the present invention provides a kind of server, comprising: the first acquisition module, for obtain shortest path length and the described router extremely equivalent number of described user node of described router to user node from the router being connected with described server; The second acquisition module, for obtaining the optimum link weight of each virtual link, described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node Ns, and described virtual link is that the described dummy node and the described server that in network topology, increase form; The second difference computational unit, for calculating described dummy node through any virtual link (Ns, a s _i) to shortest path length and the extremely difference of the shortest path length of described user node of described dummy node of described user node, wherein, Ns represents described dummy node, s _irepresent to form described virtual link (Ns, s _i) another node; The second ration of division determining unit, for calculating gained difference according to described router to equivalent number and described second difference computational unit of described user node, determines described total flow demand any virtual link (Ns, s in described each virtual link _i) on the ration of division; Traffic demand computing unit, for calculating the product of described total flow demand and the described ration of division, obtain described total flow demand with described any virtual link (Ns, a s _i) traffic demand of distributing to described user node on corresponding server.

The embodiment of the present invention provides a kind of optimum link Weight Acquisition device, comprise: conversion module, be used for by increase dummy node in network topology, each server node in described network topology to the traffic demand of user node is converted into the virtual flow of described dummy node to described user node, described virtual flow is constantly equal to the total flow demand of described user node, between each server node in described dummy node and described network topology, forms virtual link; Optimal flux is asked for module, for the constant take described virtual flow as many article network flow MCF problem user's request bound term, solve the optimal flux that described MCF problem obtains described each link of network topology, each link of described network topology comprises physical link and the described virtual link in described network topology, and in described MCF problem, user's request bound term is:

$\underset{(s,v)\∈E}{\mathrm{\Σ}}{f}_{s,v}^t-\underset{(u,s)\∈E}{\mathrm{\Σ}}{f}_{u,s}^t=D(s,t),\∀s\∈V,t\∈V,$ Wherein, (s, v) ∈ E represents network topological diagram G (V, E) the oriented link from server node s to node v in, (u, s) ∈ E represents the oriented link from node u to server node s, and V and E represent respectively network topology G (V, E) node set on and link set

represent oriented link (s, belongs to the traffic demand of user node t on v), represent to belong on oriented link (u, s) traffic demand of user node t, D (s, t) represents the traffic demand from server node s to user node t, and described D (s, t) is the constant of described user's request bound term; Optimum link weight is asked for module, solves network entropy maximization NEMR problem for ask for the optimal flux that module asks for according to the optimal flux on described link, obtains the optimum link weight on described virtual link and physical link; Distribution module, for being distributed to the router node in described network topology by described optimum link weight.

The combined optimization system that the embodiment of the present invention provides a kind of traffic engineering and server to select, comprising: optimum link Weight Acquisition device, server and router, described optimum link Weight Acquisition device, be used for by increase dummy node in network topology, each server node in described network topology to the traffic demand of user node is converted into the virtual flow of described dummy node to described user node, described virtual flow is constantly equal to the total flow demand of described user node, between each server node in described dummy node and described network topology, form virtual link, take the constant of described virtual flow user's request bound term in many article network flow MCF problem, solve the optimal flux that described MCF problem obtains described each link of network topology, each link of described network topology comprises physical link and the described virtual link in described network topology, in described MCF problem, user's request bound term is:

$\underset{(s,v)\∈E}{\mathrm{\Σ}}{f}_{s,v}^t-\underset{(u,s)\∈E}{\mathrm{\Σ}}{f}_{u,s}^t=D(s,t),\∀s\∈V,t\∈V,$ Wherein, (s, v) ∈ E represents network topological diagram G (V, E) the oriented link from server node s to node v in, (u, s) ∈ E represents the oriented link from node u to server node s, V and E represent respectively node set and link set

represent oriented link (s, belongs to the traffic demand of user node t on v),

represent oriented link (u, s) on, belong to the traffic demand of user node t, D (s, t) represent the traffic demand from server node s to user node t, described D (s, t) is the constant of described user's request bound term, solves network entropy maximization NEMR problem according to the optimal flux on described link, obtain the optimum link weight on described virtual link and physical link, described optimum link weight is distributed to the router node in described network topology, described server, for obtain shortest path length and the described router extremely equivalent number of described user node of described router to user node from connected router, obtain the optimum link weight of each virtual link, described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node, described virtual link is that the described dummy node and the described server that in network topology, increase form, according to described optimum link weight, described shortest path length, described router is to the equivalent number and the total flow demand that belongs to described user node of described user node, determine the traffic demand of distributing to described user node on described server, described router, for determining that carrying belongs to each outside link of the flow of same user node, obtain the optimum link weight of described each outside link, described optimum link weight obtains by the network entropy maximization problem NEMR that solves the network topology that has increased dummy node, according to described optimum link weight and described in belong to the flow of same user node, described in determining, belong to flow any outside link (u in described each outside link of same user node, v) the segmentation strategy of going up, wherein, u represents the node at described router place, v represents to form described outside link (u, v) another node.

From the invention described above embodiment, router is by obtaining the optimum link weight of its each outside link, just can determine according to this optimum link weight and the flow that belongs to same user node the flow segmentation strategy on any outside link in each outside link that belongs to same user node.Therefore, the router in above-described embodiment is equivalent to only need the sub-optimization problem of a similar TE of solution, compared with prior art, do not need the calculating of many wheels, greatly reduced the transmission capacity of information, accelerated the convergence time of system, still less consuming time generally, determine that the efficiency that flow cuts apart is also higher.

Accompanying drawing explanation

In order to be illustrated more clearly in the technical scheme of the embodiment of the present invention, to the accompanying drawing of required use in prior art or embodiment description be briefly described below, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skills, do not paying under the prerequisite of creative work, can also obtain as these accompanying drawings other accompanying drawing.

