CN108737261B

CN108737261B - Two-stage virtual network function forwarding graph design method

Info

Publication number: CN108737261B
Application number: CN201810442401.0A
Authority: CN
Inventors: 虞红芳; 周廷枢; 孙罡; 许都
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2018-05-10
Filing date: 2018-05-10
Publication date: 2020-03-17
Anticipated expiration: 2038-05-10
Also published as: CN108737261A

Abstract

The invention discloses a two-stage virtual network function forwarding graph design method, which divides VNF-FG design into two stages by considering that a dependency relationship exists between partial VNFs in a network service request and the difference exists in flow change rates caused by the working characteristics of different VNFs. A single SFC structure is designed in the first stage, so that the total consumption of the bandwidth of a logic link is minimized; and the second stage integrates a plurality of SFCs, and reduces the number of VNF nodes. In the mapping work, each logical link needs to be mapped to one or more bottom-layer physical links, and bandwidth resources are allocated; VNF nodes need to be mapped onto the underlying physical servers and each VNF node needs to run in an instantiated virtual machine. Therefore, link bandwidth consumption and the number of VNF nodes are reduced as much as possible in the VNF-FG design, and link mapping cost and virtual machine instantiation cost can be reduced in the subsequent mapping process.

Description

Two-stage virtual network function forwarding graph design method

Technical Field

The invention belongs to the technical field of virtual network functions, and particularly relates to a design method of a two-stage virtual network function forwarding graph.

Background

With the rapid expansion of network scale and the iterative innovation of network technology, the old network architecture is increasingly fraught with complexity. In order to solve the problem, people try to reconstruct the existing Network architecture, and NFV (Network function virtualization) is a core technology in the Network architecture and is paid close attention by the communication industry. The NFV technique extracts network functions from proprietary hardware, decoupling hardware and software of legacy devices, such as multimedia caches, quality of service monitors, video transcoders, gateways, and various agents. The NFV deploys network element functions in a virtual machine through a virtualization technology, a virtual resource layer separates the network element functions from bottom layer resources, the network element functions on the upper layer cannot sense the bottom layer hardware physical resources, and an operator only needs to manage and maintain the bottom layer hardware physical resources.

In the existing operator network, a message transmitted by an end-to-end Service needs to go through a plurality of functional module nodes according to a certain specific sequence, and the concept of an SFC (Service Function Chain) is generated. The SFC is an ordered set including network functions, and performs a series of network function processing on a data stream according to a predetermined classification rule and policy. The SFC is widely applied in an NFV environment as a providing method of a Network service, and a Network Function that is conventionally implemented by hardware, that is, a VNF (virtual Network Function) is implemented by a software programming method, and has the characteristics of dynamic creation and deletion, high flexibility, and extensibility. The SFC logically represents the VNF types and order that the data flow needs to go through. A VNF-FG (VNF Forwarding Graph) is also a logical topology diagram for characterizing VNF connections and traffic flow direction, and may contain multiple SFC flows, providing multiple services.

The structure of the VNF-FG determines the specific way of providing the service (the order of VNF processing, the connection relationship of VNFs). The VNF-FG has source and destination end systems, which may be a functional entity or a device port, such as a tunnel endpoint, a virtual network card, etc., indicating the physical source and destination locations of network traffic in the underlying network.

The network service request of the user is deployed on the underlying network, and the resource allocation is carried out, which can be divided into the following four steps:

(1) a user requests a Service provider for a network function, where the request may include both a functional requirement and a non-functional requirement, the functional requirement refers to a specific network function, such as Deep Packet Inspection (DPI), Firewall (FW), and the like, and the non-functional requirement generally refers to Quality of Service (QoS).

(2) The service provider designs a VNF-FG topology based on the user's network service request, where it is important to compute the resources required for each functionality's needs to meet the non-functional requirements.

(3) A service provider formulates a deployment scheme, and maps VNFs and virtual links in VNF-FG to NFVI (network function virtualization infrastructure), thereby ensuring that functional requirements and non-functional requirements are simultaneously satisfied.

(4) According to the deployment scheme, the service provider allocates resources on the physical host, starts the virtual machines and deploys the VNF software, and finally connects the virtual machines through virtual links.

Of the four steps of service deployment described above, the result of the second step is the input to the third step, and these two steps greatly affect the QoS and overall cost of the entire service. Currently, the VNF-FG mapping problem of the third step is mainly studied in network service deployment, and the VNF-FG design work study of the second step is relatively lacked.

