WO2016082867A1

WO2016082867A1 - Orchestrator and method for virtual network embedding

Info

Publication number: WO2016082867A1
Application number: PCT/EP2014/075543
Authority: WO
Inventors: Riccardo GUERZONI; Zoran Despotovic; Sergio Beker
Original assignee: Huawei Technologies Co., Ltd
Priority date: 2014-11-25
Filing date: 2014-11-25
Publication date: 2016-06-02
Also published as: EP3213461A1

Abstract

A virtual network embedding orchestrator (10) for orchestrating embedding of virtual networks into a physical network is provided. The virtual network embedding orchestrator (10) comprises a problem formulation unit (11), which is adapted to formulate an embedding problem (13) based upon information regarding the virtual networks, information regarding the physical network (12) at input parameters (17). Moreover, the problem formulation unit (11) is adapted to provide the embedding problem (13) to an external problem solving unit. The virtual network embedding orchestrator (10) moreover comprises a feedback processing unit (14) adapted to receive embedding results (16) and at least one output parameter (16) from the problem solving unit and determine at least one input parameter (17) of the problem formulation unit (11) based upon the embedding results (16) and the at least one output parameter (16) of the problem solving unit. The solver can a MIP solver. The at least one input parameter can be timeout, MIP gap or solving method.

Description

Orchestrator and method for virtual network embedding

TECHNICAL FIELD

The invention relates to embedding virtual networks into a physical network. BACKGROUND When embedding a number of virtual networks into a physical network, conventionally, an input problem comprising information regarding the virtual networks, the physical networks and one or more input parameters is formulated by a user and provided to a solver unit, which calculate a possible network embedding. On the one hand, the quality of the embedding solution within given computation time constraints might be too low, resulting in a sub- optimal network utilization of the physical network. On the other hand, the computation time needed to reach a high quality solution might be too high. At present, it is up to the user to specify the input problem based upon his experience for achieving an acceptable embedding quality and computation time. Therefore, a great user experience is needed while still only a sub-optimal embedding result can reliably be achieved.

SUMMARY Accordingly, an object of the present invention is to provide an apparatus and method, which allow for an increase in embedding quality, a decrease in computation time and at the same time require very low user experience.

The object is solved by the features of claim 1 for the apparatus and by the features of claim 13 for the method. Further it is solved by the features of claim 15 for the associated computer program. The dependent claims contain further developments.

According to a first aspect of the present invention, a virtual network embedding orchestrator for orchestrating embedding of virtual networks into a physical network is provided. The virtual network embedding orchestrator comprises a problem formulation unit, which is adapted to formulate an embedding problem based upon information regarding the virtual networks, information regarding the physical network at input parameters. Moreover, the problem formulation unit is adapted to provide the embedding problem to an external problem solving unit. It is pointed out that the term "external" here means that the problem solving unit is not part of the embodiment of the virtual network embedding orchestrator, but merely interacts therewith. The virtual network embedding orchestrator moreover comprises a feedback processing unit adapted to receive embedding results and at least one output parameter from the problem solving unit and determine at least one of the input parameters of the problem formulation unit based upon the embedding results and the at least one output parameter of the problem solving unit. Thereby, a feedback loop between the output of the problem solving unit and the input of the problem formulation unit is achieved. With minimal user experience requirements, an increase of the embedding quality and reduction of computational complexity can be achieved.

According to a first implementation form of the virtual network embedding orchestrator according to the first aspect, the feedback processing unit is adapted to determine if the received embedding results are final embedding results, based upon the received output parameters, and provide the determined at least one input parameter to the problem formulation unit, if the received embedding results are not final embedding results, triggering the problem formulation unit to formulate a new embedding problem. This significantly limits the computational complexity, since the further iterations are stopped, as soon as the final embedding results have been determined. According to a second implementation form of the virtual network embedding orchestrator according to the first aspect or the first implementation form of the first aspect, the information regarding the physical network comprises information regarding nodes and physical links connecting the nodes. Additionally or alternatively, the information regarding the virtual networks comprises information regarding at least one virtual link connecting at least two virtual nodes. Additionally or alternatively, the embedding results comprise a mapping of the physical links of the physical network to the virtual links of the virtual networks and/or a mapping of the physical nodes of the physical network to the virtual nodes of the virtual networks. A very accurate and time efficient determination of the embedding results is thereby possible.