Fig. 1 is the combined optimization method schematic flow sheet that the traffic engineering that provides of the embodiment of the present invention and server are selected;

Fig. 2 is oriented link schematic diagram in the topological network figure that provides of the embodiment of the present invention;

Fig. 3 is the path schematic diagram of the node u that provides of the embodiment of the present invention to user node t;

Fig. 4 is the combined optimization method schematic flow sheet that the traffic engineering that provides of another embodiment of the present invention and server are selected;

Network topology schematic diagram after the increase dummy node that Fig. 5 embodiment of the present invention provides;

Fig. 6 is the combined optimization method schematic flow sheet that the traffic engineering that provides of another embodiment of the present invention and server are selected;

Fig. 7 is a kind of router logic structural representation that the embodiment of the present invention provides;

Fig. 8 is a kind of router logic structural representation that another embodiment of the present invention provides;

Fig. 9 is a kind of server logic structural representation that the embodiment of the present invention provides;

Figure 10 is a kind of server logic structural representation that another embodiment of the present invention provides;

Figure 11 is the optimum link Weight Acquisition device logical construction schematic diagram that the embodiment of the present invention provides;

Figure 12 is the combined optimization system logical construction schematic diagram that a kind of traffic engineering of providing of the embodiment of the present invention and server are selected.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is only the present invention's part embodiment, rather than whole embodiment.Based on the embodiment in the present invention, those of ordinary skills, not making the every other embodiment obtaining under creative work prerequisite, belong to the scope of protection of the invention.

Referring to accompanying drawing 1, is the combined optimization method schematic flow sheet that a kind of traffic engineering of providing of the embodiment of the present invention and server are selected, and mainly comprises step:

S101, router determines that carrying belongs to each outside link of the flow of same user node.

So-called outwards link (Outgoing Link), at directed graph G(V, E) (for example, network topological diagram etc., wherein, V represents the node set in directed graph, and E represents the link set in directed graph) in can be so a kind of link: for any node, if flow is the neighbor node that flows out and flow into this node from this node, the link between this node and such neighbor node is exactly the outside link of this node.As shown in Figure 2, if the use direction of arrow represents the flow direction of data (flow), node u and neighbor node v ₁, neighbor node v ₂..., neighbor node v _noriented link (the v forming ₁, u), oriented link (u, v ₂) ... with oriented link (u, v _n) in, because data are to flow out from node u, from neighbor node v ₂with neighbor node v _nflow into oriented link (u, v ₂) and oriented link (u, v _n) be exactly the outside link of node u, and oriented link (v ₁, u) be not the outside link of node u.

In network topological diagram, for given router (being a node in network topological diagram), the link being made up of this router has many.Therefore, need to determine which is the outside link that carrying belongs to the flow of same user node.In embodiments of the present invention, router can determine that carrying belongs to each outside link of the flow of same user node by searching the network topological informations such as routing table.

S102, router obtains the optimum link weight of each outside link.

In embodiments of the present invention, optimum link weight obtains by solving the network entropy maximization problem (NEMR, Network Entropy Maximization Relaxed) of the network topology that has increased dummy node.Solving of NEMR problem can be consulted the below embodiment shown in Fig. 4.The solution of NEMR problem is the optimum link weight of each link (comprising outside link) in network topological diagram, and this optimum link weight can be distributed to each router, and router obtains the optimum link weight of each outside link thus.

Outwards the optimum link weight of link can also have other to solve and ways of distribution, is not limited to the embodiment shown in Fig. 4 of the present invention.

S103, according to optimum link weight and the flow that belongs to same user node, determines the flow any an outside link (u, the segmentation strategy v) gone up in each outside link that belong to same user node.

So-called user node, can be the final purpose node of flow, for example, and the terminal of PC in real network (Personal Computer, PC), mobile phone and so on; Also can be general destination node, such node have converged the flow of distributing to the some users of lower floor.

Represent actual router with the node u in network topological diagram below, node t represents user node, the discrepancy node-flow that the direction of arrow represents data (flow) to, illustrate that router how to confirm belongs to flow any an outside link (u, the segmentation strategy v) gone up in each outside link of same user node t.

As shown in Figure 3, be the path schematic diagram of node u to user node t.Suppose that any path that arrives user node t through node v is the path u-v-c-...-t in figure in these paths.According to explanation above, (u, is exactly v) an outside link of node u to link, and node v is link (u, another node v).Router is easy to obtain it, and through outside link, (u, v) to the shortest path of user node t, supposes that this shortest path is u-v-c-d-t; Further, router also can obtain the length of shortest path u-v-c-d-t, and through any outside link, (u to the shortest path length of user node t is v) note router

here, w (u, v) represents that (this link weight is also the optimal solution of described NEMR problem, i.e. optimum link weight above to link for u, link weight (being also the length of link) v).Similarly, router is also easy to obtain its shortest path to user node t and (it will be appreciated by those skilled in the art that, this shortest path may be also that router is through outside link (u, v) to the shortest path of user node t, for example, the path u-v-c-d-t of the present embodiment), suppose that this shortest path is u-e-f-t; Further, router also can obtain the length of this shortest path u-e-f-t, and note router to the shortest path length of user node t is

then, calculate

with

difference, this difference is used

represent,

$h_{u,v}^t=d_v^t+w(u,v)-d_u^t.$

Then, according to difference

with the equivalent number of router to user node t, determine the flow above-mentioned any an outside link (u, the ration of division v) going up in each outside link that belong to same user node t.Below provide an embodiment who calculates this ration of division.

S1, calculates the equivalent number of router and the product of an attenuation parameter, and this attenuation parameter is the difference of natural Exponents e

the inverse of inferior power, calculates (u, flow segmentation function are v) used to obtain above-mentioned any outside link

represent, wherein,

be the equivalent number of router to user node t, its value is defined as

in this expression formula,

the i paths length from node v to user node t,

the shortest path length of node v to user node t.Flow segmentation function

with above-mentioned any outside link (u, v) correspondence.

It should be noted that, router can be under the request of server, and it is sent to coupled server to the shortest path length of user node t with to the equivalent number of user node t.

S2, calculate flow segmentation function that each outside link is corresponding and.

(u, flow segmentation function is v) similar, can use identical method to calculate the flow segmentation function that each outside link is corresponding, calculates with calculating outside link in S1

wherein, (u, j) ∈ E represents any outside link (u, j) in the outside link set of node u, and then summation, calculates

${\mathrm{\Σ}}_{(u,j)\∈E}\mathrm{\Γ}\left(h_{u,j}^t\right).$

S3, asks for flow segmentation function

with

ratio, calculate $\frac{\mathrm{\Γ}\left(h_{u,v}^t\right)}{{\mathrm{\Σ}}_{(u,j)\∈E}\mathrm{\Γ}\left(h_{u,j}^t\right).}$

By the calculating of S3, the flow that obtains belonging to same user node t in each outside link of node u any outside link (u after the ration of division v) going up, calculates and belongs to the flow of same user node t and the product of the ration of division, be i.e. calculating

in this expression formula,

represent to flow to from node u the flow of same user node t.Belong to flow any outside link (u, the flow of cutting apart of v) going up use in above-mentioned each outside link of same user node

represent,

In embodiment described in Fig. 1, router is by obtaining the optimum link weight of its each outside link, just can determine according to this optimum link weight and the flow that belongs to same user node the flow segmentation strategy on any outside link in each outside link that belongs to same user node.Therefore, the router in above-described embodiment is equivalent to only need the sub-optimization problem of a similar TE of solution, compared with prior art, do not need the calculating of many wheels, greatly reduced the transmission capacity of information, accelerated the convergence time of system, still less consuming time generally, determine that the efficiency that flow cuts apart is also higher.