The VNF-FG contains the logical VNF nodes and their connections and specifies the order in which the flows traverse the VNF. The VNF-FG design refers to that a topological structure meeting VNF type requirements and VNF dependency relations is designed according to network service requests of users, and the VNF-FG is used as an input parameter of a mapping stage.

VNF dependency means that if VNF2 depends on VNF1, user traffic must first go through VNF1 before it can be processed by VNF2 (for example, a decoder must be located after an encoder).

The rate of change of flow of the VNF refers to the ratio of the flow rate exiting after processing by the VNF to the flow rate entering the VNF. The VNF may affect the outgoing rate of the traffic processed by the VNF (for example, the firewall may reduce the data volume, and the VPN agent may increase the data volume due to adding IPsec header overhead to the data packet), and the bandwidth requirement of the outgoing link may change, so that the bandwidth requirement of the logical link is small in the VNF-FG design, which may reduce the link mapping bottleneck and save the link mapping cost.

The VNF nodes in the VNF-FG are finally mapped to physical nodes, VNF software runs on a virtual machine to provide services, and certain cost exists in instantiation of the virtual machine and running of the virtual machine, so that the number of the instantiated virtual machines can be reduced by reducing the number of the nodes in the VNF-FG, and the purpose of reducing the operation cost is achieved.

Disclosure of Invention

The invention aims to provide a two-stage virtual network function forwarding graph design method, which designs a virtual network function forwarding graph for minimizing the total consumption of logical link bandwidth and the number of VNF nodes in an offline scene.

The technical scheme of the invention is as follows: a two-stage virtual network function forwarding graph design method comprises the following steps:

s1, single SFC design: and constructing a VNF dependency relationship diagram according to the VNF dependency relationship in the network service request, performing topology sequencing on the VNF dependency relationship diagram, converting the VNF dependency relationship diagram into a VNF topology sequence, and designing a single SFC by taking the VNF topology sequence as the order of VNFs in the SFC.

Step S1 includes the following substeps:

s11, constructing a directed acyclic VNF dependency relationship graph according to the VNF dependency relationship in the network service request.

And taking each required VNF in the network service request as a node in the VNF dependency graph, taking each dependency as an edge in the VNF dependency graph, and pointing the dependent item to the dependent item, thereby constructing the directed acyclic VNF dependency graph.

And S12, selecting the node with the minimum flow rate change rate from all the nodes with the degree of 0 in the VNF dependency graph, and putting the node into the VNF topology sequence.

And S13, subtracting 1 from the degree of entrance of the node succeeding the node selected in the step S12.

And S14, judging whether the VNF dependency graph has residual nodes, if so, returning to the step S12, and otherwise, entering the step S15.

S15, taking the VNF topological sequence as the order of VNFs in the SFC, and designing a single SFC;

s16, judging whether there is any network service request without SFC design, if yes, returning to step S11, otherwise, entering step S2.

S2, integrating multiple SFCs: and grouping all the SFCs, dividing the SFCs with the same source and target end systems into one group, and integrating the SFCs in each group to obtain the VNF-FG.

Step S2 includes the following substeps:

and S21, grouping all the SFCs, and dividing the SFCs with the same source and destination system into a group.

And S22, judging whether a group which is not subjected to SFC integration exists, if so, entering a step S23, and if not, finishing the design of the virtual network function forwarding graph.

S23, selecting a group of SFCs, integrating the SFCs in the group to obtain VNF-FG of the group, and returning to the step S22.

Step S23 includes the following substeps:

s231, selecting a group of SFCs, and taking one SFC with the most VNF nodes as an initial VNF-FG.

And S232, judging whether the group of SFCs still have residual SFCs, if so, entering the step S233, otherwise, finishing the integration of the group of SFCs, and returning to the step S22.

And S233, taking out the SFC with the most VNF nodes in the group of SFCs as the SFC to be integrated.

And S234, traversing all paths of the current VNF-FG, and calculating the longest common subsequence Lcs of each path and the SFC to be integrated.

S235, selecting the longest lc as lc _ max, integrating the lc _ max with the SFC to be integrated to form a new VNF-FG, and returning to the step S232.

If the VNF in the SFC to be integrated is the VNF contained in the Lcs _ max, the VNF is shared by the SFC to be integrated and the current VNF-FG; and if the VNF in the SFC to be integrated is the VNF not included in the Lcs _ max, newly adding the VNF node in the current VNF-FG and adding a link.