According to a third implementation form of the virtual network embedding orchestrator according to the first aspect or one of the previously described implementation forms of the first aspect, the at least one output parameter of the problem solving unit received and processed by the feedback processing unit comprises at least one problem solving quality parameter and/or a problem solving time duration parameter. An optimization of the embedding regarding a problem solving quality and regarding a problem solving time duration can thereby be achieved. According to a fourth implementation form of the virtual network embedding orchestrator according to the third implementation form of the first aspect, the at least one problem solving quality parameter comprises an average number on links of the physical network used for the links of the virtual network and/or a fraction of utilization of the links and nodes of the physical network, and/or an amount of virtual networks embedded into the physical network. A very accurate optimization of the embedding can thereby be achieved.

According to a fifth implementation form of the virtual network embedding orchestrator according to the third of fourth implementation form of the first aspect, the feedback processing unit is adapted to determine the at least one input parameter of the problem formulation unit by comparing the at least one problem solving quality parameter to at least one quality threshold and amending the at least one input parameter to increase a problem solving quality, if the at least one problem solving quality parameter is below the at least one quality threshold. Additionally or alternatively, the feedback processing unit is adapted to determine the at least one input parameter of the problem formulation unit by comparing the problem solving time duration parameter to a time duration threshold, and amending the at least one input parameter to decrease a problem solving time duration, if the problem solving time duration parameter is above the time duration threshold. It is thereby possible to automatically - without user influence - change the input problem in order to optimize the embedding.

According to a sixth implementation form of the virtual network embedding orchestrator according to the first aspect or any of the previously described implementation forms of the first aspect, the at least one input parameter of the problem formulation unit comprises an amount of virtual networks for which the embedding into the physical network is to be calculated simultaneously. Thereby, a strong influence on the problem solving quality parameter and the problem solving time duration parameter is achieved.

According to a seventh implementation form of the virtual network embedding orchestrator according to the first aspect or any of the previously described implementation forms of the first aspect, the at least one input parameter of the problem formulation unit comprises a weighting parameter for weighting costs of a solution of the problem solving by the problem solving unit against benefits of the solution of the problem solving by the problem solving unit. The problem formulation unit is than adapted to formulate the embedding problem to incorporate the weighting parameter. The costs of a solution of the problem solving by the problem solving unit are in this case based on the infrastructure utilization of the physical network. The benefits of a solution of the problem solving by the problem solving unit are in this case based on an amount of virtual networks embedded into the physical network by the embedding solution. A very strong influence of the input parameter on the problem solving quality and time duration is thereby achieved.

According to an eighth implementation form of the virtual network embedding orchestrator according to the first aspect or any of the previously described implementation forms of the first aspect, the virtual network embedding orchestrator further comprises an input buffer adapted to successfully receive and store the information regarding the virtual networks. The problem formulation is then adapted to formulate the embedding problem based upon information regarding the virtual networks stored in the input buffer. A very efficient embedding is thereby achieved.

According to a second aspect of the present invention, a virtual network embedding system is provided. The virtual network embedding system comprises a before-described virtual network embedding orchestrator and a problem solving unit. The problem formulation unit of the virtual network embedding orchestrator is than adapted to provide the embedding problem to the problem solving unit.

According to a first implementation form of the virtual network embedding system according to the second aspect, the problem solving unit is adapted to determine the solution to the embedding problem by integer programming or mixed integer programming. A very efficient problem solution calculation is thereby achieved.

According to a third aspect of the present invention, a virtual network system is provided. The virtual network system comprises a before-described virtual network embedding system and at least two virtual networks embedded into a physical network.

According to a fourth aspect of the present invention, a method for orchestrating embedding of virtual networks into a physical network is provided. The method comprises formulating an embedding problem based upon information regarding the virtual networks, information regarding the physical network and input parameters, providing the embedding problem to a problem solving unit, receiving embedding results and at least one output parameter from the problem solving unit, and determining at least one of the input parameters for formulating the embedding problem based upon the embedding results and the at least one output parameter of the problem solving unit. It is thereby possible to achieve a reduction in computation time, an increase of embedding quality while only requiring minimal user experience.

