CN105187322A - Virtual network mapping method based on load balancing - Google Patents

Abstract

The present invention disclose a virtual network mapping method based on load balancing. The method of the present invention comprises a step 1 of establishing a two dimensional load balancing model of the time and the resource; a step 2 of establishing a mathematical programming model having the minimum two dimensional load balancing degree; a step 3 of mapping a virtual network based on the two dimensional load balancing model and the mathematical programming model. The two dimensional load balancing model comprises the node two dimensional load intensity wn, the link two dimensional load intensity we, the node two dimensional load balancing performance and the link two dimensional load balancing performance, and the mathematical programming model having the minimum two dimensional load balancing degree comprises a minimization objective function u. The virtual network mapping method based on load balancing of the present invention can map the virtual network real-timely, and enables the virtual network mapping success rate to be improved and a bottom-layer physical network to be located in a load balancing state for a long time.

Description

A kind of virtual net mapping method based on load balancing

Technical field

The invention belongs to network virtualization technical field, be specifically related to a kind of virtual net mapping method based on load balancing.The present invention relates to the structure of the virtual net of the two dimensional attributes of time response and spatial character, the angle based on load balancing utilizes the method for discrete weightings to carry out the method for virtual net mapping.

Background technology

Come in the past few decades, the development of the Internet be unprecedented soon, owing to supporting Distributed Application and heterogeneous network technologies, the Internet achieves huge success, becomes the main channel of people's quick obtaining information transmission of information.It has stimulated fast development and the extensive use of network technology greatly.But, because there is multiple supplier the Internet, adopt new network architecture to be not only the change of individual host and route, and need the joint agreement between Internet Service Provider.But Internet provider considers that number one is difficult to reaching an agreement in new technique large scale deployment, therefore also just result in network and to ossify problem.Network virtualization technology provides up-and-coming mode and solves network and to ossify problem.In network virtualization, isomery virtual network is built in the bottom-layer network that multiple service provider leases in one or more infrastructure supplier, and network service is end to end provided, network technology innovation can be carried out and service promotes simultaneously separately to reach.

Along with the rise of network virtualization, researcher proposes a series of virtual net developing algorithm, and these algorithms are mostly all based on the one-dimensional space being run in same bottom-layer network the application of multiple virtual net simultaneously.Do not consider the impact that time factor maps virtual net, because virtual net request has running time and duration, thus cause bottom-layer network resource to worsen along with the carrying out of time.

The present invention, from the angle of Network Load Balance, proposes the method for the two-dimensional discrete weighting of time and resource load.This method can well metric physical network resource status, makes physical network resource can be in the state of load balancing on the whole always, thus improves network resource utilization.

Summary of the invention

The object of the invention is for the deficiencies in the prior art, propose a kind of virtual net mapping method based on load balancing.

This to solve the technical scheme that adopts of its technical problem as follows:

The two-dimentional Load Balancing Model of step 1, settling time and resource

Step 2, set up the minimized mathematical programming model of two-dimentional load balancing degrees

Step 3, based on two-dimentional Load Balancing Model and mathematical programming model, virtual net to be mapped

The foundation of the two-dimentional Load Balancing Model of the time described in step 1 and resource, specific as follows:

Two dimension Load Balancing Model comprises node two dimension intensity of load w _n, link two dimension intensity of load w _e, node two dimension load equilibrium and link two dimension load equilibrium;

Node two dimension intensity of load w _nbe calculated as follows:

$w_n=\underset{i=1}{\overset{\mathrm{\∞}}{\Σ}}\left(S_n\right(t_i)\×{\∫}_{t_i}^{t_{i+1}}w\left(t\right)dt)$ Formula (1)

Wherein, S _nt () is node load intensity, w (t) is the weight function of a monotone decreasing and meets i is positive integer.

Link two dimension intensity of load w _ebe calculated as follows:

$w_e=\underset{i=1}{\overset{\mathrm{\∞}}{\Σ}}\left(S_e\right(t_i)\×{\∫}_{t_i}^{t_{i+1}}w\left(t\right)dt)$ Formula (2)

Wherein S _et () represents intensity of load on t link, w (t) is the weight function of a monotone decreasing and meets i is positive integer.