In real network, because the flow forwarding from a node A is after several intermediate nodes forward, this flow is likely got back to again node A, that is, flow may meet with " Routing Loop " when route in network.For fear of occurring Routing Loop, in embodiments of the present invention, while selecting outside link, preferentially select so outside link, each node to the shortest path length of user node t in each outside link is less than router (node is u) to the shortest path length of user node t.That is to say, each router (u) locate, and flow is always being cut apart flow from destination node on those nodes close to more by node.Therefore, in embodiments of the present invention, any outside link in each outside link (u, v) go up flow segmentation function and be expressed as:

$\mathrm{\Γ}\left(h_{u,v}^t\right)=\left\{\begin{array}{cc}{e}^{-h_{u,v}^t}{\mathrm{\γ}}_{v,}^t\mathrm{if}& d_u^t>d_v^t\\ 0,& \mathrm{otherwise}\end{array}\right.$

Referring to accompanying drawing 4, is the combined optimization method schematic flow sheet that a kind of traffic engineering of providing of another embodiment of the present invention and server are selected, and mainly comprises step:

S401, by increase dummy node in network topology, is converted into the virtual flow of dummy node to user node by each server node in network topology to the traffic demand of user node.

In the present embodiment, the dummy node of increase can and network topology on each server node between form virtual link.As shown in Figure 5, when having increased after dummy node Ns, each server node (s ₁, s ₂.., s _n) equal dummy node Ns directly to the virtual flow of user node t to traffic demand (Traffic Demand) sum of user node t, use D (Ns, t) to represent.So-called virtual flow, refers to the flow in fact not existing from dummy node Ns to user node t, just increases after dummy node Ns, and for the needs of dealing with problems, a flow of supposing.Although for user node t, each (s _i, t) to (wherein, i be 1,2 ..., traffic demand on n) changes, but each (s _i, be t) the total flow demand that is constantly equal to user node t to upper traffic demand sum, therefore, dummy node Ns directly to the virtual flow of user node t be a constant that is constantly equal to the total flow demand of user node t.

S402, the constant of user's request bound term in (MCF, Multi-Commodity Flow) problem take virtual flow as many article network flow, solves MCF problem and obtains the optimal flux of each link of network topology.

So-called MCF problem refers to the problem that uses following expression formula to represent:

$\min \underset{(u,v)\∈E}{\mathrm{\Σ}}\mathrm{\Φ}\left(f_{u,v,}{c}_{u,v}\right)$

Constraints:

$\left(1\right),f_{u,v}=\underset{t\∈V}{\mathrm{\Σ}}{f}_{u,v,}^t\∀(u,v)\∈E$

$\left(2\right),f_{u,v}\≤c_{u,v,}\∀(u,v)\∈E$

$\left(3\right),\underset{(s,v)\∈E}{\mathrm{\Σ}}{f}_{s,v}^t-\underset{(u,s)\∈E}{\mathrm{\Σ}}{f}_{u,s}^t=D(s,t),\∀s\∈V,t\∈V$

Variable: $f_{u,v}^t\≥0$

In MCF problem, Φ (f _u,v, c _u,v) represent link (u, v) cost, it is link flow (being also link load) f _u,vwith link capacity c _u,vfunction.Conventionally elect link cost Φ as convex function, for example, the cost of link is taken as to the utilance f of link _u,v/ c _u,v, or the piecewise linear function of link utilization, the target that solves MCF problem is exactly the cost sum that minimizes all links.The 1st of constraints represents the flow f of link (u, v) _u,vequal the flow through all different users of this link

sum; The 2nd of constraints represents that the flow of link (u, v) can not exceed the capacity c of this article of link _u,v; The 3rd of constraints represents user's request constraint, and it meets the requirement of flow conservation, and wherein, D (s, t) represents the traffic demand from server s to user t.In TE problem, D (s, t) is a constant.Above-mentioned MCF problem is a typical protruding optimization problem, and it can obtain optimal solution, i.e. the optimal flux f of every link in polynomial time _u,v.

Traditional TE considers on each node, router how by the assignment of traffic of process on outside link to flow to adjacent node.It should be noted that the traffic demand between traditional TE hypothesis source node and destination node is constant, traffic matrix (being made up of each D (s, t)) is constant.

Do not increase before dummy node in network topology, as shown in Figure 6, each node s _i(for example, server node) is to the traffic demand D (s that is positioned at user node t transmission _i, t) comprising: represent from node s _isend the CP flow D of user node t to ^cp(s _i, t) with background (background) the flow D that sends user t to ^bg(s _i, t), background traffic D ^bg(s _i, t) be constant, D ^cp(s _i, t) be exactly the variable in SS problem.

But, for the combined optimization of TE and SS, the traffic demand D (s of user node t _i, t) be no longer a constant, this be because, each node s _ito the CP flow D of user t ^cp(s _i, t) in SS problem, be a variable.Although each node s _ito the CP flow D of user t ^cp(s _i, t) in SS problem, be a variable, but for certain user node t, its total flow demand is a constant, i.e. all node s that serve user node t _ithe CP flow sum providing

it is a constant.

In network topological diagram, increase after dummy node Ns and virtual link the CP flow D that individual server is changed unique user ^cp(s _i, t) be converted into multiple servers of serving this user to the constant total CP flow of this user

due to D ^bg(s _i, t) be constant, in network topological diagram, increase after dummy node Ns and virtual link the traffic demand D (s that individual server is changed unique user _i, t) be also just converted into multiple servers of serving this user to constant total flow demand D (s, t) with regard to this user.