According to a fist implementation form of the method for orchestrating according to the fourth aspect, the method additionally comprises determining if the received embedding results are not final embedding results based upon the received output parameters, and providing the determined at least one input parameter for formulating the embedding problem, if the received embedding results are not final embedding results. This significantly limits the computational complexity, since the further iterations are stopped, as soon as the final embedding results have been determined. According to a second implementation form of the method for orchestrating according to the fourth aspect or according to the first implementation form of the fourth aspect, the information regarding the physical network comprises information regarding nodes and physical links connecting the nodes and/or the information regarding the virtual networks comprises information regarding at least one virtual link connecting at least two virtual nodes. Additionally or alternatively, the embedding results comprise a mapping of the physical links of the physical network to the virtual links of the virtual networks and/or a mapping of the physical nodes of the physical network to the virtual nodes of the virtual networks. A very accurate and time efficient determination of the embedding results is thereby possible. According to third implementation form of the method for orchestrating according to the fourth aspect or according to any of the preciously described implementation forms of the fourth aspect, the at least one output parameter of the problem solving unit comprises at least one problem solving quality parameter and/or a problem solving time duration parameter. An optimization of the embedding regarding a problem solving quality and regarding a problem solving time duration can thereby be achieved.

According to a fourth implementation form of the method for orchestrating according to the fourth aspect or according to the third implementation form of the fourth aspect, the at least one problem solving quality parameter comprises an average number of links of the physical network used for the links of the virtual networks, and/or a fraction of utilization of the links and nodes of the physical network, and/or an amount of virtual networks embedded into the physical network. A very accurate optimization of the embedding can thereby be achieved.

According to a fifth implementation form of the method for orchestrating according to the fourth aspect or according to the third or fourth implementation form of the fourth aspect, the at least one input parameter of the problem formulation unit is determined by comparing the at least one problem solving quality parameter to at least one quality threshold, and amending the at least one input parameter to increase a problem solving quality, if the at least one problem solving quality parameter is below the at least one quality threshold. Alternatively or additionally, the at least one input parameter of the problem formulation unit is determined by comparing the problem solving time duration parameter to a time duration threshold, and by amending the at least one input parameter to decrease a problem solving time duration, if the problem solving time duration parameter is above the time duration threshold. It is thereby possible to automatically - without user influence - change the input problem in order to optimize the embedding.

According to a sixth implementation form of the method for orchestrating according to the fourth aspect or according to any of the preciously described implementation forms of the fourth aspect, the at least one input parameter of the problem formulation unit comprises an amount of virtual networks for which the embedding into the physical network is to be calculated simultaneously. Thereby, a strong influence on the problem solving quality parameter and the problem solving time duration parameter is achieved.

According to seventh implementation form of the method for orchestrating according to the fourth aspect or according to any of the preciously described implementation forms of the fourth aspect, the at least one input parameter for problem formulation comprises a weighting parameter for weighting costs of a solution of the problem solved by the problem solving unit against benefits of the solution of the problem solving by the problem solving unit. The embedding problem is formulated to incorporate the weighting parameter. The costs of a solution of the problem solving by the problem solving unit are based on the infrastructure utilization of the physical network. The benefits of a solution of the problem solving by the problem solving unit are based on an amount of virtual networks embedded into the physical network by the solution. A very strong influence of the input parameter on the problem solving quality and time duration is thereby achieved.

According to an eighth implementation form of the method for orchestrating according to the fourth aspect or according to any of the preciously described implementation forms of the fourth aspect, information regarding the virtual networks is successively received and stored. During problem formulation, the successfully received and stored information regarding the virtual networks is used. The problem formulation is then adapted to formulate the embedding problem based upon information regarding the virtual networks stored in the input buffer. A very efficient embedding is thereby achieved.

According to fifth aspect of the present invention, a computer program with program code for performing the above-described method when the computer program runs on a computer, is provided. Generally, it has to be noted that all arrangements, devices, elements, units and means and so forth described in the present application could be implemented by software or hardware elements or any kind of combination thereof. Furthermore, the devices may be processors or may comprise processors, wherein the functions of the elements, units and means described in the present applications may be implemented in one or more processors. All steps which are performed by the various entities described in the present application as well as the functionality described to be performed by the various entities are intended to mean that the respective entity is adapted to or configured to perform the respective steps and functionalities. Even if in the following description or specific embodiments, a specific functionality or step to be performed by a general entity is not reflected in the description of a specific detailed element of that entity which performs that specific step or functionality, it should be clear for a skilled person that these methods and functionalities can be implemented in respect of software or hardware elements, or any kind of combination thereof.