The invention has the beneficial effects that:

(1) the invention makes up the deficiency of VNF-FG design link in the current service deployment and resource allocation work, and provides a complete flow for calculating a VNF-FG structure (representing a specific network service providing mode) according to a network service request of a user in an off-line scene.

(2) The consideration factors are comprehensive: the invention considers that the VNFs in the network service requests have dependency relationships and the flow change rates of the VNFs have differences, and the total consumption of the logical link bandwidth in the SFC is minimum in the single SFC design stage on the premise of ensuring the dependency relationship; considering the determination of the end systems of the SFCs at the bottom layer physical location, in the SFC integration stage, the SFCs having the same source and destination end systems are grouped into one group, and the integration is performed in the group, so that it is possible to prevent additional link consumption caused by the fact that multiple SFCs with end systems far apart need to share one VNF instance in the subsequent mapping work due to the SFC integration of different source and destination end systems. Meanwhile, no ring exists in the VNF-FG formed by SFC integration, and link consumption caused by a flow generating loop in the mapping process can be avoided.

(3) The mapping cost can be reduced: the invention ensures the minimum total consumption of the logical link bandwidth when a single SFC is designed, and the same VNF is aggregated as much as possible at the integration stage of a plurality of SFCs, thereby reducing the link mapping cost and the virtual machine instantiation cost.

Drawings

Fig. 1 is a flowchart of a two-stage virtual network function forwarding graph design method according to an embodiment of the present invention.

Fig. 2 is a schematic diagram illustrating a dependency graph constructed according to a network service request according to an embodiment of the present invention.

Fig. 3 is a flowchart illustrating a substep of step S23 according to an embodiment of the present invention.

Fig. 4 is a schematic diagram of a VNF-FG path according to an embodiment of the present invention.

Fig. 5 is a schematic diagram of an SFC to be integrated according to an embodiment of the present invention.

Fig. 6 is a schematic diagram of a new VNF-FG after SFC aggregation according to an embodiment of the present invention.

Detailed Description

Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It is to be understood that the embodiments shown and described in the drawings are merely exemplary and are intended to illustrate the principles and spirit of the invention, not to limit the scope of the invention.

The embodiment of the invention provides a two-stage virtual network function forwarding graph design method, as shown in fig. 1, comprising the following steps S1-S2:

s1, single SFC design: considering two factors, namely, the dependency relationship existing in part of VNFs and the VNF traffic change rate, in the embodiment of the invention, a VNF dependency relationship graph is constructed according to the VNF dependency relationship in the network service request, the VNF dependency relationship graph is subjected to topology sequencing, the VNF dependency relationship graph is converted into a VNF topology sequence, and the VNF topology sequence is used as the VNF sequence in the SFC to design a single SFC.

Step S1 includes the following substeps S11-S16:

The network service request may include the VNF types required by the request and the dependency relationships between VNFs, and as an example of a network service request shown in fig. 2, each required VNF in the network service request is taken as a node in a VNF dependency relationship graph, each dependency relationship is taken as an edge in the VNF dependency relationship graph, and the dependent item points to the dependent item (for example, in fig. 2, if VNF3 depends on VNF1, an edge pointing from VNF1 to VNF3 is added), so as to construct a VNF dependency relationship graph. Since there is no circular dependency for the network service request, i.e. there is no situation such as a dependency B, B dependency C, C dependency a, the VNF dependency graph constructed must be loop-free.

And S15, taking the VNF topological sequence as the order of VNFs in the SFC, and designing a single SFC.

Steps S12-S15 are topology sorting operations performed on the VNF dependency graph constructed in step S11. In order to ensure the dependency of the VNF and to make the dependent item in front of the dependent item, the present invention performs topology sorting on the VNF dependency graph constructed in step S11 in a topology sorting manner, each time a point with an entry of 0 in the VNF dependency graph is entered into the topology sequence, after an element at the head of the sequence is popped up, the entry of a subsequent node is decremented by 1, and if a new node with an entry of 0 is generated, the new node is placed into the topology sequence, thereby ensuring that the dependent item is always in front of the dependent item.