Described in step S401, after increasing dummy node Ns in network topology, dummy node Ns is to the virtual flow D (Ns of user node t, t) be a constant, the total flow demand D (s, t) of this virtual flow D (Ns, t) and user node t equates.In other words, after increasing dummy node Ns, due to the traffic demand D (s of individual server to unique user _i, t) be change and the 3rd this situation of constraints of not meeting MCF problem changed, dummy node Ns is constant and met the 3rd of the constraints of MCF problem to the virtual flow D (Ns, t) of user node t.So far, in embodiments of the present invention, the 3rd of the constraints of MCF problem can be expressed as $\underset{(\mathrm{Ns},v)\∈E^{*}}{\mathrm{\Σ}}{f}_{\mathrm{Ns},v}^t-\underset{(u,\mathrm{Ns})\∈E^{*}}{\mathrm{\Σ}}{f}_{u,\mathrm{Ns}}^t=D(\mathrm{Ns},t),\∀\mathrm{Ns}\∈V^{*},t\∈V^{*}.$

Through above-mentioned by former network topology G(V, E) (node set that represents V in network topological diagram, E represents link set) be converted into new network topology G ^*(V ^*, E ^*) after, the node set V of new network topology ^*comprise node s _i, dummy node Ns and router node (for example, the node of previous embodiment u) etc., link set E ^*comprise all physical links and virtual link (Ns, s _i), the requirement matrix in MCF problem becomes constant.So far, be equivalent to by the non-MCF problem in the combined optimization scheme of TE and SS (in this problem, the traffic demand D (s of individual server to unique user _i, t) change) and change a standard MCF problem into.Solving standard MCF problem is prior art, does not repeat herein.

At new network topology G ^*(V ^*, E ^*) in, the optimal solution of MCF problem comprises physical link and virtual link (Ns, s _i) on optimal flux.Virtual link (N _s, s _i) on flow represent server node s _iupper total CP flow, $f_{N_s,s_i}=\underset{t}{\mathrm{\Σ}}{D}^{\mathrm{cp}}\left(s_{i,}t\right).$

S403, solves network entropy maximization problem (NEMR, Network Entropy Maximization Relaxed) according to optimal flux on link, obtains the optimum link weight on virtual link and physical link.

So-called NEMR problem refers to the problem that uses following expression formula to represent:

NEMR:

$\max \underset{s,t\∈V}{\mathrm{\Σ}}D(s,t)(\underset{p_{s,t}^j}{\mathrm{\Σ}}-x_{s,t}^j\log x_{s,t}^{\text{j}})$

Constraints:

$\left(1\right),\underset{s,t,j:(u,v)\∈P_{s,t}^i}{\mathrm{\Σ}}D(s,t)x_{s,t}^j\≤f_{u,v},\∀(u,v)\∈E$

$\left(2\right),\underset{s,t,j:(N_s,s_i)\∈P_{N_s,t}^j}{\mathrm{\Σ}}{d}^t\·x_{s,t}^j\≤\underset{t}{\mathrm{\Σ}}{d}^{t+1,\∀(N_s,S_i)}$

$\left(3\right),\begin{array}{cc}\underset{j}{\mathrm{\Σ}}{x}_{s,t}^j=1,& \∀s,t\∈V\end{array}$

Variable: $x_{s,t}^j\≥0$

Utilize Duality Decomposition principle can solve above-mentioned NEMR problem, suppose λ _u,vthe Lagrangian of introducing during for loose NEMR problem the 1st item constraint condition, i.e. the dual variable of lagrange duality problem.If dual variable λ _u,voptimal solution be

obvious, for virtual link (N _s, s _i) traffic constraints condition (the 2nd item constraint condition), inequality left-half is always not more than its right half part.Therefore, make dual variable optimum on all virtual links by the NEMR form of this particular design

be 0.

The solution of NEMR problem has provided the corresponding dual variable of every link (comprising physical link and virtual link) in topological diagram.How the optimal value of these dual variables can be used for calculating on each node the assignment of traffic of identical destination in different adjacent nodes.

NEMR is solved to the optimum dual variable obtaining

with

as the link weight (Link Weights) of each link, i.e. optimum link weight, note optimum link weight is w(u, v), u and v are two nodes that form link.It should be noted that, due to dual variable optimum on all virtual links

be 0, therefore, virtual link (Ns, s _i) on optimum link weight be 0, i.e. w(Ns, s _i)=0.

S404, is distributed to the router node in network topology by optimum link weight.

Utilize this group optimum link weight can realize optimum link flow.Each node can calculate flow for the identical object allocation proportion on each outside link.Suppose that certain node s has the outside link of N bar.The destination node of the flow of this node of process can be any one node in network.For any one destination node t, what need to determine is the flow of its s that the flows through allocation proportion on the outside link of N bar.In specific implementation process, each router of ISP can independently calculate the allocation proportion for any destination node according to the weight of the topological sum link of whole network.

Referring to accompanying drawing 6, is the combined optimization method schematic flow sheet that a kind of traffic engineering of providing of another embodiment of the present invention and server are selected, and mainly comprises step:

S601, server obtains the shortest path length and this router equivalent number to user node of this router to user node from connected router.

This step can adopt distributed method to realize.Particularly, the server s of CP _ifrom contiguous router u _j(router being connected with each server of CP) obtains some information, and these information comprise the shortest path length of this router to each user node t and equivalent number

because each router can calculate and by shortest path length and equivalent number

storage.As the server s of CP _ito contiguous router u _jwhile sending control information, router is by the shortest path length of storage and equivalent number be sent to server s etc. information _i.Server s _ican utilize these information calculation servers s _ito the shortest path length of user node t

and equivalent number

when Servers-all has all obtained shortest path length

and equivalent number

after, these servers can exchange these information mutually.After exchange finishes, each server can independently calculate its corresponding dispense flow rate

and then adopt distributed way to make server selection strategy.

ISP need to know the information of traffic matrix in the time carrying out combined optimization, and this has comprised all background traffic and CP flow.It should be noted that in traffic matrix, all CP flows are all from dummy node.ISP can obtain these information by the measurement at bottom-layer network and estimation, also can select to obtain these information from CP.

S602, server obtains the optimum link weight of each virtual link.

In embodiments of the present invention, optimum link weight obtains by the NEMR problem that solves the network topology that has increased dummy node, can consult above to explanation embodiment illustrated in fig. 4.Virtual link is that the dummy node and the server that in network topology, increase form.

S603, to the equivalent number of user node t with belong to the total flow demand of user node, determines the traffic demand of distributing to user node on this server according to the optimum link weight of obtaining, shortest path length, router.

First, calculate dummy node Ns through any virtual link (Ns, a s _i) difference to user node t shortest path length and dummy node Ns to user node t shortest path length.

