CN105159780A - Multi-layer cloud application oriented high-availability virtual network mapping method and apparatus - Google Patents

Abstract

The present invention relates to a multi-layer cloud application oriented high-availability virtual network mapping method and apparatus, and belongs to the technical field of multi-layer cloud application. The method provided by the present invention comprises: firstly, acquiring multi-layer application requests sent by a user; secondly, performing, according to a quantity of virtual machines required by each layer, descending sorting on the requests sent by the user; then, allocating virtual machine resources to each layer in a descending order of the quantity of virtual machines required by each layer, which is specifically: firstly calculating a feasible vector Se and a minimum feasible value, as shown in the specification, and a corresponding target vector Pe and a corresponding target value, as shown in the specification, of all allocation solutions according to a tree-shaped topology structure of a current cloud platform for the quantity of virtual machines required by the request of the layer, and then acquiring an optimal allocation solution of each link for the layer; and finally, outputting the optimal allocation solution of the multi-layer application requests of the user. Compared with the prior art, in the resource allocation solution provided by the present invention, both resource utilization and feasibility are considered, and efficiency of a whole system is improved while feasibility is ensured.

Description

The high availability virtual network mapping method of oriented multilayer time cloud application and device

Technical field

The present invention relates to a kind of mapping method of virtual network and device, in particular to high availability virtual network mapping method and the device of time cloud application of a kind of oriented multilayer, belong to multi-level cloud applied technical field, specifically belong to the fields such as virtual network availability and resource utilization.

Background technology

In recent years, increasing service provider was by among their multi-level application migration to public cloud network.In traditional cloud network environment, cloud provider has physical equipment, can provide virtual network resource to service provider, after each service provider obtains an independent virtual network resource, namely can be the service that each user provides certain.Virtual network embeds and achieves the mapping of virtual network to actual physical networks.In order to realize better application service quality; service provider can have the specific demand of some Resource Guarantee (as resources of virtual machine, bandwidth resources) etc. usually to virtual network, but the inefficacy of some physical networks (as switch, link) often makes a big impact to the availability of service.The more existing method having researched and proposed guarantee data center network availability, such as design a kind of novel topological structure with network redundancy, but data center loses efficacy, the virtual network Resource Availability caused reduced the thing remaining an a lot of people of puzzlement.Therefore, service provider has very large demand for raising virtual network availability and these two targets of guarantee virtual resource.

There is now multiple virtual network to operate on identical equipment simultaneously, so a kind of virtual network of necessary design embeds mechanism, while realizing the efficient utilization of resources, meet the demand of service provider.But because high efficiency and availability itself are conflicting, so to realize these two targets very difficult simultaneously.The virtual network that the present invention proposes high availability embeds framework, improves the availability of resource utilization and system simultaneously.In addition, the present invention can measure resource utilization expense and availability expense respectively, these two expenses is solved as optimization problem.

In multi-level cloud application, the utilization factor of resource and availability are the targets of two contradictions own, consider this two targets simultaneously, and the balance obtaining these two targets has very important significance for multi-level cloud application, but regrettably lack relevant effective solution for the problems referred to above are current.

Summary of the invention

The object of the invention is cannot take into account the target of resource utilization and these two contradictions of availability for solving prior art and carry out the problem of virtual network mapping, the high availability virtual network mapping method of a kind of oriented multilayer time cloud application is proposed, the method effectively can improve availability and the resource utilization of virtual network, meanwhile, the present invention have that dirigibility is high, availability is strong, available and efficient feature.

Thought of the present invention is according to user's request, calculates the resource utilization expense under often kind of Resource Allocation Formula and availability expense respectively, and Resourse Distribute when acquisition expense sum is minimum, is optimal resource allocation scheme of the present invention.