The topological ordering may have a plurality of sequence results satisfying the dependency relationship, in order to obtain the total consumption B of the logical link bandwidth_totalThe smallest SFC configuration requires that the VNF that minimizes the rate of change of flow be placed as far forward in the SFC as possible. Total logical link bandwidth consumption B in the embodiment of the invention_totalThe definition is as follows:

wherein E_sRepresenting the logical link set in the SFC, e is the logical link in the SFC, and B (e) represents the bandwidth requirement of the logical link e. Therefore, when the VNF dependency graph constructed in step S11 is topologically sorted, each time a VNF with the smallest traffic change rate is popped from the node queue with an ingress of 0 and placed into the topological sorting sequence. Thus, the VNF that minimizes the rate of change of traffic is placed at the front position in the SFC as much as possible, while ensuring that the dependency is satisfied.

S16, judging whether there is any network service request without SFC design, if yes, returning to step S11, performing SFC design on the next network service request, otherwise, indicating that the SFC design is completed on all network service requests, and entering step S2.

S2, integrating multiple SFCs: all SFCs are grouped, the SFCs with the same source and target end system are divided into one group, the SFCs in each group are integrated to obtain an acyclic VNF-FG, and the situation that a loop is not generated in a mapping stage to cause extra link consumption is guaranteed.

Step S2 includes the following substeps S21-S23:

and S21, grouping all the SFCs, and dividing the SFCs with the same source and destination system into a group. Grouping by the same source-destination system is to avoid additional link consumption caused by the fact that multiple SFCs with end systems far apart in subsequent mapping work need to share one VNF instance due to SFC integration of different source-destination systems.

Integrating SFC into VNF-FG, the problems to be solved are: which VNF nodes in the SFC to be integrated are aggregated with which VNF nodes in the VNF-FG, respectively. In order to aggregate a larger number of VNF nodes, it is necessary to traverse all paths from the source node to the destination node in the VNF-FG, where VNFs on each path sequentially form a sequence L1, VNFs in the SFC to be integrated sequentially form a sequence L2, and calculate the longest common subsequence Lcs of L1 and L2. Each path in the VNF-FG has one lc with the SFC to be integrated, and in the embodiment of the invention, the longest lc is selected as an integration scheme lc _ max.

As shown in fig. 3, step S23 includes the following substeps S231-S235:

In VNF-FG, there may be more than one path from the source node to the destination node. In the embodiment of the present invention, as shown in fig. 4, in the VNF-FG, there are 3 paths, which are L1: { VNF1, VNF2, VNF4, VNF7}, L2: { VNF1, VNF2, VNF5, VNF7}, L3: { VNF1, VNF3, VNF6, VNF7 }. In order to aggregate as many identical VNFs as possible to reduce the number of nodes, all paths in the VNF-FG need to be traversed, and the longest common subsequence Lcs of the VNF sequence of the SFC to be integrated and the VNF sequence of the path is calculated. SFC as shown in FIG. 5: { VNF1, VNF3, VNF6, VNF8}, which is the longest common subsequence of the three paths in VNF-FG in fig. 4 as Lcs 1: { VNF1}, Lcs 2: { VNF1}, Lcs 3: { VNF1, VNF3, VNF6 }.

The longest Lcs (i.e., the Lcs with the largest number of nodes) is selected from the Lcs in step S234 and recorded as Lcs _ max, and the SFC is integrated into the corresponding path according to the elements in the Lcs _ max. For example, Lcs _ max of SFC shown in fig. 5 and VNF-FG shown in fig. 4 is Lcs 3: { VNF1, VNF3, VNF6}, SFC is thus integrated on path L3 as nodes in Lcs 3. In the integration process, if the VNF appears in Lcs3, the SFC to be integrated shares the VNF with the original VNF-FG; if there is a VNF not included by Lcs3 in the SFC to be integrated, the VNF node is newly added in the VNF-FG, and a link is added. As shown in fig. 6, the SFCs to be integrated share VNF1, VNF3, and VNF6 on L3 in the original VNF-FG, and a VNF8 node is newly added, and two links between VNF6 and VNF8, and between VNF8 and the destination node are added.

And integrating each SFC in the group according to the steps of S233-S235, wherein each integration forms a new VNF-FG and is used as the VNF-FG of the next SFC integration. When all SFCs are integrated, the VNF-FG finally formed is the VNF-FG designed according to the group division.

In the NFV architecture of ETSI, MANOs (Management and organization) are composed of VIMs (Virtualized Infrastructure Management), VNFM (VNF Manager, virtual Network Function Manager), and NFVO (Network Function Virtualization organization), where VIM is used to manage virtual resources, VNFM is used to manage Network functions, and NFVO is used for lifecycle Management and global cross-domain resource scheduling of Network services. The embodiment of the invention can be deployed in NFVO in MANO architecture under NFV environment, and determines the specific providing mode of network service (determines the number of VNFs and the connection relation) through VNF-FG design.

