Background
At present, the mobile network industry is rapidly evolving to 5g, and three new application fields of mobile broadband enhancement, large-scale internet of things and low-delay high-reliability communication play an important role. The 5g network has high flexibility to cope with the service change of mobile operators, and particularly, the proposal of the network function virtualization concept enables the infrastructure to flexibly meet the diversification of the vertical application requirements. The network slice is a technology for flexibly configuring resources in a wireless virtual network, and can be quickly deployed and centrally managed. Limited physical resources are divided and recombined by means of Software Defined Networking (SDN) and Network Function Virtualization (NFV) technologies to form logically mutually independent virtual network resources for each slice network, so that repeated and efficient utilization of network resources is realized, cost input and operation expenditure of operators are reduced, better quality is provided for tenants, and the utilization rate of the network resources is improved. In a slice network, each service request is composed of several different Virtual Network Functions (VNFs), which are interconnected to be called Service Function Chains (SFCs). The different requirements of each service request can cause different sets of VNFs on each SFC, and how to effectively deploy SFCs so as to maximize the revenue of virtual operators while satisfying the slice quality of service (QOS) is a hot issue of research in 5G networks.
The low-delay scene of the 5G mobile communication system has higher requirement on the delay of the communication service, and the user millisecond-level end-to-end network service needs to be supported. Reducing the end-to-end latency of network services is one of the key issues that SFC deployments need to address. The existing invention takes the processing delay and the transmission delay as the end-to-end delay for analysis. The processing time delay is set as a fixed value, the influence of the node load on the time delay is ignored, the method is not practical, the scale of the applicable network service is limited, and the time delay problem caused by the node position and the reliability is not considered.
During SFC deployment, either a hardware failure (e.g., failure of a physical node executing a VNF) or a software failure (e.g., misconfiguration of the VNF itself) destroys the entire link, resulting in service suspension. To date, most existing inventions employ redundancy-based VNF deployment policies to achieve reliability of SFC deployment. It should be noted that the presence of these redundant VNFs will increase the length of the service function chain, thereby increasing the end-to-end latency of the service, which is very disadvantageous for latency-limited SFCs and reduces the utilization of resources. Therefore, it is necessary to develop a new SFC deployment scheme to improve the service quality of SFC deployment.
Disclosure of Invention
In view of this, an object of the present invention is to provide a QoS guarantee-based dynamic service function chain deployment method, which can dynamically allocate resources to an SFC according to underlying resources, improve QoS for SFC deployment, that is, ensure reliability requirements of the SFC while minimizing end-to-end delay. Furthermore, it is another object of the present invention to be able to accept more SFC requests to maximize the utilization of the underlying resources. In order to achieve the purpose, the invention provides the following technical scheme:
a QoS guarantee-based service function chain dynamic deployment method comprises the following steps:
s1: aiming at the QoS problem of communication service in a slice network, establishing a Service Function Chain (SFC) deployment model facing the reliability requirement, wherein the model aims at minimizing end-to-end time delay;
s2: the method comprises the steps of comprehensively considering node positions and reliability, and utilizing a novel node sorting method to sort Virtual Network Functions (VNFs) in physical nodes and service function chains respectively;
s3: determining a node mapping priority according to a node sequencing result, and selecting nodes meeting resource constraints to deploy virtual network functions to realize load balancing;
s4: in the link mapping process, a time delay shortest path which meets the link resource constraint and ensures the reliability requirement is searched for mapping.
Further, in step S1, the service function chain deployment network model is:
the underlying network is formalized as an undirected graph G
S=(N
S,L
S) In which N is
SRepresenting a set of underlying nodes, each of which may deploy one or more VNFs, L
SRepresents the set of all underlying links; each bottom node m is belonged to N
SHas a CPU capacity of
The node position is loc (m), and a link l connecting the nodes m and n
mnHas a bandwidth of
Reliability is R (l)
mn);
Is a loop-free set of paths between nodes m and n;
chains of SFCs are formalized into a directed graph, denoted G
V=(N
V,L
V),N
VRepresenting all VNF sets, L
VA set representing all virtual links connecting the VNFs; each SFC consists of ordered VNF functions, Q denotes the strength of the SFC request arriving per unit time, and the set of SFCs is denoted S ═ S
q1,2,. cndot.Q }, each SFC ∈ Q consisting of
And a virtual link connecting two adjacent VNFu and v
Composition is carried out; the CPU resource requirement of each VNFu in the SFC is
Virtual link l
uvThe bandwidth requirement is
Defining a binary variable
Representing a virtual link l
uvWhether or not to map to physical link/
mn∈L
SThe above step (1); the position of the virtual machine represented by each VNF is loc '(u), the maximum distance offset cannot exceed lp' (u) when the virtual machine is mapped to the bottom layer node, and the time delay of the virtual link after the virtual link is mapped to the bottom layer link is d
uvReliability after mapping to underlying link cannot be lower than R (l)
uv)。
Further, in step S1, the reliability-oriented requirement is to select a node and a link with high reliability for SFC deployment and the selected link satisfies the minimum requirement of the link for reliability.