The object of the invention is to be achieved through the following technical solutions:

A high availability virtual network mapping method for oriented multilayer time cloud application, comprises the following steps:

1. service provider (user) proposes the request T={T with the application of k layer to cloud provider _i(1≤i≤k), wherein n _irepresent the quantity of the virtual machine that application i-th layer needs; represent the bandwidth that i-th layer of upper each virtual machine intralayer communication needs; represent the bandwidth that i-th layer of upper each virtual machine interlayer communication needs;

2. cloud provider obtains the tree topology G=(V, E) under current cloud platform, and wherein V represents the set of Servers-all, and E represents the set of all links, root be in E with link; And according to applying the virtual machine quantity N of every layer of demand in request T _icarry out descending sort, and result is put in array T';

3. for the request T of the every one deck in T' _i, calculate the virtual machine feasible vector S of every bar link subtree _ewith minimum feasible value feasible vector S _eexpression as shown in table 1:

Table 1 feasible vector S _e

x	0	1	……	N i
					y	0/1	0/1	……	0/1

Represent the Resource Allocation Formula F of link e subtree, namely when distributing x resources of virtual machine in this link e subtree, whether feasible (y=1 represents feasible, otherwise infeasible); For variable x, if meet x≤r _v, then corresponding y value is 1, r _vrepresent the quantity of server v dead slot; Minimum feasible value when representing y=1, the minimum value of x;

4. according to feasible vector S _ethe virtual resource allocation scheme F of middle correspondence calculates object vector P corresponding to link e _ewith minimum target value

P _e＝C(F)+αA(F)(1)

Wherein, C (F) represents the bandwidth cost sum of data center's structure all links under allocative decision F, is calculated by following formula:

$C\left(F\right)=\underset{e\∈E}{\Σ}{w}_e\frac{b_e}{c_e}---\left(2\right)$

Wherein, w _erepresent the weight that link e occupies, c _erepresent the capacity of link e, b _erepresent the bandwidth demand of link under allocative decision F, calculated by following formula:

$b_e=\underset{i\∈k}{\Σ}\min \left(N_i^{-}{B}_i^{in},N_i^{+}{B}_i^{in}\right)+\min \left(\underset{i\∈k}{\Σ}{N}_i^{-}{B}_i^{out},\underset{i\∈k}{\Σ}{N}_i^{+}{N}_i^{out}\right)---\left(3\right)$

Wherein, represent the i-th layer of virtual machine quantity applied contained in the subtree be connected with G link e, represent in G the i-th layer of virtual machine quantity applied removing the remainder that is connected subtree with link e and contain, represent the subtree that is connected with link e and remain the bandwidth set required for intralayer communication, represent the subtree that is connected with link e and remain the bandwidth set required for interlayer communication; Wherein, the bandwidth distributing to link e is not more than the remaining bandwidth capacity of link e, i.e. b _e≤ r _e, r _erepresent the residual capacity of link e, if b _e> r _e, illustrate that the remaining bandwidth of link e is not enough to meet the demand of user under allocative decision F, then give up the program;

A (F) represents the availability expense of data center's structure under allocative decision F, is calculated by following formula:

$A\left(F\right)=\underset{i\∈k}{\Σ}\underset{j\∈J}{\Σ}{p}_j{\left({\mathrm{\θ}}_{ij}\right)}^2---\left(4\right)$

Wherein, J represents the server set that may lose efficacy in data center, p _j, (j ∈ J) represents the possibility that in J, each server j lost efficacy;

θ _ij＝∑ _v∈Vf _ivz _vj/N _i(5)

θ _ijwhen representing that in J, j lost efficacy, i-th layer of resources of virtual machine quantity of distributing in j takies the ratio of the i-th layer of resource requirement in family, f _ivrepresent to the quantity of the virtual resource of the i-th Layer assignment of application in server v, z _vjrepresent whether server v is j, therefore z _vjvalue is 0 or 1; As v=j, z _vj=1; As v ≠ j, z _vj=0; α is the parameter of availability expense A (F), is specified by cloud provider; Minimum target value for object vector P _eminimum value;

5. cloud provider is according to feasible vector S _ewith minimum feasible value and object vector P _ewith minimum target value i-th layer of optimum virtual resource allocation scheme is obtained by traversal method;

As preferably, described traversal method is recursion method alloc (r, e, n), and the method obtains the virtual resource quantity of distributing to i-th layer of application, and the N asked with user _icontrast, if both are equal, then return true, the allocative decision that existence i-th layer is applied is described; If both are unequal, then return false, the allocative decision that there is not i-th layer of application is described; Wherein, parameter r represents that user asks, and parameter e represents link, and parameter n represents the virtual resource quantity of user's request dispatching; Further, using with the original execution link of link root as method alloc, by T _i, root, N _iin substitution method alloc (r, e, n); The workflow of alloc (r, e, n) is as follows:

(1) if level (e)=0 (level represents the level of link e in the data residing for the heart, and during level (e)=0, link e is directly connected with server), the virtual resource quantity of distribution is represent as object vector P _eget minimum value time corresponding l value, judge that doneVM is at feasible vector S _ewhether legal (namely as x=doneVM, whether y is 1), if illegal, then-1 is returned; If legal, then return doneVM value, in namely give i-th layer to be applied in server v that link e is connected, quantity allotted is the resources of virtual machine of l, and mathematical expression is f _iv=l;

(2) if level (e) ≠ 0, doneVM ← 0d, on sets eVseMt (e+) t as the link set of mp link e subtree level (e)-1 layer, link e' ∈ set (e), to the every bar link e' in set (e) according to value carries out ascending order arrangement, calculates virtual resource quantity available base, for the every bar link e' in set (e), calculate with tmp=alloc (r, e', n _e'), tmp is the numerical value that recurrence manner of execution alloc (step 5) obtains; If tmp ≠-1, base and doneVM carries out following assignment: with doneVM ← doneVM+tmp; If tmp=-1, then alloc returns-1 value; After executing the every bar link e' in set (e) and operating, judge that doneVM is at feasible vector S _ewhether legal (namely as x=doneVM, whether y is 1), if illegal, then-1 is returned; If legal, then return doneVM value;

6., after execution of step 5, the Resource Allocation Formula { f of virtual network i-th layer of application can be obtained _iv, v ∈ V, i fixes, and performs the request of applying other layer in the 2nd step in T' successively, obtains the Resource Allocation Formula { f of other layer _jv, 1≤j≤k, j ≠ i, v ∈ V, the final virtual network that obtains embeds solution F={f _iv, 1≤i≤k, v ∈ V, f _ivrepresent to the quantity of i-th layer of virtual resource distributed in server v of application.

A high availability virtual network mapping device for oriented multilayer time cloud application, comprise connect successively user's request receiving module, request computing module, resource distribution module and Resourse Distribute output module; User's request receiving module asks T={T for receiving user _i(1≤i≤k), and export this user request to request computing module; Request computing module is used for asking the tree topology under T and current cloud platform to calculate the feasible vector of link and corresponding object vector respectively according to user, and exports user's request and result of calculation to resource distribution module: feasible vector, object vector; Resource distribution module is the certain resources of virtual machine of each Layer assignment of application for calling alloc method according to user's request, feasible vector and object vector, and allocative decision is outputted to Resourse Distribute output module; Resourse Distribute output module is used for Resource Allocation Formula to export to cloud provider.

As preferably, described request computing module is made up of feasible vector computing module, object vector computing module, bandwidth cost computing module and availability overhead computational module further, feasible vector computing module is directly connected with object vector computing module, and object vector computing module is connected with availability overhead computational module with bandwidth cost computing module respectively; Described feasible vector computing module is used for enumerating all possible Resource Allocation Formula under every bar link subtree; Bandwidth cost computing module is for calculating the link bandwidth utilization rate (i.e. bandwidth cost) under certain Resource Allocation Formula; Availability overhead computational module is for calculating under certain Resource Allocation Formula server failure to the impact (i.e. availability expense) of the overall situation; Object vector computing module is used for bandwidth cost and availability expense sum under certain Resource Allocation Formula.

Beneficial effect

Under the prior art, a lot of research is all for raising data center resource utilization factor (i.e. bandwidth cost) or availability expense, the research both not combined, this causes available data center or resource utilization is low, availability is low, and both are difficult to reach optimal balance effect simultaneously.And the high availability virtual network mapping method of a kind of oriented multilayer that the present invention proposes time cloud application, improve availability and the resource utilization of virtual network.Meanwhile, the invention allows for the method for a new quantification multilayer virtual network availability expense, bandwidth sum availability expense is solved as an optimization problem; In addition, because this optimization problem is a np hard problem, so the present invention devises a heuritic approach, this problem can be solved in polynomial time, reduce time complexity.The present invention can solve the resource allocation problem of all tree topology of data center, has the advantages that dirigibility is high; The method that the present invention proposes can be deployed on data center at once, has the advantages that availability is strong; Meanwhile, Resource Allocation Formula of the present invention takes into account resource utilization and availability, while ensure that availability, improves the efficiency of whole system.