Currently, many VIM products in the industry are based on OpenStack. Therefore, OpenStack can be used as a VIM module of the NFV cloud platform, a Tacker project is used as a VNFM, and the NFVO module is developed on the basis, a network service provider can deploy the two-stage virtual network function forwarding graph design method provided by the embodiment of the invention in the NFVO in the MANO architecture, and the NFVO stores the VNF-FG as a VNF-FG Descriptor (VNF-FG Descriptor, VNF-FGD) after the VNF-FG is customized for subsequent network service arrangement and mapping work.

Aiming at an offline scene, when a plurality of network service requests arrive, the NFVO calls a virtual network function forwarding graph mapping method deployed in the NFVO, and a corresponding SFC structure is designed for each request, which is the work of a first stage; and then grouping the SFCs by the same source and target end system and integrating, wherein each group constructs a VNF-FG, which is the work of the second stage. And finally, storing the designed VNF-FG as a VNF-FGD template for use in a mapping stage.

It will be appreciated by those of ordinary skill in the art that the embodiments described herein are intended to assist the reader in understanding the principles of the invention and are to be construed as being without limitation to such specifically recited embodiments and examples. Those skilled in the art can make various other specific changes and combinations based on the teachings of the present invention without departing from the spirit of the invention, and these changes and combinations are within the scope of the invention.

Claims

1. A two-stage virtual network function forwarding graph design method is characterized by comprising the following steps:

s1, single SFC design: constructing a VNF dependency relationship graph according to the VNF dependency relationship in the network service request, performing topology sequencing on the VNF dependency relationship graph, converting the VNF dependency relationship graph into a VNF topology sequence, and designing a single SFC by taking the VNF topology sequence as the order of VNFs in the SFC;

s2, integrating multiple SFCs: grouping all SFCs, dividing the SFCs with the same source and target end systems into one group, and integrating the SFCs in each group to obtain VNF-FG;

the step S1 includes the following sub-steps:

s11, constructing a directed acyclic VNF dependency relationship graph according to the VNF dependency relationship in the network service request;

s12, selecting the node with the minimum flow rate change rate from all nodes with the degree of income of 0 in the VNF dependency relationship graph, and putting the node into the VNF topology sequence;

s13, subtracting 1 from the degree of entrance of the node succeeding the node selected in the step S12;

s14, judging whether the VNF dependency graph has residual nodes, if yes, returning to the step S12, and if not, entering the step S15;

s16, judging whether there is any network service request without SFC design, if yes, returning to step S11, otherwise, entering step S2;

the step S11 specifically includes: taking each required VNF in the network service request as a node in the VNF dependency graph, taking each dependency as an edge in the VNF dependency graph, and pointing the dependent item to the dependent item, thereby constructing a directed acyclic VNF dependency graph;

the step S2 includes the following sub-steps:

s21, grouping all SFCs, and dividing the SFCs with the same source and target end system into a group;

s22, judging whether there is a group without SFC integration, if yes, entering step S23, otherwise ending the design of virtual network function forwarding graph;

s23, selecting a group of SFCs, integrating the SFCs in the group to obtain VNF-FG of the group, and returning to the step S22;

the step S23 includes the following sub-steps:

s231, selecting a group of SFCs, and taking one SFC with the most VNF nodes as an initial VNF-FG;

s232, judging whether the group of SFCs still have residual SFCs, if so, entering the step S233, otherwise, completing the integration of the group of SFCs, and returning to the step S22;

s233, taking out the SFC with the most VNF nodes in the group of SFCs as the SFC to be integrated;

s234, traversing all paths of the current VNF-FG, and calculating the longest common subsequence Lcs of each path and the SFC to be integrated;

s235, selecting the longest Lcs as the Lcs _ max, integrating the Lcs _ max and the SFC to be integrated to form a new VNF-FG, and returning to the step S232;

the step S235 specifically includes: if the VNF in the SFC to be integrated is the VNF contained in the Lcs _ max, the VNF is shared by the SFC to be integrated and the current VNF-FG; and if the VNF in the SFC to be integrated is the VNF not included in the Lcs _ max, newly adding the VNF node in the current VNF-FG and adding a link.