BRIEF DESCRIPTION OF DRAWINGS

The present invention is the following explained in detail in relation to embodiments of the invention in reference to the enclosed drawings, in which

Fig. 1 shows a first embodiment of the inventive virtual network embedding orchestrator; Fig. 2 shows virtual networks, a physical network and an exemplary embedding solution; Fig. 3 shows an example of a virtual network embedding system;

Fig. 4 shows a second embodiment of the inventive virtual network embedding orchestrator;

Fig. 5 shows a third embodiment of the virtual network embedding orchestrator;

Fig. 6 shows a fourth embodiment of the virtual network embedding orchestrator;

Fig. 7 shows an exemplary virtual network and an exemplary physical network;

Fig. 8 shows an exemplary embedding solution;

Fig. 9 shows a first time/quality diagram of an embedding problem;

Fig. 10 shows a second time/quality diagram of an embedding problem;

Fig. 11 shows an algorithm of a fifth embodiment of the inventive virtual network embedding orchestrator, and

Fig. 12 shows an embodiment of the method for orchestrating embedding virtual networks into a physical network. DESCRIPTION OF EMBODIMENTS

First, we briefly describe a first embodiment of the inventive virtual network embedding orchestrator along Fig. 1. Afterwards, along Fig. 2 and 3, the underlying problem of network embedding is described. Along Fig. 4-6, further embodiments of the inventive virtual network embedding orchestrator are described in detail. Afterwards, along Fig. 7-10 more details regarding virtual network embedding are given. Along Fig. 11 , further details of an embodiment of the inventive virtual network embedding orchestrator are described. Finally, along Fig. 12, the function of an embodiment of the inventive method is described. Similar entities and reference numbers and different figures have been partially omitted.

In Fig. 1, a first embodiment of the inventive virtual network embedding orchestrator 10 is depicted in a block diagram. The virtual network embedding orchestrator 10 comprises a problem formulation unit 11 which is connected to a feedback processing unit 14 also comprised by the virtual network embedding orchestrator 10.

The problem formulation unit 11 is provided with an input problem. This input problem comprises information regarding virtual networks 12, information regarding a physical network 12, the virtual networks are to be embedded into. The problem formulation unit 11 is adapted to formulate an embedding problem 13 based upon the information regarding the virtual networks 12, and the information regarding the physical network 12 within the input problem. Moreover, the problem formulation unit 11 uses input parameters 17, which are provided by the feedback processing unit 14. As a result, the problem formulation unit 11 provides a formulated embedding problem 13 to a non-depicted external problem solving unit. The problem solving unit then solves the embedding problem and provides embedding results 16 and at least one output parameter 16 to the feedback processing unit 14. The feedback processing unit 14 receives the embedding results 16 and the at least one output parameter 16 and determines at least one input parameter 17 of the problem formulation unit 11 based upon the embedding results and the at least one output parameter 16.

Moreover, the feedback processing unit 14 determines, if the received embedding results 16 are the final embedding results 15 based upon the received output parameters 16. If the received embedding results 16 are final embedding results 15, they are output. In case the received embedding results 16 are not final embedding results, the feedback processing unit 14 provides the input parameter 17 to the problem formulation unit 11, which formulates a further embedding problem 13, so that a further iteration by the problem solution unit is performed. Further details regarding the function of the virtual network embedding orchestrator 10 are given regarding Fig. 4 - Fig. 6.

When embedding virtual networks one aims at mapping multiple virtual networks 20, 21, 22 onto a physical network 24, as illustrated in Fig. 2. A possible embedding 23 is also depicted here.

A physical network 24 advantageously is an undirected graph of nodes and links representing the resources offered by a physical infrastructure. For example, in a physical network representing interconnected hosts, a physical node can be a host with certain computing power, memory and storage space available, while each physical link might have bandwidth and delay as the two relevant properties.

A virtual network 20, 21, 22 is advantageously a directed graph of virtual nodes and links representing the resource requirements of an application that must be implemented in the physical infrastructure. These requirements are seen as constraints that should be fully satisfied when mapping a required virtual network 20, 21, 22 onto the physical network 24. For example, if the requirement for a virtual link is to provide 10Mbps bandwidth, it is not allowed to provide a bandwidth less than 10Mbps by, for example, allocating only 1Mbps bandwidth physical link.

The challenge in virtual network embedding is to allocate resources to as many as possible of these virtual networks 20, 21, 22 while staying within the physical limits, as specified in the physical network 24. Generally speaking, among all possible assignments of virtual links and virtual nodes across all virtual networks 20, 21, 22 to the physical links and physical nodes, we have to find one that optimizes a suitably selected goal function. However, this problem is NP- hard. So we cannot rely on an exhaustive search or expect to find a polynomial time algorithm to solve it. In order to perform such a network embedding, solvers are typically used to determine possible solutions of mixed integer programming (MIP) model instances. The performance of the solver, in terms of convergence to a solution and quality of the solution, depends on (i) the structure of the input problem and (ii) the configuration parameters of the solver. In the following the terms solver, MIP solver and problem solving unit are used interchangeably. Also, the terms problem, embedding problem and MIP embedding problem are also used interchangeably. We point out that the problem solving unit does not necessarily have to use mixed integer programming as problem solving method. Also the use of integer programming or other methods is possible. It is of no avail that the problem solving unit is in some instances referred to as MIP solver. This is not to be understood as limitation to this problem solving method.