Accompanying drawing explanation

Fig. 1 is the virtual resource allocation schematic diagram of low availability, low bandwidth overhead;

Fig. 2 is the virtual resource allocation schematic diagram of high availability, high bandwidth expense;

Fig. 3 is that k layer application layer is interior, interlayer communication bandwidth demand schematic diagram;

The initial primary topology schematic diagram of Tu4Shi data center;

Fig. 5 is the topological structure schematic diagram of data center after Resourse Distribute;

Fig. 6 is the inventive method schematic flow sheet;

Fig. 7 is the inventive method apparatus structure schematic diagram.

Embodiment

For making target of the present invention, technical scheme and advantage are clearly clear and definite, are described in detail embodiments of the invention below in conjunction with accompanying drawing.The present embodiment carries out actual practice with technical scheme of the present invention veritify for instructing, and gives detailed embodiment and concrete operating process simultaneously, but protection scope of the present invention has more than and is limited to following embodiment.

Fig. 1 and Fig. 2 represents the virtual resource allocation schematic diagram of low availability, low bandwidth overhead and high availability, high bandwidth expense respectively.The structure of two figure is identical, all separately containing a core router (Cor), two convergence routers (Agg), four servers, the most multipotency of each server puts four virtual machines, core router connects two convergence routers, and convergence router connects two servers.Lost efficacy if the low availability of Fig. 1 is embodied in server 1, then whole system can paralyse state; The communication that its low bandwidth overhead embodies in the server 1 between four virtual machines can realize at server internal, and need not communicate with the virtual machine in other server, the bandwidth of whole like this communication remains on comparatively low degree.If the high availability of Fig. 2 is embodied in four servers an inefficacy, other three servers still can work, and are unlikely to total system and paralyse; Four virtual machines that its high bandwidth expense is embodied in distribution are distributed in four different servers, and the communication between virtual machine spans different servers, has higher bandwidth cost.

Fig. 3 represents every interior, interlayer communication bandwidth demand from level to level that k layer is applied.For ground floor, intralayer communication is undertaken by virtual switch (Intra-tierVS) in layer, and the bandwidth of needs is interlayer communication is undertaken by interlayer virtual switch (Cross-tierVS), and the bandwidth of needs is

Fig. 4 represents the initial primary topology of data center, and comprise a convergence router (Agg), two servers, each server can hold four virtual machines, and every bar link capacity is 1Gbps.Convergence router connects two servers, and the probability that server 1,2 lost efficacy is respectively 0.01,0.02.

Be described for the implementation process of data center's structure to the inventive method shown in Fig. 4 below: first, in setting formula (5a), parameter alpha is 200, and weight is respectively 1.

Be illustrated in figure 6 the inventive method schematic flow sheet, comprise the following steps:

1. user files a request P, in table 2:

Table 2 user asks P

Parameter <1,0,150> represent that user asks ground floor to need 1 virtual resource, and layer is interior, interlayer communication bandwidth demand difference 0,150Mbps; Parameter <3,0,200> represent that user asks the second layer to need 3 virtual resources, and layer is interior, interlayer communication bandwidth demand difference 0,200Mbps.

2. the virtual resource demand descending sort of pair each layer of request obtains T '=<3,1>.

3., for the 2nd layer, calculate the feasible vector S of every bar link ₁, S ₂, as shown in Table 3, 4:

Table 3 link l ₁feasible vector S ₁

x	0	1	2	3
					y	1	1	1	1

Now, minimum feasible value

Table 4 link l ₂feasible vector S ₂

x	0	1	2	3
					y	1	1	1	1

Now, minimum feasible value

4. the object vector that the feasible vector of each link above-mentioned is corresponding is respectively:

Table 5 object vector P ₁

x	0	1	2	3
					p	16/9	12/5	26/15	4

Now, minimum feasible value

Table 6 object vector P ₂

x	0	1	2	3
					P	4	26/15	12/5	16/9

Now, minimum feasible value

5. by T ₂, root, 3 are updated in alloc method, due to level (root) ≠ 0, set (root)={ l ₁, l ₂, according to ascending order arrangement obtains { l ₁, l ₂, base=3-(0+0)=3, for the every bar link l in set (root) ₁, l ₂, perform following steps: (1) by T ₂, l ₁, 3 are updated in alloc method, due to level (l ₁)=0, so tmp=2, so be 2 in the virtual machine quantity of server 1 distribution, because tmp ≠-1, so base=3-(2-0)=1, doneVM=0+2=2.(2) by T ₂, l ₂, 1 is updated in alloc method, due to level (l ₂)=0, so tmp=1, so be 1 in the virtual machine quantity of server 2 distribution, because tmp ≠-1, so base=1-(1-0)=0, doneVM=2+1=3, finally, alloc (T _i, root, N _i) result that returns is doneVM=3, now meets alloc (T ₂, root, N ₂)=N ₂, so the program is feasible.In summary, for application the 2nd layer, allocative decision is: distribute 2 resources of virtual machine in the server 1, in server 2, distribute 1 resources of virtual machine, i.e. f ₂₁=2, f ₂₂=1.