Node two dimension load equilibrium N _σbe calculated as follows:

$N_{\mathrm{\σ}}=\sqrt{\frac{\underset{\left|N_s\right|}{\Σ}{\left(w_n\right(N_s)-N_{avg})}^2}{\left|N_s\right|}}$ Formula (3)

Wherein, n _σrepresent node two dimension load equilibrium on physical network, particularly the standard deviation of each node two dimension intensity of load, N _srepresent physical node.

Link two dimension load equilibrium L _σbe calculated as follows:

$L_{\mathrm{\σ}}=\sqrt{\frac{\underset{\left|E_s\right|}{\Σ}{\left(w_e\right(E_s)-L_{avg})}^2}{\left|E_s\right|}}$ Formula (4)

Wherein, l _σrepresent physical network uplink two dimension load equilibrium, particularly the standard deviation of each link two dimension intensity of load, E _srepresent physical link.

The minimized mathematical programming model of two-dimentional load balancing degrees described in step 2 comprises and minimizes target function u:

Minu=α N _σ+ β L _σformula (5)

Wherein, α, β are the parameter factors of 0-1;

Described in step 3 based on two-dimentional Load Balancing Model and mathematical programming model, map virtual net, detailed process is as follows:

The dummy node of virtual request is pressed the non-increasing arrangement of cpu resource by 3-1., maps successively to all dummy nodes, dummy node is mapped to the node two dimension intensity of load w meeting resource constraint _non minimum physical node;

Described resource constraint is namely the cpu resource of dummy node request is less than the remaining cpu resource of physical node.

The virtual link of virtual request is pressed the non-increasing arrangement of bandwidth constraint by 3-2., successively all virtual links are mapped, from bottom-layer network, find out current virtual link meet bandwidth constraint and the K shortest path that can map, current virtual link maps is concentrated link two dimension intensity of load w to path _eon minimum path.

Described K is positive integer, and span is 3-7;

Described bandwidth constraint is namely the bandwidth resources of virtual link request are less than the remaining bandwidth resources of physical link.

3-3. based on the mapping of step 3-1 and 3-2, according to propose two-dimentional load balancing degrees minimized mathematical programming model computational minimization target function u.

3-4. is based on the mapping of step 3-1 and 3-2, and each dummy node completes following process successively:

Find out the physical node set M be not assigned with meeting the constraint of current virtual node resource, if after the physical node that a physical node replacement current virtual node in set M maps, the value minimizing target function u can be made to reduce, then with the physical node that this physical node replacement current virtual node maps.

Beneficial effect of the present invention is as follows:

The present invention can map virtual net in real time, can improve virtual net and be mapped to power, and makes bottom physical network be in the state of load balancing for a long time.

Accompanying drawing explanation

Fig. 1 is virtual net map example schematic diagram of the present invention

Fig. 2 is discrete weightings right value function schematic diagram

Embodiment

Below in conjunction with accompanying drawing, the present invention is further illustrated.

Based on a virtual net mapping method for load balancing, comprise the steps:

The two-dimentional Load Balancing Model of step 1, settling time and resource

Step 2, set up the minimized mathematical programming model of two-dimentional load balancing degrees

Step 3, based on two-dimentional Load Balancing Model and mathematical programming model, virtual net to be mapped

The foundation of the two-dimentional Load Balancing Model of the time described in step 1 and resource, specific as follows:

Two dimension Load Balancing Model comprises node two dimension intensity of load w _n, link two dimension intensity of load w _e, node two dimension load equilibrium and link two dimension load equilibrium;

Node two dimension intensity of load w _nbe calculated as follows:

$w_n=\underset{i=1}{\overset{\mathrm{\∞}}{\Σ}}\left(S_n\right(t_i)\×{\∫}_{t_i}^{t_{i+1}}w\left(t\right)dt)$ Formula (1)

Wherein, S _nt () is node load intensity, w (t) is the weight function of a monotone decreasing and meets i is positive integer.