Further, in step S1, the dynamic deployment model with the objective of minimizing the end-to-end latency is:
the first part is to maximize the number of requests to access the SFC, in order to maximize the utilization of resources; the second part is to minimize the end-to-end delay; calculating end-to-end delay by using the delay superposition of each hop of link in the SFC; the delay per hop link is expressed as the processing delay
And transmission delay
The sum of (1); in which the processing delay and the load that the node needs to handle
In relation, the processing load is defined as the ratio of the total VNF resource demand to be processed to the processing capacity of the underlying node, when the CPU load of a node increases, the processing delay of the node increases rapidly, assuming that the processing delay is a convex function of the processing load, a convex delay curve is approximated using piecewise linearization; the transmission delay is related to the number of hops that the SFC link that needs to be processed maps to the underlying link.
Further, in step S2, the node position and reliability are:
the node position is represented by a node connectivity G (m), the validity E (m) of the node and an adaptability T (m); wherein the connectivity of the node is determined by the total number of adjacent links; the effectiveness of the node is expressed by the node efficiency, the node efficiency is defined as the reciprocal of the distance between the node and other nodes, the distance represents the hop count of a link, and the shorter the transmission distance is, the higher the node efficiency is; the node adaptability means that after the node m fails, all other nodes connected with the shortest path through the node m are used as the minimum value of the communication distance increased for recovering the link connected with the node m, the smaller the increased communication distance is, the shorter the recovery time is, the larger the node adaptability value is, otherwise, the smaller the node adaptability value is;
the node reliability can be represented by the normal working probability of the node; and the normal working probability and the failure rate of the node are lambdam(ii) related; node failure rates are expressed in terms of Mean Time Between Failure (MTBF) and Mean Time To Repair (MTTR); the normal operation probability of the node can be expressed as 1-lambdam;
Assuming that the link failure satisfies the poisson distribution, the probability of no failure occurring in the time interval t is
So that the link has a reliability of
Wherein λ
mnIs a link l
mnThe failure rate of (t) is the link l
mnTime delay of (3); due to link factors taken into account when VNF deployment is performed, under the delay constraint, the following transformations are made:
maximizing the above equation amounts to minimizing the following equation:
due to d
u'
vThe value mapped to each physical link is the same and can be further translated into the following equation: min lambda
mn(ii) a The link influencing factor is expressed in node ordering as:
further, in step S2, the novel node ranking method is: the significance of the node is redefined by adopting the idea in the PageRank algorithm of Google, and is expressed as the following formula:
where r (m) represents the score of node m, which characterizes the importance of the node, γ is a damping coefficient between 0 and 1, J (m) represents the set of nodes adjacent to node m,
represents a normalized node resource state, represented as
According to the idea of PageRank, obtaining a final node iteration score expression which is expressed as the following vector expression, and obtaining the importance of all nodes through continuous iteration:
r=(1-γ)C+γHr。
further, in step S3, the deployment of the virtual network function specifically includes: carrying out VNF deployment according to the node score result; firstly, nodes and VNF scores are arranged in a descending order, and then the VNF with a large score value is deployed to a physical node with a large score value according to the idea of merging and sorting, wherein the physical node must meet the resource requirement constraint of the VNF; if all nodes in the physical network can not meet the VNF resource constraint after circulation, the SFC is rejected; carrying out the deployment of VNFs in the SFC in sequence according to the steps until the deployment is completed or the VNFs do not meet the resource constraint and are rejected; certain changes are required when VNF sorting is performed; since the VNF in the SFC does not need to consider the normal operation probability of the node and the reliability property of the link, the resource state in the node normalization is represented as
Due to the link reliability constraint per hop link in the SFC, the impact condition of the link in the SFC changes
Further, in step S4, the link mapping process specifically includes: firstly sorting the sizes of links according to the reliability requirement of each hop of the links in the SFC, firstly selecting the links with higher reliability requirement for mapping, deleting all bottom links which do not meet the requirement of the corresponding links of the SFC, then executing a K-shortest path algorithm to select K paths with shortest time delay, sorting the paths in an increasing way according to the time delay size, and finally selecting the path with the shortest time delay which meets the link reliability constraint of the SFC.
The invention has the beneficial effects that: the invention establishes a service function chain deployment model facing the reliability requirement, and designs a service function chain dynamic deployment scheme based on QoS guarantee by taking the minimized end-to-end time delay as a target. The method comprehensively considers the position and the reliability of the node, the VNF is deployed by using a novel node sequencing method, the node with high reliability and high reliability of the link connected with the node is selected as far as possible to deploy the VNF, and preparation is made for link mapping. In addition, the VNF selects nodes with relatively more resources to deploy, so that load balance is realized, and processing time delay is reduced. And in the link mapping, selecting a time delay shortest path meeting the reliability requirement for deployment. The invention reduces the end-to-end time delay of the service function chain, ensures the reliability of deployment, and improves the request acceptance rate and the resource utilization rate by the method of load balancing and no resource reservation in the process.