Fig. 3 shows the typical use of a MIP solver 32 to determine a solution for an MIP embedding problem 31 provided by a user 30. The MIP embedding problem 31 is described by a set of constraints and a cost function to be maximized. The problem 31 is described using the syntax of the MIP solver 32. The solution delivered by the solver 32 is also influenced by parameter settings 34 of the solver 32, for instance, timeout, MIP gap, solving method, etc.. In order to perform such a problem solving, a great deal of experience of the user 30 is required to define the input problem 31 and to set the solver parameters 34. By use of the inventive virtual network embedding orchestrator, this is no longer necessary.

In Fig. 4, a further embodiment of the virtual network embedding orchestrator 10 is shown. Here, also a problem solving unit 41, which is not part of the virtual network embedding orchestrator 10 is shown. This problem solving unit 41 corresponds to the solver 32 of Fig. 3. The virtual network embedding orchestrator 10 and the problem solving unit 41 form a virtual network embedding system 40, which corresponds to an embodiment of the second aspect of the present invention. In the following, more information regarding the function of the virtual network embedding orchestrator is given. It is also applicable to the embodiments of Fig. 1. The information regarding the physical network 12 comprises information regarding nodes and physical links connecting the nodes. The information regarding the virtual networks 12 comprises information regarding at least one virtual link connecting at least two nodes. The embedding results 16 comprise a mapping of the physical links of the physical network to the virtual links of the virtual networks and/or a mapping of the physical nodes of the physical network to the virtual nodes of the virtual networks.

The at least one output parameter 16 of the problem solving unit 41 received and processed by the feedback processing unit 14 comprises at least one problem solving quality parameter and/or a problem solving time duration parameter.

The at least one problem solving quality parameter comprises an average number of links of the physical network used for the links of the virtual networks and/or a fraction of utilization of the links and nodes of the physical network and/or an amount of virtual networks embedded into the physical network.

The feedback processing unit 14 determines the at least one input parameter of the problem formulation unit 11 by comparing the at least one problem solving quality parameter to at least one quality threshold and accordingly amending the input parameter to increase a problem solving quality parameter, if the at least one problem solving quality parameter is below the at least one quality threshold. Also, the feedback processing unit 14 can determine the at least one input parameter by comparing the problem solving time duration parameter to a time duration threshold and amending the at least one input parameter to decrease a problem solving time duration, if the problem solving time duration parameter is above the time duration threshold.

In case it is necessary to amend one of the parameters in the opposite direction, in order to meet the criteria of the other parameter, this is possible, as long as no threshold is crossed.

Alternatively, the thresholds can also be defined in opposite directions. In this case, it is desirable that the problem solving quality parameter in below the at least one quality threshold. Additionally or alternatively, the thresholds can be defined so that is desirable to keep the time duration parameter above the time duration threshold.

The input parameter of the problem formulation unit in this example is an amount of virtual networks for which the embedding into the physical network is to be calculated

simultaneously. Therefore, the problem formulation unit 11 formulates the embedding problem 13 to incorporate the amount of virtual networks for which the embedding into the physical network is to be calculated simultaneously.

As an alternative input parameter or as an additional input parameter of the problem formulation unit 11 , a weighting parameter for weighting costs of the solution of the problem solving by the problem solving unit 41 against benefits of a solution can be used. The problem formulation unit 11 then formulates the embedding problem to incorporate this weighting parameter. In this case, the costs of a solution of the problem solving are the infrastructure utilization of a physical network and benefits of a solution are a number of virtual networks embedded into the physical network by the according solution.