Link two dimension intensity of load w _ebe calculated as follows:

$w_e=\underset{i=1}{\overset{\mathrm{\∞}}{\Σ}}\left(S_e\right(t_i)\×{\∫}_{t_i}^{t_{i+1}}w\left(t\right)dt)$ Formula (2)

Wherein S _et () represents intensity of load on t link, w (t) is the weight function of a monotone decreasing and meets i is positive integer.

Node two dimension load equilibrium N _σbe calculated as follows:

$N_{\mathrm{\σ}}=\sqrt{\frac{\underset{\left|N_s\right|}{\Σ}{\left(w_n\right(N_s)-N_{avg})}^2}{\left|N_s\right|}}$ Formula (3)

Wherein, n _σrepresent node two dimension load equilibrium on physical network, particularly the standard deviation of each node two dimension intensity of load, N _srepresent physical node.

Link two dimension load equilibrium L _σbe calculated as follows:

$L_{\mathrm{\σ}}=\sqrt{\frac{\underset{\left|E_s\right|}{\Σ}{\left(w_e\right(E_s)-L_{avg})}^2}{\left|E_s\right|}}$ Formula (4)

Wherein, l _σrepresent physical network uplink two dimension load equilibrium, particularly the standard deviation of each link two dimension intensity of load, E _srepresent physical link.

The minimized mathematical programming model of two-dimentional load balancing degrees described in step 2 comprises and minimizes target function u:

Minu=α N _σ+ β L _σformula (5)

Wherein, α, β are the parameter factors of 0-1;

Described in step 3 based on two-dimentional Load Balancing Model and mathematical programming model, map virtual net, detailed process is as follows:

The dummy node of virtual request is pressed the non-increasing arrangement of cpu resource by 3-1., maps successively to all dummy nodes, dummy node is mapped to the node two dimension intensity of load w meeting resource constraint _non minimum physical node;

Described resource constraint is namely the cpu resource of dummy node request is less than the remaining cpu resource of physical node.

The virtual link of virtual request is pressed the non-increasing arrangement of bandwidth constraint by 3-2., successively all virtual links are mapped, from bottom-layer network, find out current virtual link meet bandwidth constraint and the K shortest path that can map, current virtual link maps is concentrated link two dimension intensity of load w to path _eon minimum path.

Described K is positive integer, and span is 3-7;

Described bandwidth constraint is namely the bandwidth resources of virtual link request are less than the remaining bandwidth resources of physical link.

3-3. based on the mapping of step 3-1 and 3-2, according to propose two-dimentional load balancing degrees minimized mathematical programming model computational minimization target function u.

3-4. is based on the mapping of step 3-1 and 3-2, and each dummy node completes following process successively:

Find out the physical node set M be not assigned with meeting the constraint of current virtual node resource, if after the physical node that a physical node replacement current virtual node in set M maps, the value minimizing target function u can be made to reduce, then with the physical node that this physical node replacement current virtual node maps.

Embodiment:

As shown in Figure 1, virtual network requests 1 interior joint mapping scheme is { a → A, b → B}, link maps scheme is { (a, b) → (A, B) }, mapping scheme in virtual net request 2 is { c → D, d → C, f → F}, link maps scheme is { (c, d) → (D, B, C), (c, f) → (D, E, F) }. { c → D, d → A can be changed into by mapping scheme after amendment iteration map, f → C}, link maps scheme is { (c, d) → (D, B, A), (c, f) → (D, E, C) }.

We simply can not try to achieve the intensity of load phase Calais of timeslice each in a period of time the metric of load balancing degrees.Preassignment because we have just started virtual net resource, if we need when preallocated resource shifts, from current time more close to the cost that shifts will be higher.As shown in Figure 2, the method for discrete weightings can be adopted, introduce function w (t) of a monotone decreasing, and this function meets and using each timeslice beginning and ending time as range of integration, the value that resource load degree product weight function being carried out to integration and this timeslice is obtained as the weights of timeslice intensity of load, and based on

This is weighted summation, thus obtains the two-dimentional intensity of load within this virtual net operation beginning and ending time.