6., for the 1st layer, calculate the feasible vector S of every bar link ₁, S ₂, as shown in table 7,8:

Table 7 link l ₁feasible vector S ₁

x	0	1
			y	1	1

Now, minimum feasible value

Table 8 link l ₂feasible vector S ₂

x	0	1
			y	1	1

Now, minimum feasible value

The object vector that the feasible vector of each link above-mentioned is corresponding is respectively:

Table 9 object vector P ₁

x	0	1
			p	89/20	49/20

Now, minimum feasible value

Table 10 object vector P ₂

x	0	1
			P	49/20	89/20

Now, minimum feasible value

By T ₁, root, 1 is updated in alloc method, due to level (root) ≠ 0, set (root)={ l ₁, l ₂, according to ascending order arrangement obtains { l ₁, l ₂, base=1-(0+0)=1, for the every bar link l in set (root) ₁, l ₂, perform following steps: (1) by T ₂, l ₁, 1 is updated in alloc method, due to level (l ₁)=0, so tmp=1, so be 1 in the virtual machine quantity of server 1 distribution, because tmp ≠-1, so base=1-(1-0)=0, doneVM=0+1=1.(2) by T ₂, l ₂, 0 is updated in alloc method, due to level (l ₂)=0, so tmp=0, so be 0 in the virtual machine quantity of server 2 distribution, because tmp ≠-1, so base=0-(0-0)=0, doneVM=1+0=1, finally, alloc (T _i, root, N _i) result that returns is doneVM=1, now meets alloc (T ₁, root, N ₁)=N ₁, so the program is feasible.In summary, for application the 1st layer, allocative decision is: distribute 1 resources of virtual machine in the server 1, in server 2, distribute 0 resources of virtual machine, i.e. f ₁₁=1.

7. so allocative decision is: server 1 distributes 1 virtual machine to application ground floor, distribute 2 virtual machines to the application second layer, server 2 distributes 1 virtual machine to the application second layer, is expressed as F={f ₂₁=2, f ₂₂=1, f ₁₁=1}.As shown in Figure 5.

Be illustrated in figure 7 the high availability virtual network mapping apparatus structure schematic diagram of a kind of oriented multilayer time cloud application made based on said method, as can be seen from Figure, apparatus of the present invention comprise connect successively user's request receiving module, request computing module, resource distribution module and Resourse Distribute output module; User's request receiving module asks T={T for receiving user _i(1≤i≤k), and export this user request to request computing module; Request computing module is used for asking the tree topology under T and current cloud platform to calculate the feasible vector of link and corresponding object vector respectively according to user, and exports user's request and result of calculation to resource distribution module: feasible vector, object vector; Resource distribution module is the certain resources of virtual machine of each Layer assignment of application for calling alloc method according to user's request, feasible vector and object vector, and allocative decision is outputted to Resourse Distribute output module; Resourse Distribute output module is used for Resource Allocation Formula to export to cloud provider.

As preferably, described request computing module is made up of feasible vector computing module, object vector computing module, bandwidth cost computing module and availability overhead computational module further, feasible vector computing module is directly connected with object vector computing module, and object vector computing module is connected with availability overhead computational module with bandwidth cost computing module respectively; Described feasible vector computing module is used for enumerating all possible Resource Allocation Formula under every bar link subtree; Bandwidth cost computing module is for calculating the link bandwidth utilization rate (i.e. bandwidth cost) under certain Resource Allocation Formula; Availability overhead computational module is for calculating under certain Resource Allocation Formula server failure to the impact (i.e. availability expense) of the overall situation; Object vector computing module is used for bandwidth cost and availability expense sum under certain Resource Allocation Formula.

The bandwidth cost solution of the link that the present invention proposes: by distributing the ratio of bandwidth in link, and giving link certain weight, truly reflecting the link bandwidth expense in data center's tree topology; The availability expense solution proposed: by the square proportional that virtual resource is affected by server failure, the impact because server failure brings can be distinguished better, conform to actual conditions.