It is pointed out that an increase of the number of virtual networks for which the embedding is calculated simultaneously results in an increase in problem solving quality while it results in an increase of problem solving time duration. Also, a decrease of this number results in a decrease of problem solving quality and time duration. Moreover, a decrease of the weighting parameter results in an increase of the problem solving quality and in an increase of problem solving time duration, while an increase of the weighting parameter results in a decrease of the problem solving quality and a decrease of the problem solving time duration. The problem formulation unit 11 formulates the embedding problem as a mixed integer program or an integer program, which is solved by the problem solving unit 41. The function of integer programming and mixed integer programming is not shown in detail here. In Fig. 5, a third embodiment of the virtual network embedding orchestrator 10 is shown. Here also a problem solving unit 41 is depicted. In comparison to Fig. 4, here additionally the inputs and outputs are shown in greater detail. Moreover, the virtual network embedding orchestrator 10 here comprises an input buffer 50, which is used for successively receiving and storing the information regarding the virtual networks 12 from the input problem. The problem formulation unit 11 is then adapted to formulate the embedding problem 13 based upon information regarding the virtual networks 12 stored in the input buffer 50. It is thereby possible to start the calculations before all information regarding the virtual networks is present. In Fig. 6, a fourth embodiment of the virtual network embedding orchestrator 10 is shown. Here, the virtual network embedding orchestrator 10 and the problem solving unit 41 form a virtual network embedding system 40. The virtual network embedding system 40, virtual networks 20, 21 and 22 and a physical network 24 form a virtual network system 60. The virtual networks 20, 21 and 22 are embedded into the physical network 24 by the virtual network embedding system 40 resulting in an exemplary embedding 23.

Moreover, the problem formulation unit 11 decides on the correct parameterization of the problem formulation 13. For example, it can decide to submit to the problem solving unit 41 one large problem instance containing all received virtual networks 20, 21 and 22 within a giving time interval or to split the set of received requests into a number of subsets and to submit those to the problem solving unit 41 separately. As another example, there might be parameters in the respective goal function of the embedding problem 13 which can be tuned. The problem formulation unit may select different values for those parameters. For example, G is the total number of instances to be embedded, the requests are embedded sequentially in groups of G = G I r.

In Fig. 7, a virtual network 70 which is to be implemented into a physical network 71 is shown. The virtual network 70 comprises a number of nodes v3g, vlg, v2g, v4g, v5g and a number of links elg, e2g, e3g, e4g, e5g.

The physical network 71 comprises a number of nodes nl, n2, n3, n4 and n5 and a number of links 112, 114, 142, 152, 113, 135.

The embedding problem is characterized by: • Input: one physical network 70 and many virtual network (VN) requests corresponding to a number of virtual networks 70, given in the form of graphs, and

• Output: assign virtual networks 70 to the physical network 71 resources so as to maximize a goal function while fulfilling operational requirements and staying within available capacities of the physical network 71.

The virtual networks 70 are described in the form of directed graphs G_g ^v = (V_g ^v, E_g ^v), where:

• g=l...G and G is the number of Virtual networks simultaneously embedded;

• V_g ^v is a set of H_g virtual nodes Vh_g with h=l...H_g, each one characterized by node capacity requirements;

• E_g ^v is a set of K_g virtual links e_kg with k=l...K_g, each one characterized by link capacity requirements.

The physical network 71 is described in the form of undirected graphs G^P= (V^P, E^p), where:

• V^P is a set of N physical nodes n_; with i=l...N, each one characterized by available node capacity;

• E^p is a set of virtual links ly with i,j=l...N, each one characterized by available link capacity.

The embedding model is characterized by the following variables:

• Virtual node embedding variables y'_hg, Boolean variables equal to 1 if and only if the virtual node Vh_g is embedded in the physical node n_;;

• Virtual link embedding variables x^kgij , Boolean variables equal to 1 if and only if the virtual link e_kg is embedded in the physical node ly;

• Virtual graph embedding variables y_g, Boolean variables equal to 1 if and only if the virtual graph G_g ^v is fully embedded in G^P. The graph is fully embedded when all the virtual nodes and links belonging to the graph are embedded.

In Fig. 8, a virtual network 80 and an according embedding solution into a physical network 81 is depicted. Here, an embedding solution for two virtual nodes VHG, VH'G and one virtual link EKG is shown.

Finally the embedding problem is characterized by a goal function to be maximized. An example of goal function is:

This function is meant to maximize the number of embedded virtual networks, i.e. the summation of y_g, while minimizing the number of physical links involved in the solution, i.e. the summation of x^kgi_j.

The parameter w tunes the balance between revenues (the summation of y_g) and costs (the summation of χ¾. The parameter w corresponds to the earlier-described weighting parameter.