Dummy node Ns is through any virtual link (Ns, a s _i) equal virtual link (Ns, s to user node t shortest path length _i) optimum link weight w(Ns, s _i) and node s _i(server) is to user node t shortest path length sum,

due to w(Ns, s _i) be 0, therefore, dummy node Ns is through virtual link (Ns, s _i) actually to user node t shortest path length be in fact,

wherein, node u _jrepresent and node s _ithe router that (server) is adjacent, node u _jbe positioned at from s _ito the shortest path of t,

represent node s _ito adjacent node u _jlink (s _i, u _j) length, when while obtaining in S601,

be not difficult to try to achieve.

Similarly, dummy node Ns is to the shortest path length employing of user node t

represent, can use above-mentioned similar approach to obtain, do not do repeat specification.

Shortest path from dummy node Ns to user t need pass through certain virtual link, therefore,

again due to virtual link (N _s, s _i) optimum link weight w (Ns, s _i) be 0, therefore, dummy node Ns is through any virtual link (N _s, s _i) (it will be appreciated by those skilled in the art that this shortest path may be also that dummy node Ns is through any virtual link (N to user node t shortest path length and dummy node Ns to user node t shortest path _s, s _i) to user node t shortest path) difference of length is $h_{N_s,s_i}^t=d_{s_i}^t+w(\mathrm{Ns},s_i)-d_{N_s}^t=d_{s_i}^t-d_{N_s}^t.$

Then, equivalent number and the difference to user node t according to router

determine total flow demand any virtual link (N in each virtual link _s, s _i) on the ration of division.

It should be noted that, due to every virtual link (N _s, s _i) in another node be exactly server (node s _i), for the total flow demand of user node t, total flow demand is any virtual link (N in each virtual link _s, s _i) on the ration of division namely total flow demand need to be assigned to corresponding server (node s _i) flow proportional.

As one embodiment of the invention, determine total flow demand any virtual link (N in each virtual link _s, s _i) on the ration of division can adopt following method:

S1, obtains the equivalent number of server to user node t by router to the equivalent number of user node t.

Router (node u _j) to the equivalent number of user node t

in step S601, obtain, be not difficult to obtain server (node s by router to the equivalent number of user node t _i) to the equivalent number of user node t

S2, calculation server, to the equivalent number of user node and the product of an attenuation parameter, obtains any virtual link (N _s, s _i) traffic demand segmentation function.

Attenuation parameter is the difference of natural Exponents e

the inverse of inferior power, that is,

therefore, any virtual link (N _s, s _i) traffic demand segmentation function be

traffic demand segmentation function

with above-mentioned any virtual link (N _s, s _i) correspondence.

S3, calculate flow segmentation function that each virtual link is corresponding and.

The flow segmentation function that each virtual link is corresponding and

S4, asks for traffic demand segmentation function

with

ratio, obtain aggregate demand flow at any virtual link (N _s, s _i) on the ration of division,

Finally, calculate total flow demand and the ration of division

product, obtain total flow demand

At any virtual link (N _s, s _i) on the flow cut apart

due to every virtual link (N _s, s _i) in another node be exactly server (node s _i), for the total flow demand of user node t, total flow demand is any virtual link (N in each virtual link _s, s _i) on the flow cut apart namely the total flow demand of user node t need to be assigned to this virtual link (N _s, s _i) corresponding server (node s _i) on flow.

It should be noted that, be to adopt distributed solution SS problem although above-mentioned, is not limited to distributed.In embodiments of the present invention, the server selection course of CP also can adopt centralized fashion to process.Specifically, the server s of CP _ifrom contiguous router u _jthere obtains the shortest path

and equivalent number

and then calculate server s _ito the shortest path length of user node t

and equivalent number from distributed different, each server s _ishortest path length by it to user node t

and equivalent number

feed back to the centralized management system of CP.Based on these information, the centralized management system of CP can be made server selection strategy, and need to not obtain complete topology information from ISP there.This guarantees that ISP and CP can also obtain optimum solution in information each other in the least possible exposure.

The embodiment providing from above-mentioned Fig. 6, how the considerably less information that CP can provide according to ISP determine independently customer flow demand assignment to different servers, thereby solve SS problem; Or, can distributed earth between server exchange shortest path length and equivalent number information, each server can distributed earth, determine independently it self assignment of traffic to different user.Due in optimizing process, a small amount of variable information that CP or its server only need to obtain from its adjacent router (for example, each router is to shortest path length information and the equivalent number information of user node) and do not need the key message of bottom-layer network, (for example, topological sum link-state information), can independently upgrade oneself SS strategy.Therefore,, even in the situation that ISP is unwilling to share bottom-layer network information with CP, the TE based on PEFT and SS combined optimization (PETS, PEFT based joint TE and SS) are still suitable for.

Refer to Fig. 7, a kind of router logic structural representation that the embodiment of the present invention provides.For convenience of explanation, only show the part relevant to the embodiment of the present invention.This router comprises outside link determination module 71, acquisition module 72 and segmentation strategy determination module 73, wherein:

Outwards link determination module 71, for determining that carrying belongs to each outside link of the flow of same user node;

Acquisition module 72, for obtaining the optimum link weight of each definite outside link of outside link determination module 71, this optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node;

Segmentation strategy determination module 73, for the optimum link weight of obtaining according to acquisition module 72 and the flow that belongs to same user node, determine the flow any outside link (u in each outside link that belongs to same user node, v) the segmentation strategy of going up, wherein, u represents the node at described router place, and v represents to form described outside link (u, another node v).

Further, segmentation strategy determination module 73 comprises the first difference computational unit 81, the first ration of division determining unit 82 and cuts apart flow rate calculation unit 83, the router that another embodiment of the present invention provides as shown in Figure 8, wherein:

The first difference computational unit 81, for calculating router, (node is u) through any outside link (u, the v) difference to the shortest path length of user node t and router to the shortest path length of user node t;

It should be noted that, router to the shortest path of user node t may be also that (u) through any outside link, (u, v) to the shortest path of user node t for node for router.

The first ration of division determining unit 82, for the difference of trying to achieve to equivalent number and first difference computational unit 81 of user node according to router, determine the flow any an outside link (u, the ration of division v) going up in each outside link that belong to same user node;

Cut apart flow rate calculation unit 83, belong to the flow of same user node and the product of the ration of division for calculating, obtain belonging to flow any outside link (u, flow of cutting apart of v) going up in each outside link of same user node.

Further, the first ration of division determining unit 82 comprises:

The first product computing unit, for calculating the product of equivalent number and an attenuation parameter, obtains any outside link (u, v) flow segmentation function, this flow segmentation function and any outside link (u, v) corresponding, attenuation parameter is the inverse of the described difference time power of natural Exponents e;

The first sum unit, for calculate flow segmentation function that each outside link is corresponding and;

The first ratio is asked for unit, for ask for flow segmentation function that flow segmentation function is corresponding with described each outside link and ratio, obtain belonging to flow any an outside link (u, the ration of division v) going up in each outside link of same user node.