Above-described specific descriptions; further detailed description has all been carried out to the object of inventing, technical scheme and advantage benefit; be understood that; the foregoing is only specific embodiments of the invention; the protection domain be not intended to limit the present invention; within the spirit and principles in the present invention all, any amendment made, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (4)

1. a high availability virtual network mapping method for oriented multilayer time cloud application, is characterized in that: comprise the following steps:

(1) service provider (user) proposes the request T={T with the application of k layer to cloud provider _i(1≤i≤k), wherein n _irepresent the quantity of the virtual machine that application i-th layer needs; represent the bandwidth that i-th layer of upper each virtual machine intralayer communication needs; represent the bandwidth that i-th layer of upper each virtual machine interlayer communication needs;

(2) cloud provider obtains the tree topology G=(V, E) under current cloud platform, and wherein V represents the set of Servers-all, and E represents the set of all links, and this Rigen's link is root; And according to applying the virtual machine quantity N of every layer of demand in request T _icarry out descending sort, and result is put in array T';

(3) for the request T of the every one deck in T' _i, calculate the virtual machine feasible vector S of every bar link subtree _ewith minimum feasible value feasible vector S _erepresent the Resource Allocation Formula F of link e subtree, namely when distributing x resources of virtual machine in this link e subtree, whether feasible (y=1 represents feasible, otherwise infeasible); For variable x, if meet x≤r _v, then corresponding y value is 1, r _vrepresent the quantity of server v dead slot; Minimum feasible value when representing y=1, the minimum value of x;

(4) according to feasible vector S _ethe virtual resource allocation scheme F of middle correspondence calculates object vector P corresponding to link e _ewith minimum target value

P _e＝C(F)+αA(F)(1)

Wherein, C (F) represents the bandwidth cost sum of data center's structure all links under allocative decision F, is calculated by following formula:

$C\left(F\right)=\underset{e\&Element;E}{\&Sigma;}{w}_e\frac{b_e}{c_e}---\left(2\right)$

Wherein, w _erepresent the weight that link e occupies, c _erepresent the capacity of link e, b _erepresent the bandwidth demand of link under allocative decision F, calculated by following formula:

$b_e=\underset{i\&Element;k}{\&Sigma;}\min \left(N_i^{-}{B}_i^{in},N_i^{+}{B}_i^{in}\right)+\min \left(\underset{i\&Element;k}{\&Sigma;}{N}_i^{-}{B}_i^{out},\underset{i\&Element;k}{\&Sigma;}{N}_i^{+}{N}_i^{out}\right)---\left(3\right)$

Wherein, represent the i-th layer of virtual machine quantity applied contained in the subtree be connected with G link e, represent in G the i-th layer of virtual machine quantity applied removing the remainder that is connected subtree with link e and contain, represent the subtree that is connected with link e and remain the bandwidth set required for intralayer communication, represent the subtree that is connected with link e and remain the bandwidth set required for interlayer communication; Wherein, the bandwidth distributing to link e is not more than the remaining bandwidth capacity of link e, i.e. b _e≤ r _e, r _erepresent the residual capacity of link e, if b _e> r _e, illustrate that the remaining bandwidth of link e is not enough to meet the demand of user under allocative decision F, then give up the program;

A (F) represents the availability expense of data center's structure under allocative decision F, is calculated by following formula:

$A\left(F\right)=\underset{i\&Element;k}{\&Sigma;}\underset{j\&Element;J}{\&Sigma;}{p}_j{\left({\mathrm{\&theta;}}_{ij}\right)}^2---\left(4\right)$

Wherein, J represents the server set that may lose efficacy in data center, represent the possibility that in J, each server j lost efficacy;

θ _ij＝Σ _v∈Vf _ivz _vj/N _i(5)

θ _ijwhen representing that in J, j lost efficacy, i-th layer of resources of virtual machine quantity of distributing in j takies the ratio of the i-th layer of resource requirement in family, f _ivrepresent to the quantity of the virtual resource of the i-th Layer assignment of application in server v, z _vjrepresent whether server v is j, therefore z _vjvalue is 0 or 1; As v=j, z _vj=1; As v ≠ j, z _vj=0;

α is the parameter of availability expense A (F), is specified by cloud provider; Minimum target value for object vector P _eminimum value;