In Fig. 9, a time/quality diagram is depicted. The diagram of Fig. 9 shows a problem instance parameter on the x-axis and time/quality of the y-axis. Assume that the problem solving unit is operating at point A characterized by the value vl of the problem instance parameter or as a set thereof. By changing the problem parameter to v2 (point B), we can heavily reduce the solution computation time, as indicated on the time curve. At the same time, the solution quality is not seriously affected as indicated on the quality curve. For example, Fig. 10 shows the problem size as the problem instance parameter. Assume that the orchestrator has G virtual networks to embed. The orchestrator can formulate the embedding problem such that all 6 requests are submitted simultaneously or it can decide to split the input into a number r of subsets of C ⁼ ^j r requests which are submitted to the solver sequentially. However, this possibility is followed by certain quality losses, i.e. it is to expect that the acceptance ratio (the fraction of the initial ^ request that is actually embedded) will be smaller than if all requests are submitted at once. However, the rate at which the computing time grows with the input size is much higher than the rate of growth of the acceptance rate. It is pointed out that the problem instance parameter corresponds to the at least one input parameter 170f of the problem formulation 11.

In Fig. 11, an algorithm for use by the virtual network embedding orchestrator and the method for embedding virtual networks is shown.

The algorithm shown here controls the input problem size and decides on the best acceptable value, i.e. how to split the original input problem into advantageously equally sized subsets which are sequentially submitted to the problem solving unit.

Essentially, the algorithm checks, if a problem solving by the problem solving unit occurs fast enough. If this is the case, the problem solving can continue with the current parameters. If this is not the case, the number of simultaneously embedded virtual networks is reduced by half. The problem solving is then initiated again.

Another possibility is to tune the goal function to achieve the tradeoffs from Fig. 9 and Fig. 10. Take for example the following goal function:

which is based on the following reasoning. We are interested in accepting as many virtual networks as possible (this goal is captured by variables y_g) while minimizing the infrastructure utilization (variables Now, it turns out that different weights w give rise to very different embedding times, while not affecting the acceptance rate, i.e. the fraction of virtual network requests which can be embedded. High values of w soften the impact of the link utilization in the goal function, so that the solver may accept solutions involving unnecessarily long paths to embed the virtual links. This results in shorter convergence time. Lower values of w yield shorter paths, hence better quality solutions, and longer convergence time. An algorithm similar to that from Fig. 11 can be introduced to select the most suitable value of weight w.

Finally, in Fig. 12, a flow diagram of an embodiment of the method for orchestrating embedding of virtual networks into a physical network is shown. In step 100, information regarding virtual networks and a physical network at input parameters are provides. In step 101, an embedding problem is formulated based upon the provided information regarding the virtual networks and the physical network and based upon the provided input parameters. In step 102, embedding results and at least one output parameter are determined by solving the embedding problem. In step 103, it is determined, if the embedding results are final embedding results. To do so, the output parameter is used. For example, the output parameter is compared to a threshold. If it exceeds the threshold, it is determined that the present embedding results are not final embedding results. In this case, in step 104 at least one input parameter is determined based upon the embedding results and the at least one output parameter. This new input parameter is supplied to step 101, in which a new input problem is formulated.

In case the embedding results determined in step 102 are determined to be final embedding results in step 103, the final embedding results are output in step 105.

The invention has been described in conjunction with various embodiments herein. However, other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure and the appended claims. In the claims, the word "comprising " does not exclude other elements or steps and the indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in usually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. A computer program may be stored/distributed on a suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the internet or other wired or wireless communication systems.

Claims

1. Virtual network embedding orchestrator (10) for orchestrating embedding of virtual networks (20, 21, 22, 70, 80) into a physical network (24, 71, 81), comprising

- a problem formulation unit (11) adapted to

- formulate an embedding problem (13) based upon information regarding the virtual networks (12), information regarding the physical network (12) and input parameters (17), and

- provide the embedding problem (13) to an external problem solving unit (41), and - a feedback processing unit (14) adapted to

- receive embedding results (16) and at least one output parameter (16) from the problem solving unit (41), and

- determine at least one of the input parameters (17) of the problem formulation unit (11) based upon the embedding results (16) and the at least one output parameter (16) of the problem solving unit (41).

2. Virtual network embedding orchestrator (10) according to claim 1,

wherein the feedback processing unit (14) is adapted to

- determine if the received embedding results (16) are final embedding results (15), based upon the received output parameters (16), and

- provide the determined at least one input parameter (17) to the problem formulation unit (11), if the received embedding results (16) are not final embedding results (15), triggering the problem formulation unit (11) to formulate a new embedding problem (13).

