CN105162613A - Resource quantitative distribution method based on optical network architecture service plane - Google Patents

Abstract

The invention discloses a resource quantitative distribution method based on an optical network architecture service plane, which comprises the steps of reducing physical topology nodes into abstract network topology with only boundary nodes being left in physical network topology; enabling links of different target-source node pairs in the abstract network topology to share the same practical link, wherein different practical links can be selected to carry out data transmission of the links of the target-source node pairs, and the data flow of each practical link can be only along the same direction; calculating a bandwidth value of a shortest path between each internal node and all of the boundary nodes in the physical topology, and merging the internal node to the boundary node corresponding to a maximum value of the bandwidth value. According to the invention, linear constraint conditions are defined, bandwidth resources of all links are sufficiently considered, thereby not only being high in resource request hit ratio, but also fully integrating scattered bandwidth resources, and enabling the utilization ratio of the resources to be improved correspondingly.

Description

A kind of resource based on optical-fiber network framework service plane quantizes distribution method

Technical field

The present invention is based on service-oriented optical-fiber network architecture, carries out abstract and distributes, be specifically related to the quantizing distribution algorithm of the abstract of network physical topology and related resource to Internet resources and IT resource.

Background technology

In recent years, Internet service achieved huge development, and network application service and number of users are also increasing rapidly.But various network application and huge userbase bring stern challenge to the development of optical interconnection network technology.These challenges mainly comprise: the harmony of Internet resources and IT resource (computational resource and storage resources), the high service reliability that user requires, the diversity of Network and extensibility etc.

In order to meet such challenge, optical-fiber network with the addition of service plane on original three-tier architecture (management plane, control plane, transport plane) basis.Service plane is deployed between user and basic network equipment, its main task is the correlative detail that shielding Internet Transmission realizes on the one hand, on the other hand the role zone of operator is divided into network provider (or resource provider NP) and ISP (SP), namely carries out abstract to resource and abstract resource is reasonably distributed.Service plane has two kinds of interfaces: a kind of is interface CSI (Customer-ServicePlaneInterface) between user, can the request of response user rapidly and efficiently; Another kind is the interface UNI (UsertoNetworkInterface) between service plane and basic network equipment, it controls one group of protocol groups, as: negotiate marking commutation (MPLS), general multiple protocols exchange (Fig. 1) such as (GMPLS).Can also operation exchange be realized by this service plane and Service control is separated, make service generation independent of the running environment of business.Operation exchange and Service control are separated the traditional approach changing and provided different Added Business by switching system.Switching system is only responsible for exchanging and access function, and what make business provides the supplier not relying on service plane relevant device.

Service plane is made up of the abstract mechanism of one group of self adaptation and intelligence, for service and application provide the interface being connected to underlying resource framework.These abstract mechanisms are platforms flexibly, can be deployed in fast and efficiently by new services and applications in a public bottom layer node.But physical resource supplier (NP) is in order to the confidentiality of business, and what be supplied to ISP (SP) is the abstract of network topology.The message recipient of service plane receives the request of user and the network abstraction topology of hardware provider simultaneously, then abstract topology is carried out virtual quantification and request is processed, the demand for services (Fig. 2) of user can be met finally by the composition of serving and service adaptive polo placement Current resource.The flow process of user's RQ cycle of the supplier (SP) served in service plane, user (USER) and hardware provider (NP) three as shown in Figure 3.After ISP receives the service request of user, according to the abstract resource that physical resource supplier provides, service being replied, judge whether to accept request, if do not accepted, directly returning; If accepted, to the corresponding physical resource of resource provider application, if real resource meets user's request, then provide corresponding resource; If real resource does not meet, then reject the request of ISP.Again abstract to surplus resources after an assignment period.The Resource Abstract algorithm of a high accuracy can make service plane more accurate to service request response, comprising two aspects:

(1) request of ISP's acceptance, resource provider should have corresponding real resource;

(2) request of ISP's refusal also should guarantee that resource provider does not have enough resources to provide.These 2 is also the main quantitatively evaluating standard of abstract algorithm.

Summary of the invention

In order to overcome deficiency of the prior art, the object of the invention is the Resource Abstract quantizing distribution algorithm of the service plane design pinpoint accuracy for ASON.