Claims (4)

1., based on a virtual net mapping method for load balancing, it is characterized in that comprising the steps:

The two-dimentional Load Balancing Model of step 1, settling time and resource;

Step 2, set up the minimized mathematical programming model of two-dimentional load balancing degrees;

Step 3, based on two-dimentional Load Balancing Model and mathematical programming model, virtual net to be mapped.

2. a kind of virtual net mapping method based on load balancing as claimed in claim 1, is characterized in that the foundation of the two-dimentional Load Balancing Model of time described in step 1 and resource, specific as follows:

Two dimension Load Balancing Model comprises node two dimension intensity of load w _n, link two dimension intensity of load w _e, node two dimension load equilibrium and link two dimension load equilibrium;

Node two dimension intensity of load w _nbe calculated as follows:

$w_n=\underset{i=1}{\overset{\mathrm{\∞}}{\Σ}}\left(S_n\right(t_i)\×{\∫}_{t_i}^{t_{i+1}}w(t\left)dt\right)$ Formula (1)

Wherein, S _nt () is node load intensity, w (t) is the weight function of a monotone decreasing and meets i is positive integer;

Link two dimension intensity of load w _ebe calculated as follows:

$w_e=\underset{i=1}{\overset{\mathrm{\∞}}{\Σ}}\left(S_e\right(t_i)\×{\∫}_{t_i}^{t_{i+1}}w(t\left)dt\right)$ Formula (2)

Wherein S _et () represents intensity of load on t link, w (t) is the weight function of a monotone decreasing and meets i is positive integer;

Node two dimension load equilibrium N _σbe calculated as follows:

$N_{\mathrm{\σ}}=\sqrt{\frac{\underset{\left|N_s\right|}{\Σ}{\left(w_n\right(N_s)-N_{avg})}^2}{\left|N_s\right|}}$ Formula (3)

Wherein, n _σrepresent node two dimension load equilibrium on physical network, particularly the standard deviation of each node two dimension intensity of load, N _srepresent physical node;

Link two dimension load equilibrium L _σbe calculated as follows:

$L_{\mathrm{\σ}}=\sqrt{\frac{\underset{\left|E_s\right|}{\Σ}{\left(w_e\right(E_s)-L_{avg})}^2}{\left|E_s\right|}}$ Formula (4)

Wherein, l _σrepresent physical network uplink two dimension load equilibrium, particularly the standard deviation of each link two dimension intensity of load, E _srepresent physical link.

3. a kind of virtual net mapping method based on load balancing as claimed in claim 1, the minimized mathematical programming model of two-dimentional load balancing degrees that it is characterized in that described in step 2 comprises and minimizes target function u:

Minu=α N _σ+ β L _σformula (5)

Wherein, α, β are the parameter factors of 0-1.

4. a kind of virtual net mapping method based on load balancing as claimed in claim 1, it is characterized in that described in step 3 based on two-dimentional Load Balancing Model and mathematical programming model, map virtual net, detailed process is as follows:

The dummy node of virtual request is pressed the non-increasing arrangement of cpu resource by 3-1., maps successively to all dummy nodes, dummy node is mapped to the node two dimension intensity of load w meeting resource constraint _non minimum physical node;

Described resource constraint is namely the cpu resource of dummy node request is less than the remaining cpu resource of physical node;

The virtual link of virtual request is pressed the non-increasing arrangement of bandwidth constraint by 3-2., successively all virtual links are mapped, from bottom-layer network, find out current virtual link meet bandwidth constraint and the K shortest path that can map, current virtual link maps is concentrated link two dimension intensity of load w to path _eon minimum path;

Described K is positive integer, and span is 3-7;

Described bandwidth constraint is namely the bandwidth resources of virtual link request are less than the remaining bandwidth resources of physical link;

3-3. based on the mapping of step 3-1 and 3-2, according to propose two-dimentional load balancing degrees minimized mathematical programming model computational minimization target function u;

3-4. is based on the mapping of step 3-1 and 3-2, and each dummy node completes following process successively:

Find out the physical node set M be not assigned with meeting the constraint of current virtual node resource, if after the physical node that a physical node replacement current virtual node in set M maps, the value minimizing target function u can be made to reduce, then with the physical node that this physical node replacement current virtual node maps.