Detailed Description
Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
FIG. 1 is a schematic diagram of an example of a scenario in which embodiments of the present invention may be applied. Consider a network function virtualization architecture composed of an NFV orchestration and control framework. To the left of fig. 1, the orchestration and management functions of the system are represented, consisting of management and orchestrator (MANO) and Software Defined Network (SDN) controllers. The MANO comprises three parts of service arrangement, virtual network function management and virtualization infrastructure management, and is respectively responsible for SFC service requests, VNF connection and infrastructure global resource management on the right side of the figure. The infrastructure of a Physical Network (PN) is composed of an access network and a core network, and the access network and the core network are connected through an SDN network. By isolation, multiple VNFs may run on the same underlying node without affecting each other. The MANO may perform resource allocation based on the resource status and the service request. Furthermore, the introduction of access networks requires SDN controllers to enhance MANO functionality, enabling MANOs with a wider range of connection configuration management capabilities. The SDN controller may perform traffic management according to bandwidth consumption, delay, and the like, and may also report network state information to the MANO for resource allocation. The end-to-end SFC service request is orderly composed of different VNFs and is mapped to an underlying network for service according to the resource requirements of the VNFs. In the mapping process, in consideration of the low delay and reliability requirements of the SFC, a scheme is required to be found for minimizing the end-to-end delay of the SFC deployed on the underlying network and simultaneously ensuring the reliability of the SFC service request.
FIG. 2 is a model diagram of a node sorting process in the present invention. The flow is used to order VNFs in the underlying physical node and SFC requests. The method comprises the following steps:
step 201: initializing a network topology G ═ N, L, and presetting a small positive value sigma;
step 202: calculating a matrix H and an initial vector C;
step 203: defining iteration times k, and initializing k to be 0; defining a variable w, and initializing w ═ infinity;
step 204: judging whether the w & gt sigma condition is met, if not, executing the step 208; otherwise, go on to step 205;
step 205: calculating r ═ (1-gamma) (I-gamma H)-1C;
Step 206: let k be k + 1;
step 207: calculating w ═ abs (r)k+1-rk) Returning to step 204 to continue execution;
fig. 3 is a schematic diagram of SFC request deployment process in the present invention, which includes the following steps:
step 301: collecting SFCs arriving in a unit time;
step 302: is there SFC left unprocessed? If yes, go to step 303; otherwise, the deployment is finished;
step 303: based on the node position and reliability, calculating the priorities of the nodes and VNFs in the bottom nodes and the VNFs in the SFCs according to a node sorting method, and sorting the nodes and the VNFs in a descending order;
step 304: and executing a VNF deployment algorithm, and deploying the VNF with a large score value to the physical node with the large score value meeting the resource constraint of the VNF to realize load balancing. If successful, go to step 305; otherwise, rejecting the SFC and continuing to process the next SFC;
step 305: and executing a link mapping algorithm, and searching a time delay shortest path which meets the virtual link resource constraint in the SFC and can ensure the reliability requirement for mapping. If successful, go on to step 306; otherwise, rejecting the SFC, returning to the step 302 to continue processing the next SFC;
step 306: if step 305 is successful, the SFC is accepted, the underlying network resources are updated, and the process returns to step 302 to continue execution until all SFCs are processed.
Fig. 4 is a schematic VNF deployment flow chart in the present invention, including the following steps:
step 401: storing the sorted physical nodes and the VNF priority order in the SFC;
step 402: is it determined whether all VNFs in the SFC have been completely deployed? If yes, go to step 406 to perform link mapping; otherwise, go on to step 403;
step 403: deploying the VNFs ranked in the front first according to the VNF priorities;
step 404: searching a physical node meeting the VNF resource constraint according to the priority of the physical node, namely deploying the VNF with high priority to the physical node with high priority;
step 405: if a physical node satisfying the condition is found, the VNF is deployed to the physical node, and the physical node is not considered in subsequent VNF deployment, and then the procedure returns to step 402 to continue processing the next VNF;
fig. 5 is a schematic diagram of a virtual link mapping process in the present invention, which includes the following steps:
step 501: sequencing links of each hop in the SFC according to the reliability requirement, and firstly deploying links with high reliability requirements;
step 502: is it determined whether each hop link in the SFC has been deployed completed? If yes, executing step 507 to deploy the next SFC, otherwise, continuing to execute step 503;
step 503: deleting physical links which do not meet the resource requirement of the hop link;
step 504: obtaining K time delay shortest paths by using a K shortest path method, and sequencing the K time delay shortest paths from small to large;
step 505: selecting a time delay shortest path meeting the reliability requirement of the hop link from the K paths;
step 506: the jump link is mapped to the shortest delay path, and the process returns to step 502 to continue processing until each jump link is processed.
Finally, it is noted that the above-mentioned preferred embodiments illustrate rather than limit the invention, and that, although the invention has been described in detail with reference to the above-mentioned preferred embodiments, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the scope of the invention as defined by the appended claims.