This algorithm utilizes linear programming equation to carry out abstract to Internet resources.Main thought is: different abstracting link may share the physical link of same reality, and same abstracting link also can use different physical links to carry out the transmission of data simultaneously.After a request dispatching cycle completes, service plane can carry out the abstract of a surplus resources, and abstract data are passed to ISP.The request that the request-reply correspondence that ISP carries out next round according to new abstract data again meets, ISP sends resource request to resource provider.Resource provider is checked if the resource of reality meets the demands, and provides resource, does not meet and then refuses.

The present invention adopts following technical scheme to achieve these goals: a kind of resource based on optical-fiber network framework service plane quantizes distribution method, comprises the following steps:

(1) physical topology is abstract, in physical network topology, is the abstract network topology of only remaining boundary node by physical topology node abatement.

(2) Internet resources is abstract, by link sharing same physical link right for target-source nodes different in abstract network topology, the transfer of data of the link that target-source node is right can select different physical links to carry out, and wherein the data flow of each physical link all can only along same direction.

(3) IT resource is abstract, and in Computational Physics network topology, each internal node is to the bandwidth value of the shortest path of all boundary nodes, internal node is integrated into the boundary node corresponding to maximum of described bandwidth value.

The Advantageous Effects that the present invention has is: compared with existing Internet resources abstract algorithm SILK, abstract more accurate to resource of algorithm herein, traditional SILK algorithm just considers the bandwidth of certain single link, although ISP shoots straight to request, the utilance of physical resource is low.This algorithm is by the condition of definition linear restriction, and take into full account the bandwidth resources of all links, not only resource request shoots straight, and fully incorporates the bandwidth resources of dispersion, and the utilance of resource is also improved accordingly.

Accompanying drawing explanation

Fig. 1 is optical network system framework;

Fig. 2 is optical-fiber network service plane structure;

Fig. 3 is optical-fiber network service plane RQ cycle flow chart;

Fig. 4 is network physical topological abstract example;

Fig. 5 is Internet resources abstract algorithm flow process;

Fig. 6 is boundary node merger schematic diagram;

Fig. 7 is the bar chart of amount of bandwidth and the actual amount of bandwidth of distributing in abstract algorithm;

Fig. 8 is the refusal return rate curve charts of two algorithms in three topological diagrams.

Embodiment

In order to understand technical scheme of the present invention better, describe in detail below in conjunction with accompanying drawing.

One, physical topology is abstract

The abstract of network topology is carried out by the method for Full-mesh.Physical topology node abatement is the abstract topology of only remaining boundary node by it.Boundary node is exactly the node be connected with other network domains in the network domains of self.

Network physical topological diagram G (V, E) represents, wherein V={v _nrepresent the node of network, E={e _{(u, v)}| (u, v) ∈ V ²represent the limit of topological diagram.|| V||=N represents the number of node, || E||=L represents the number on limit, B _{(u, v)}represent limit e _{(u, v)}bandwidth, unit is bytes-per-second.The abstract of physical topology is exactly by subset G'(V', the E' of topological diagram G (V, E) with it) represent, wherein V'={v _bbe the set of the boundary node of topological G, || V'||=N' is boundary node number, V " be the set of internal node, N "=|| V " || be the number of internal node.E'={e' _{(s, d)}| (s, d) ∈ V' ², e' _{(s, d)}represent the limit of connected node s and node d.|| E'||=L' represents the number on abstract topological limit, is a undirected full-mesh figure (Fig. 4).

Two, Internet resources is abstract

(1) parameter is abstract: nowadays the abstract of most of Internet resources is SimpleLink-SILK algorithm.The main thought of SILK algorithm is the bandwidth value of choosing a maximum access in the practical topology bandwidth as abstracting link.The accuracy of this algorithm Resourse Distribute is higher, but the utilance of resource is low.In order to solve the low problem of SILK algorithm resource utilization, the present invention proposes MILP (mixed integer linear program) equation to carry out the abstract of Internet resources.Main thought is that different abstracting link can share same physical link, and the transfer of data of same abstracting link can be carried out along different physical links.Because total transmission bandwidth cannot exceed the bandwidth of physical link, therefore define the constraints of linear programming.

MILP equation defines:

According to the definition of above actual physics network topology, set following parameter:

(2) target function:

$\max [\underset{v_s\∈V^{\′}}{\mathrm{\Σ}}\underset{v_d\∈V^{\′}}{\mathrm{\Σ}}\underset{k=1}{\overset{P_{\max }}{\mathrm{\Σ}}}{Z}_k^{s,d}+\mathrm{\α\ζ}]---\left(1\right)$

Formula (1) represents each node v _s∈ V' is to other any one nodes v _dthe bandwidth sum that all path k of ∈ V' can provide adds the bandwidth be assigned with.Target function is exactly the maximum obtaining the available bandwidth resources of whole abstract topology.Formula latter half α ζ is tending towards even to allow the right allocated bandwidth of all " target-source " node, and when minimum value ζ is more close to mean value, the value of equation is larger.