Refer to Fig. 9, a kind of server logic structural representation that the embodiment of the present invention provides.For convenience of explanation, only show the part relevant to the embodiment of the present invention.This server comprises the first acquisition module 91, the second acquisition module 92 and flow determination module 93, wherein:

The first acquisition module 91, for obtaining shortest path length and the described router extremely equivalent number of described user node of described router to user node from the router being connected with described server;

The second acquisition module 92, for obtaining the optimum link weight of each virtual link, described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node Ns, and described virtual link is that the described dummy node and the described server that in network topology, increase form;

Flow determination module 93, for according to belonging to shortest path length that optimum link weight that the total flow demand of described user node, described the second acquisition module 92 obtain and described the first acquisition module 91 obtain and described router to the equivalent number of described user node, determine the traffic demand of distributing to described user node on described server.

Further, flow determination module 93 comprises the second difference computational unit 101, the second ration of division determining unit 102 and traffic demand computing unit 103, wherein:

The second difference computational unit 101, for calculating described dummy node through described any virtual link (Ns, a s _i) to shortest path length and the extremely difference of the shortest path length of described user node of described dummy node of described user node, wherein, Ns represents described dummy node, s _irepresent to form described virtual link (Ns, s _i) another node;

The second ration of division determining unit 102, for calculating gained difference according to described router to equivalent number and described second difference computational unit 101 of described user node, determine described total flow demand any virtual link (Ns, s in described each virtual link _i) on the ration of division;

Traffic demand computing unit 103, for calculating the product of described total flow demand and the described ration of division, obtain described total flow demand with described any virtual link (Ns, a s _i) traffic demand of distributing to described user node on corresponding server.

Further, the second ration of division determining unit 102 comprises that equivalent number acquiring unit, the second product computing unit, the second sum unit and the second ratio ask for unit, wherein:

Equivalent number acquiring unit, obtains the equivalent number of described server to described user node by the equivalent number of described router;

The second product computing unit, for calculating described server to the equivalent number of described user node and the product of an attenuation parameter, obtains described any virtual link (Ns, a s _i) traffic demand segmentation function, described traffic demand segmentation function and described any virtual link (Ns, a s _i) correspondence, described attenuation parameter is the negative described difference time power of natural Exponents e;

The second sum unit, for calculate the flow segmentation function corresponding with described each virtual link and;

The second ratio is asked for unit, for ask for described traffic demand segmentation function and described and ratio, obtain described total flow demand any virtual link (Ns, s in described each virtual link _i) on the ration of division.

Refer to Figure 11, the optimum link Weight Acquisition device logical construction schematic diagram that the embodiment of the present invention provides.For convenience of explanation, only show the part relevant to the embodiment of the present invention.This device comprises that conversion module 111, optimal flux are asked for module 112, optimum link weight is asked for module 113 and distribution module 114, wherein:

Conversion module 111, be used for by increase dummy node in network topology, each server node in described network topology to the traffic demand of user node is converted into the virtual flow of described dummy node to described user node, described virtual flow is constantly equal to the total flow demand of described user node, between each server node in described dummy node and described network topology, forms virtual link;

Optimal flux is asked for module 112, for the constant take described virtual flow as many article network flow MCF problem user's request bound term, solve the optimal flux that described MCF problem obtains described each link of network topology, each link of described network topology comprises physical link and the described virtual link in described network topology, and in described MCF problem, user's request bound term is:

$\underset{(s,v)\∈E}{\mathrm{\Σ}}{f}_{s,v}^t-\underset{(u,s)\∈E}{\mathrm{\Σ}}{f}_{u,s}^t=D(s,t),\∀s\∈V,t\∈V,$ Wherein, (s, v) ∈ E represents network topological diagram G (V, E) the oriented link from server node s to node v in, (u, s) ∈ E represents the oriented link from node u to server node s, V and E represent respectively node set and link set

represent oriented link (s, belongs to the traffic demand of user node t on v),

represent to belong on oriented link (u, s) traffic demand of user node t, D (s, t) represents the traffic demand from server node s to user node t, and described D (s, t) is the constant of described user's request bound term;

Optimum link weight is asked for module 113, solves network entropy maximization NEMR problem for ask for the optimal flux that module 112 asks for according to the optimal flux on described link, obtains the optimum link weight on described virtual link and physical link;

Distribution module 114, for being distributed to the router node in described network topology by described optimum link weight.

Refer to Figure 12, the combined optimization system logical construction schematic diagram that a kind of traffic engineering that the embodiment of the present invention provides and server are selected.For convenience of explanation, only show the part relevant to the embodiment of the present invention.This system comprises the server 122 of optimum link Weight Acquisition device 121, accompanying drawing 9 or accompanying drawing 10 examples and the router one 23 of accompanying drawing 7 or accompanying drawing 8 examples of accompanying drawing 11 examples, wherein:

Optimum link Weight Acquisition device 121, be used for by increase dummy node in network topology, each server 122 nodes to traffic demand of user node in described network topology is converted into the virtual flow of described dummy node to described user node, described virtual flow is constantly equal to the total flow demand of described user node, between each server 122 nodes in described dummy node and described network topology, form virtual link, take the constant of described virtual flow user's request bound term in many article network flow MCF problem, solve the optimal flux that described MCF problem obtains described each link of network topology, each link of described network topology comprises physical link and the described virtual link in described network topology, in described MCF problem, user's request bound term is:

$\underset{(s,v)\∈E}{\mathrm{\Σ}}{f}_{s,v}^t-\underset{(u,s)\∈E}{\mathrm{\Σ}}{f}_{u,s}^t=D(s,t),\∀s\∈V,t\∈V,$ Wherein, (s, v) ∈ E represents network topological diagram G (V, E) the oriented link from server node s to node v in, (u, s) ∈ E represents the oriented link from node u to server node s, V and E represent respectively node set and link set

represent oriented link (s, belongs to the traffic demand of user node t on v),

represent oriented link (u, s) on, belong to the traffic demand of user node t, D (s, t) represent the traffic demand from server node s to user node t, described D (s, t) is the constant of described user's request bound term, solves network entropy maximization NEMR problem according to the optimal flux on described link, obtain the optimum link weight on described virtual link and physical link, described optimum link weight is distributed to router one 23 nodes in described network topology;

Server 122, for obtain the equivalent number of described router one 23 to the shortest path length of user node and described router one 23 to described user node from connected router one 23, obtain the optimum link weight of each virtual link, described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node, described virtual link is that the described dummy node and the described server that in network topology, increase form, according to described optimum link weight, described shortest path length, described router one 23 is to the equivalent number and the total flow demand that belongs to described user node of described user node, determine the traffic demand of distributing to described user node on described server,

Router one 23, for determining that carrying belongs to each outside link of the flow of same user node, obtain the optimum link weight of described each outside link, described optimum link weight obtains by the network entropy maximization problem NEMR that solves the network topology that has increased dummy node, according to described optimum link weight and described in belong to the flow of same user node, described in determining, belong to flow any outside link (u in described each outside link of same user node, v) the segmentation strategy of going up, wherein, u represents the node at described router place, v represents to form described outside link (u, v) another node.