3. Virtual network embedding orchestrator (10) according to claim 1 or 2,

wherein the information regarding the physical network (12) comprises information regarding nodes and physical links connecting the nodes, and/or

wherein the information regarding the virtual networks (12) comprises

information regarding at least one virtual link connecting at least two virtual nodes, and/or wherein the embedding results (16) comprise a mapping of the physical links of the physical network (24, 71, 81) to the virtual links of the virtual networks (20, 21, 22, 70, 80) and/or a mapping of the physical nodes of the physical network (24, 71, 81) to the virtual nodes of the virtual networks (20, 21, 22, 70, 80).

4. Virtual network embedding orchestrator (10) according to any of the claims 1 to 3, wherein the at least one output parameter (16) of the problem solving unit (41) received and processed by the feedback processing unit (14) comprises at least one problem solving quality parameter and/or a problem solving time duration parameter.

5. Virtual network embedding orchestrator (10) according to claim 4, wherein the at least one problem solving quality parameter comprises an average number of links of the physical network (24, 71, 81) used for the links of the virtual networks (20, 21, 22, 70, 80), and/or a fraction of utilization of the links and nodes of the physical network (24, 71, 81), and/or an amount of virtual networks (20, 21, 22, 70, 80) embedded into the physical network (24, 71, 81).

6. Virtual network embedding orchestrator (10) according to claim 4 or 5,

wherein the feedback processing unit (14) is adapted to determine the at least one input parameter (17) of the problem formulation unit (11) by

- comparing the at least one problem solving quality parameter to at least one quality threshold, and

- amending the at least one input parameter to increase a problem solving quality, if the at least one problem solving quality parameter is below the at least one quality threshold, and/or by

- comparing the problem solving time duration parameter to a time duration threshold, and

- amending the at least one input parameter to decrease a problem solving time duration, if the problem solving time duration parameter is above the time duration threshold.

7. Virtual network embedding orchestrator (10) according to any of the claims 1 to 6, wherein the at least one input parameter (17) of the problem formulation unit (11) comprises an amount of virtual networks (20, 21, 22, 70, 80) for which the embedding into the physical network (24, 71, 81) is to be calculated simultaneously.

8. Virtual network embedding orchestrator (10) according to any of the claims 1 to 7, wherein the at least one input parameter (17) of the problem formulation unit (11) comprises a weighting parameter for weighting costs of a solution of the problem solving by the problem solving unit (41) against benefits of the solution of the problem solving by the problem solving unit (41),

wherein the costs of a solution of the problem solving by the problem solving unit (41) are based on infrastructure utilization of the physical network (24, 71, 81), and

wherein the benefits of the solution of the problem solving by the problem solving unit (41) are based on an amount of virtual networks (20, 21, 22, 70, 80) embedded into the physical network (24, 71, 81) by the solution.

9. Virtual network embedding orchestrator (10) according to any of the claims 1 to 8, further comprising an input buffer (50) adapted to successively receive and store the information regarding the virtual networks (12), and wherein the problem formulation unit (11) is adapted to formulate the embedding problem (13) based upon information regarding the virtual networks (12) stored in the input buffer (50).

10. Virtual network embedding system (40) comprising a virtual network embedding orchestrator (10) according to any of the claims 1 to 9 and a problem solving unit (41), wherein the problem formulation unit (11) is adapted to provide the embedding problem (13) to the problem solving unit (41).

11. Virtual network embedding system (40) according to claim 10,

wherein the problem solving unit (41) is adapted to determine the solution to the embedding problem (13) by integer programming or mixed integer programming.

12. Virtual network system (60) comprising a virtual network embedding system (40) according to claim 11, at least two virtual networks (20, 21, 22, 70, 80) embedded into a physical network (24, 71, 81).

13. Method for orchestrating embedding of virtual networks (20, 21, 22, 70, 80) into a physical network (24, 71, 81), comprising the following steps:

- formulating (101) an embedding problem (13) based upon information regarding the virtual networks, information regarding the physical network and input parameters

(17),

- providing the embedding problem (13) to a problem solving unit (41),

- receiving (102) embedding results (16) and at least one output parameter (16) from the problem solving unit (41), and

- determining (104) at least one of the input parameters for formulating the embedding problem (13) based upon the embedding results (16) and the at least one output parameter (16) of the problem solving unit (41).

14. Method according to claim 13,

wherein the following steps are performed:

- determining (103) ifthe received embedding results (16) are not final embedding results (15), based upon the received output parameters (16), and

- providing (104) the determined at least one input parameter for formulating the embedding problem (13), if the received embedding results (16) are not final embedding results (15).

15. A computer program with a program code for performing the method according to claim 13 or 14 when the computer program runs on a computer.