(3) constraints:

$\∀v_j\∈V,\∀k\∈[1...P_{\max }],$

$\underset{v_u}{\Σ}{X}_{u,j.k}^{s,d}-\underset{v_v}{\Σ}{X}_{j,v,k}^{s,d}=\left\{\begin{array}{cc}1,& j=d\\ -1,& j=s\\ 0,& otherwise\end{array}\right.---\left(2\right)$

V _uand v _vrepresent the node of actual physics topology, (v _u, v _v) represent a limit of practical topology.

Formula (2) constraint ensures each node v _jthe data flow of ∈ V is always transmitted along same direction in the same path k of a pair " object-source " node.But same data flow can at way set P _{(s, d)}the path that middle selection is different, each path data flow all can only along same direction, as formula (3)

$X_{u,v,k}^{s,d}+X_{v,u,k}^{s,d}\≤1,\∀e_{(u,v)}\∈E,\∀k\∈[1...P_{\max }]---\left(3\right)$

$\underset{v_s\∈V^{\′}}{\mathrm{\Σ}}\underset{v_d\∈V^{\′}}{\mathrm{\Σ}}\underset{k=1}{\overset{P_{\max }}{\mathrm{\Σ}}}{Z}_k^{s,d}*X_{u,x,d}^{s,d}\≤B_{(u,v)},\∀e_{(u,v)}\∈E---\left(4\right)$

subscript (s, d) represents abstract limit (v _s, v _d) kth paths (abstract edge-band width may comprise the bandwidth of k bar Actual path in practical topology) whether comprise actual limit (v _u, v _v).V _sand v _dall represent the abstract node of abstract network topology, (v _s, v _d) representing source-destination node pair, parameter is abstract is exactly calculate v _sto v _dmaximum amount of bandwidth, E represent actual physics topology limit set.

Formula (4) represents that all paths of the data flow of all boundary nodes can not exceed the maximum bandwidth of this edge through the summation of the bandwidth of same physics limit e.Equation (4) belongs to nonlinear restriction, because it comprises Two Variables.Therefore following three linear restrictions of definition again:

$\∀(s,d)\∈V^{\′2},\∀e_{(u,v)}\∈E,\∀k\∈[1...P_{\max }]$

$\left\{\begin{array}{c}{Y}_{u,v,k}^{s,d}\≤Z_k^{s,d}\\ Y_{u,v,k}^{s,d}\≤X_{u,v,k}^{s,d}*\mathrm{\β}\\ Y_{u,v,k}^{s,d}-Z_k^{s,d}\≥(X_{u,v,k}^{s,d}-1)*\mathrm{\β}\end{array}\right.---\left(5\right)$

Linear restriction below can be combined into according to three Linear Constraints in formula (5) and equation (4):

$\underset{v_s\∈V^{\′}}{\mathrm{\Σ}}\underset{v_d\∈V^{\′}}{\mathrm{\Σ}}\underset{k=1}{\overset{P_{\max }}{\mathrm{\Σ}}}{Y}_{u,v,k}^{s,d}\≤B_{(u,v)},\∀e_{(u,v)}\∈E---\left(6\right)$

In the minimum value of formula (7) by the data flow of path right for all " source-target " nodes of definition.

$\mathrm{\ζ}\≤\underset{k=1}{\overset{P_{\max }}{\mathrm{\Σ}}}{Z}_k^{s,d},\∀(s,d)\∈V^{\′2}---\left(7\right)$

Every bar abstracting link e' _{(s, d)}bandwidth be consider in practical topology connect two boundary node (v _s, v _d) aisled bandwidth.It is represented by following formula (8):

${B^{\′}}_{(s,d)}=\underset{k=1}{\overset{P_{\max }}{\mathrm{\Σ}}}{Z}_k^{s,d}---\left(8\right)$

B' _{(s, d)}represent abstract limit (v _s, v _d) amount of bandwidth that has.