It should be noted that, the content such as information interaction, implementation between the each module/unit of said apparatus, due to the inventive method embodiment based on same design, its technique effect bringing is identical with the inventive method embodiment, particular content can, referring to the narration in the inventive method embodiment, repeat no more herein.

One of ordinary skill in the art will appreciate that all or part of step in the whole bag of tricks of above-described embodiment is can carry out the hardware that instruction is relevant by program to complete, this program can be stored in a computer-readable recording medium, storage medium can comprise: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), disk or CD etc.

Combined optimization method, system and relevant device that the traffic engineering above embodiment of the present invention being provided and server are selected are described in detail, applied specific case herein principle of the present invention and execution mode are set forth, the explanation of above embodiment is just for helping to understand method of the present invention and core concept thereof; , for one of ordinary skill in the art, according to thought of the present invention, all will change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention meanwhile.

Claims (13)

1. the combined optimization method that traffic engineering and server are selected, is characterized in that, comprising:

Router determines that carrying belongs to each outside link of the flow of same user node;

Described router obtains the optimum link weight of described each outside link, and described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node;

Described router calculates described router through any outside link (u, v) to the shortest path length of described user node and the extremely difference of the shortest path length of described user node of described router, according to described difference and the extremely equivalent number of described user node of described router, described in determining, belong to flow any outside link (u in each outside link of same user node, v) the ration of division going up, described in calculating, belong to the flow of same user node and the product of the described ration of division, described in obtaining, belong to flow any outside link (u in described each outside link of same user node, v) the flow of cutting apart of going up, wherein, u represents the node at described router place, v represents to form described outside link (u, v) another node.

2. the method for claim 1, it is characterized in that, described according to described difference and described router the equivalent number to described user node, the flow that belongs to same user node described in determining in described each outside link any outside link (u, the ration of division v) going up comprises:

Calculate the product of described equivalent number and an attenuation parameter, obtain described any outside link (u, flow segmentation function v), described flow segmentation function and described any outside link (u, v) corresponding, described attenuation parameter is the inverse of the described difference time power of natural Exponents;

Calculate flow segmentation function that described each outside link is corresponding and;

Ask for described flow segmentation function and described and ratio, described in obtaining, belong to flow any an outside link (u, the ration of division v) going up in described each outside link of same user node.

3. the method for claim 1, is characterized in that, also comprises:

Described router is sent to described router to the equivalent number of described user node the server being connected with described router by described router to the shortest path length of described user node.

4. the method as described in claims 1 to 3 any one, is characterized in that, each node in described each outside link to the shortest path length of described user node is less than the shortest path length of described router to described user node.

5. the combined optimization method that traffic engineering and server are selected, is characterized in that, comprising:

Server obtains shortest path length and the described router extremely equivalent number of described user node of described router to user node from connected router;

Described server obtains the optimum link weight of each virtual link, described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node, and described virtual link is that the described server of the described dummy node that increases in network topology forms;

Described server calculates described dummy node through any virtual link (Ns, si) to the shortest path length of described user node and the extremely difference of the shortest path length of described user node of described dummy node, according to described equivalent number and described difference, determine total flow demand any virtual link (Ns in each virtual link, si) ration of division on, calculate the product of described total flow demand and the described ration of division, obtain described total flow demand with described any virtual link (Ns, si) on corresponding server, distribute to the traffic demand of described user node, wherein, Ns represents described dummy node, si represents to form described virtual link (Ns, si) another node.

6. method as claimed in claim 5, is characterized in that, described according to described equivalent number and described difference, determines that described total flow demand ration of division on any virtual link (Ns, si) in described each virtual link comprises:

Obtain the equivalent number of described server to described user node by described equivalent number;

Calculate described server to the equivalent number of described user node and the product of an attenuation parameter, obtain described any virtual link (Ns, si) traffic demand segmentation function, described traffic demand segmentation function and described any virtual link (Ns, si) correspondence, described attenuation parameter is the inverse of the described difference time power of natural Exponents;

Calculate flow segmentation function that described each virtual link is corresponding and;

Ask for described traffic demand segmentation function and described and ratio, obtain described total flow demand ration of division on any virtual link (Ns, si) in described each virtual link.

7. the combined optimization method that traffic engineering and server are selected, is characterized in that, comprising:

By increase dummy node in network topology, each server node in described network topology to the traffic demand of user node is converted into the virtual flow of described dummy node to described user node, described virtual flow is constantly equal to the total flow demand of described user node, between each server node in described dummy node and described network topology, forms virtual link;

Take the constant of described virtual flow user's request bound term in many article network flow MCF problem, solve described MCF problem and obtain the optimal flux on each link of described network topology, each link of described network topology comprises physical link and the described virtual link in described network topology, and in described MCF problem, user's request bound term is:

, wherein, (s, v) ∈ E represents the oriented link from server node s to node v in network topological diagram G (V, E), (u, s) ∈ E represents the oriented link from node u to server node s, and V and E represent respectively node set and link set represent oriented link (s, belongs to the traffic demand of user node t on v),

represent to belong on oriented link (u, s) traffic demand of user node t, D (s, t) represents the traffic demand from server node s to user node t, and described D (s, t) is the constant of described user's request bound term;

Solve network entropy maximization NEMR problem according to the optimal flux on described link, obtain the optimum link weight on described virtual link and physical link;

Described optimum link weight is distributed to the router node in described network topology.