(5) cloud provider is according to i-th layer of feasible vector S _ewith minimum feasible value and object vector P _ewith minimum target value i-th layer of optimum virtual resource allocation scheme is obtained by traversal method;

(6), after execution of step 5, the Resource Allocation Formula { f of virtual network i-th layer of application can be obtained _iv, v ∈ V, i fixes, and performs the request of applying other layer in the 2nd step in T' successively, obtains the Resource Allocation Formula { f of other layer _jv, 1≤j≤k, j ≠ i, v ∈ V, the final virtual network that obtains embeds solution F={f _iv, 1≤i≤k, v ∈ V, f _ivrepresent to the quantity of i-th layer of virtual resource distributed in server v of application.

2. the high availability virtual network mapping method of a kind of oriented multilayer according to claim 1 time cloud application, it is characterized in that: traversal method described in step 5 is recursion method alloc (r, e, n), the method obtains the virtual resource quantity of distributing to i-th layer of application, and the N asked with user _icontrast, if both are equal, then return true, the allocative decision that existence i-th layer is applied is described; If both are unequal, then return false, the allocative decision that there is not i-th layer of application is described; Wherein, parameter r represents that user asks, and parameter e represents link, and parameter n represents the virtual resource quantity of user's request dispatching; Its specific implementation process is as follows:

(1) if level (e)=0 (level represents the level of link e in the data residing for the heart, and during level (e)=0, link e is directly connected with server), the virtual resource quantity of distribution is $doneVM={\mathrm{argmin}}_{S_e^{\min }\&le;l\&le;n}\left\{P_e\left(l\right)\right|S_e\left(l\right)=1\},$ Represent as object vector P _eget minimum value time corresponding l value, judge that doneVM is at feasible vector S _ewhether legal (namely as x=doneVM, whether y is 1), if illegal, then-1 is returned; If legal, then return doneVM value, in namely give i-th layer to be applied in server v that link e is connected, quantity allotted is the resources of virtual machine of l, and mathematical expression is f _iv=l;

(2) if level (e) ≠ 0, if set (e) is the link set of link e subtree level (e)-1 layer, link e' ∈ set (e), to the every bar link e' in set (e) according to value carries out ascending order arrangement, calculates virtual resource quantity available base, for the every bar link e' in set (e), calculate with tmp=alloc (r, e', n _e'), tmp is that recurrence manner of execution alloc (step 5) obtains numerical value; If tmp ≠-1, base and doneVM carries out following assignment: with doneVM ← doneVM+tmp; If tmp=-1, then alloc returns-1 value; After executing the every bar link e' in set (e) and operating, judge that doneVM is at feasible vector S _ewhether legal (namely as x=doneVM, whether y is 1), if illegal, then-1 is returned; If legal, then return doneVM value.

3. a high availability virtual network mapping device for oriented multilayer time cloud application, is characterized in that: the user's request receiving module connected successively, request computing module, resource distribution module and Resourse Distribute output module; User's request receiving module asks T={T for receiving user _i(1≤i≤k), and export this user request to request computing module; Request computing module is used for asking the tree topology under T and current cloud platform to calculate the feasible vector of link and corresponding object vector respectively according to user, and exports user's request and result of calculation to resource distribution module: feasible vector, object vector; Resource distribution module is the certain resources of virtual machine of each Layer assignment of application for calling alloc method according to user's request, feasible vector and object vector, and allocative decision is outputted to Resourse Distribute output module; Resourse Distribute output module is used for Resource Allocation Formula to export to cloud provider.

4. the high availability virtual network mapping device of a kind of oriented multilayer according to claim 3 time cloud application, it is characterized in that: described request computing module is made up of feasible vector computing module, object vector computing module, bandwidth cost computing module and availability overhead computational module further, feasible vector computing module is directly connected with object vector computing module, and object vector computing module is connected with availability overhead computational module with bandwidth cost computing module respectively; Described feasible vector computing module is used for enumerating all possible Resource Allocation Formula under every bar link subtree; Bandwidth cost computing module is for calculating the link bandwidth utilization rate (i.e. bandwidth cost) under certain Resource Allocation Formula; Availability overhead computational module is for calculating under certain Resource Allocation Formula server failure to the impact (i.e. availability expense) of the overall situation; Object vector computing module is used for bandwidth cost and availability expense sum under certain Resource Allocation Formula.