Specific algorithm realizes:

Arthmetic statement:

The above-mentioned algorithm first step calculates each to boundary node to (v _s, v _d) all in practical topology link set P _{(s, d)}, as the input of algorithm.Whether the right all links in this border have bandwidth left (Line:1.1.1.1.1) to utilize previously defined Linear Constraints to judge, if there is bandwidth left, then distribute (Line:1.1.1.1.1.1).Otherwise, represent that kth paths not have bandwidth can be distributed (Line:1.1.1.1.1.2).The bandwidth of boundary node virtual link is exactly that this amount of bandwidth available to physical links all between boundary node is added (Line:2.1).Algorithm flow is as Fig. 5.

Three, IT resource is abstract

Node in abstract network topology is actually the result by some internal nodes and certain boundary node merger, for the ease of difference, is referred to as herein bunch (cluster).Internal node is integrated into boundary node and forms bunch process as shown in Figure 6.The merger problem of internal node and internal node should be integrated into the problem on which boundary node.The merger of node mainly shields transmission bandwidth therebetween.In order to avoid the waste of IT resource, herein the main thought of IT Resource Abstract algorithm to calculate in optical-fiber network topology each internal node to the bandwidth value of the shortest path of all boundary nodes, this value is integrated into the foundation of which boundary node as this internal node, maximum to certain boundary node bandwidth value, be integrated into this boundary node.

First calculate internal node v with dijkstra's algorithm _ito boundary node v _jshortest path p _{(i, j)}.If represent internal node v _ibe integrated into boundary node v _jupper desired value.V _{(i, j)}represent internal node v _ibe integrated into boundary node v _jon.K represents the set of merger result,

Specific algorithm realizes:

Arthmetic statement:

The input of the algorithm first step be the set of boundary node, the set of internal node and each internal node of utilizing dijkstra's algorithm to draw are to the set of minimal paths P of all boundary nodes _{(i, j)}.Desired value by this internal node v _ito boundary node v _j(Bucket Principle, the installation of water of wooden barrel is determined by the shortest plank) (Line:1.1.1) that the limit that in shortest path, bandwidth value is minimum determines.Calculate intermediate node v _imerger to all boundary nodes is expected obtain maximum (Line:1.1.1.2) wherein.Finally obtain the results set K after merger.

The present invention's method of single gathering (single-aggregate) carries out the abstract of IT resource.In physical topology, internal node is integrated in boundary node with above-mentioned algorithm and forms a resource node.The method of single gathering be exactly by merger after boundary node v _nin (bunch), all actual node IT resources are added.Consider calculating and storage capacity (the C ∪ S={v of these resource points herein _n∈ V'}).Abstract node is polymerized by multiple actual node, if each abstract calculating and memory node v _nall comprise individual actual node.Order represent abstract calculating and memory node v _na jth actual node, use represent actual node process number, order represent the computing capability that in a jth actual node, a kth process has, with MIPS (the average execution speed of byte fixed point instruction) for Measure Indexes.In like manner, use represent actual node the quantity of memory device, represent the storage capacity that a kth memory device has at a jth actual node, in units of GB.So to each boundary node v _nhave:

${\mathrm{\δ}}_n=\underset{j=1}{\overset{N_n^{\mathrm{clust}}}{\mathrm{\Σ}}}\underset{k=1}{\overset{N_j^{\mathrm{proc}}}{\mathrm{\Σ}}}{\mathrm{\δ}}_n^{j,k},{\∀v}_n\∈C---\left(9\right)$

${\mathrm{\φ}}_n=\underset{j=1}{\overset{N_n^{\mathrm{clust}}}{\mathrm{\Σ}}}\underset{k=1}{\overset{N_j^{\mathrm{proc}}}{\mathrm{\Σ}}}{\mathrm{\φ}}_n^{j,k},{\∀v}_n\∈S---\left(10\right)$

C, S are the set representing abstract posterior nodal point, comprise the computing capability and storage capacity these two that calculate abstract node respectively with the calculating that C, S are explanation IT resource.δ _nand φ _nrepresent abstract node v respectively _ntotal computing capability and total storage capacity.

The performance evaluation of algorithm:

The Performance evaluation criterion of algorithm is divided into two parts: the utilance of resource and the precision of Resourse Distribute.