8. a router, is characterized in that, comprising:

Outwards link determination module, for determining that carrying belongs to each outside link of the flow of same user node;

Acquisition module, for obtaining the optimum link weight of each definite outside link of described outside link determination module, described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node;

The first difference computational unit, be used for calculating described router through any outside link (u, v) to the shortest path length of described user node and the extremely difference of the shortest path length of described user node of described router, wherein, u represents the node at described router place, v represents to form described outside link (u, another node v);

The first ration of division determining unit, be used for according to described router to the equivalent number of described user node and the difference that described the first difference computational unit is tried to achieve, described in determining, belong to flow any an outside link (u, the ration of division v) going up in described each outside link of same user node;

Cut apart flow rate calculation unit, described in calculating, belong to the flow of same user node and the product of the described ration of division, described in obtaining, belong to flow any outside link (u, flow of cutting apart of v) going up in described each outside link of same user node.

9. router as claimed in claim 8, is characterized in that, described the first ration of division determining unit comprises:

The first product computing unit, for calculating the product of described equivalent number and an attenuation parameter, obtain described any outside link (u, v) flow segmentation function, described flow segmentation function and described any outside link (u, v) corresponding, described attenuation parameter is the inverse of the described difference time power of natural Exponents;

The first sum unit, for calculate the flow segmentation function corresponding with described each outside link and;

The first ratio is asked for unit, for ask for described flow segmentation function and described and ratio, described in obtaining, belong to flow any an outside link (u, the ration of division v) going up in described each outside link of same user node.

10. a server, is characterized in that, comprising:

The first acquisition module, for obtaining shortest path length and the described router extremely equivalent number of described user node of described router to user node from the router being connected with described server;

The second acquisition module, for obtaining the optimum link weight of each virtual link, described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node Ns, and described virtual link is that the described dummy node and the described server that in network topology, increase form;

The second difference computational unit, be used for calculating described dummy node through any virtual link (Ns, si) to the shortest path length of described user node and the extremely difference of the shortest path length of described user node of described dummy node, wherein, Ns represents described dummy node, si represents to form another node of described virtual link (Ns, si);

The second ration of division determining unit, for calculating gained difference according to described router to equivalent number and described second difference computational unit of described user node, determine total flow demand ration of division on any virtual link (Ns, si) in described each virtual link;

Traffic demand computing unit, for calculating the product of described total flow demand and the described ration of division, obtain described total flow demand with server corresponding to described any virtual link (Ns, si) on the traffic demand of distributing to described user node.

11. servers as claimed in claim 10, is characterized in that, described the second ration of division determining unit comprises:

Equivalent number acquiring unit, obtains the equivalent number of described server to described user node by the equivalent number of described router;

The second product computing unit, for calculate described server to the equivalent number of described user node with the product of an attenuation parameter, obtain described any virtual link (Ns, si) traffic demand segmentation function, described traffic demand segmentation function and described any virtual link (Ns, si) correspondence, described attenuation parameter is the inverse of the described difference time power of natural Exponents e;

The second sum unit, for calculate the flow segmentation function corresponding with described each virtual link and;

The second ratio is asked for unit, for ask for described traffic demand segmentation function and described and ratio, obtain described total flow demand ration of division on any virtual link (Ns, si) in described each virtual link.

12. 1 kinds of optimum link Weight Acquisition devices, is characterized in that, comprising:

Conversion module, be used for by increase dummy node in network topology, each server node in described network topology to the traffic demand of user node is converted into the virtual flow of described dummy node to described user node, described virtual flow is constantly equal to the total flow demand of described user node, between each server node in described dummy node and described network topology, forms virtual link;

Optimal flux is asked for module, for the constant take described virtual flow as many article network flow MCF problem user's request bound term, solve the optimal flux that described MCF problem obtains described each link of network topology, each link of described network topology comprises physical link and the described virtual link in described network topology, and in described MCF problem, user's request bound term is:

, wherein, (s, v) ∈ E represents network

Oriented link from server node s to node v in topological diagram G (V, E), (u, s) ∈ E represents the oriented link from node u to server node s, V and E represent respectively node set and the link set on network topology G (V, E)

represent oriented link (s, belongs to the traffic demand of user node t on v),

represent to belong on oriented link (u, s) traffic demand of user node t, D (s, t) represents the traffic demand from server node s to user node t, and described D (s, t) is the constant of described user's request bound term;

Optimum link weight is asked for module, solves network entropy maximization NEMR problem for ask for the optimal flux that module asks for according to the optimal flux on described link, obtains the optimum link weight on described virtual link and physical link;

Distribution module, for being distributed to the router node in described network topology by described optimum link weight.

The combined optimization system that 13. 1 kinds of traffic engineerings and server are selected, is characterized in that, described system comprises optimum link Weight Acquisition device, server and router;

Described optimum link Weight Acquisition device, be used for by increase dummy node in network topology, each server node in described network topology to the traffic demand of user node is converted into the virtual flow of described dummy node to described user node, described virtual flow is constantly equal to the total flow demand of described user node, between each server node in described dummy node and described network topology, form virtual link, take the constant of described virtual flow user's request bound term in many article network flow MCF problem, solve the optimal flux that described MCF problem obtains described each link of network topology, each link of described network topology comprises physical link and the described virtual link in described network topology, in described MCF problem, user's request bound term is: , wherein, (s, v) ∈ E represents the oriented link from server node s to node v in network topological diagram G (V, E), (u, s) ∈ E represents the oriented link from node u to server node s, and V and E represent respectively node set and link set

represent oriented link (s, belongs to the traffic demand of user node t on v),

represent oriented link (u, s) on, belong to the traffic demand of user node t, D (s, t) represent the traffic demand from server node s to user node t, described D (s, t) is the constant of described user's request bound term, solves network entropy maximization NEMR problem according to the optimal flux on described link, obtain the optimum link weight on described virtual link and physical link, described optimum link weight is distributed to the router node in described network topology,

Described server, for obtain shortest path length and the described router extremely equivalent number of described user node of described router to user node from connected router, obtain the optimum link weight of each virtual link, described optimum link weight obtains by the network entropy maximization NEMR problem that solves the network topology that has increased dummy node, described virtual link is that the described dummy node and the described server that in network topology, increase form, according to described optimum link weight, described shortest path length, described router is to the equivalent number and the total flow demand that belongs to described user node of described user node, determine the traffic demand of distributing to described user node on described server,

Described router, for determining that carrying belongs to each outside link of the flow of same user node, obtain the optimum link weight of described each outside link, described optimum link weight obtains by the network entropy maximization problem NEMR that solves the network topology that has increased dummy node, according to described optimum link weight and described in belong to the flow of same user node, described in determining, belong to flow any outside link (u in described each outside link of same user node, v) the segmentation strategy of going up, wherein, u represents the node at described router place, v represents to form described outside link (u, v) another node.

Images