The precision of Resourse Distribute: the target of ISP (SP) is for user (USER) provides maximum resource service, in service plane, ISP (SP) distributes according to the Resource Abstract data of network provider (NP), therefore can not ensure that NP has corresponding resource to meet whole request of SP acceptance.If R' _afor the number of requests that SP accepts, R _arepresent the number of requests that NP has respective resources to meet.But definition " refusal returns " variable θ represents the request that SP accepts the number of requests refused by NP.There is equation below:

θ＝R′ _a-R _a(11)

The precision of abstract algorithm Resourse Distribute is quantized, reject rate formulae discovery the following with the reject rate of network provider (NP):

$\widetilde{\mathrm{\θ}}=\frac{\mathrm{\θ}}{R_a^{\′}}=\frac{R_a^{\′}-R_a}{R_a^{\′}}---\left(12\right)$

The actual bandwidth total amount of resource utilization: NP is a fixed value, is set to N, definition n _krepresent kth amount of bandwidth (the k ∈ [1 that request takies that NP accepts ... R _a]).The computing formula of resource utilization is:

$\mathrm{\η}=\frac{\underset{k=1}{\overset{R_a}{\mathrm{\Σ}}}{n}_k}{N}---\left(13\right)$

Experiment conclusion

The realization of algorithm: with the node of adjacency list storage networking topology, with the amount of bandwidth between adjacency matrix memory node.In order to verify different complexity topological diagram in the performance of algorithm, respectively the topological diagram of three differing complexities (difference of internal node number) is verified.After inputting a topological diagram, program draws abstract topology, then generates one group of request by computer random, and program carries out Resourse Distribute according to abstract topology, by the request and the practical topology comparison that accept, satisfied then distribute respective resources, satisfiedly then returns refusal.One take turns distribution after, the abstract of a new round is carried out to resource remaining in practical topology and again draws abstract topology, then stochastic generation one group of request.Until bandwidth resources are used up in abstract topology.Experiment obtains the amount of bandwidth of distributing, and refusal returns (crank-back) rate.Contrast algorithm is the SILK algorithm that current application is the widest, and data as shown in Figure 7, Figure 8.Fig. 7, Fig. 8 abscissa all represents the topological diagram of three different complexities, and the former ordinate represents that amount of bandwidth unit is GB, and the latter's ordinate represents refusal return rate.Experiment shows: network topology this algorithm more complicated improves more obvious relative to the performance of SILK algorithm.

Claims (3)

1. the resource based on optical-fiber network framework service plane quantizes a distribution method, it is characterized in that comprising the following steps:

(1) physical topology is abstract, in physical network topology, is the abstract network topology of only remaining boundary node by physical topology node abatement;

(2) Internet resources is abstract, by link sharing same physical link right for target-source nodes different in abstract network topology, the transfer of data of the link that target-source node is right can select different physical links to carry out, and wherein the data flow of each physical link all can only along same direction;

(3) IT resource is abstract, and in Computational Physics network topology, each internal node is to the bandwidth value of the shortest path of all boundary nodes, internal node is integrated into the boundary node corresponding to maximum of described bandwidth value.

2. a kind of resource based on optical-fiber network framework service plane quantizes distribution method according to claim 1, it is characterized in that: the abstract employing of the physical topology described in (1) be the method for Full-mesh.

3. a kind of resource based on optical-fiber network framework service plane quantizes distribution method according to claim 1, it is characterized in that: described Internet resources abstract in also comprise minimum value and the constraints of the data flow calculating the right link of all target-source nodes, be expressed as follows respectively:

$\mathrm{\ζ}\≤\underset{k=1}{\overset{P_{\max }}{\mathrm{\Σ}}}{Z}_k^{s,d},\∀(s,d)\∈V^{\′2}$

$\underset{v_s\∈V^{\′}}{\mathrm{\Σ}}\underset{v_d\∈V^{\′}}{\mathrm{\Σ}}\underset{k=1}{\overset{P_{\max }}{\mathrm{\Σ}}}{Y}_{u,v,k}^{s,d}\≤B_{(u,v)},\∀e_{(u,v)}\∈E$

In formula: ζ represents the minimum value of the data flow of the link that all target-source nodes are right, represent the amount of bandwidth that the kth bar path from source node s to destination node d can provide, P _maxrepresent that target-source node is to the number of path, V' represents the set of the boundary node of physical network topology, represent the limit e of process the kth bar path from source node s to destination node d _{(u, v)}bandwidth, e _{(u, v)}represent actual physics topology limit, B _{(u, v)}represent actual physics topology limit e _{(u, v)}bandwidth capacity, v _sand v _dall represent the abstract node of abstract network topology, (v _s, v _d) representing source-destination node pair, E represents the set on actual physics topology limit.