CN109831319B - Network function deployment method considering multidimensional resource constraints - Google Patents
Network function deployment method considering multidimensional resource constraints Download PDFInfo
- Publication number
- CN109831319B CN109831319B CN201811620365.9A CN201811620365A CN109831319B CN 109831319 B CN109831319 B CN 109831319B CN 201811620365 A CN201811620365 A CN 201811620365A CN 109831319 B CN109831319 B CN 109831319B
- Authority
- CN
- China
- Prior art keywords
- network function
- network
- deployment
- platform
- flow
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The invention discloses a network function deployment method considering multidimensional resource constraints, which comprises the following steps: abstracting a network function deployment problem into an MVDP mathematical model; the problem in the demonstration model is the NP-hard problem; and giving off-line and on-line approximate scheduling algorithm solution. The method can obtain an optimal service chain deployment scheme based on the service chain of the network function required to pass by each flow and the resources required by each flow, so that the total service chain deployment cost is minimum, and the resources consumed by the network function do not exceed the maximum resource capacity of the server; the deployment decision of the invention can ensure that the network function required to be operated by each flow is deployed on the server node through which the flow passes.
Description
Technical Field
The invention belongs to the technical field of 5G network slicing, network function virtualization and software defined networking, and particularly relates to a network function deployment method considering multidimensional resource constraints.
Background
While 5G, which is about to be commercialized in 2020, is widely mentioned, network slicing (NetworkSlicing) has to be mentioned. As the most discussed technique in 5G, the significance of network slices to 5G can be enormous. The network slicing is to divide a physical network of an operator into a plurality of virtual logical networks, and each virtual network is divided according to different service requirements, such as time delay, bandwidth, security, reliability and the like, so as to flexibly cope with different network application scenarios.
In the 5G era, the object of the mobile network service is no longer a pure mobile phone, but various types of devices including mobile phones, tablets, fixed sensors, vehicles, and the like. Application scenarios are also diversified, such as mobile broadband, large-scale internet, mission-critical internet, and the like. The requirements to be met are also diverse, such as mobility, security, low latency, reliability, etc. All this provides a place for the network slice to use.
The rise of Network Function Virtualization (NFV) separates software from hardware, by which network managers do not need to invest in installing and maintaining expensive proprietary hardware to establish service chains for network connected devices. Instead, they may invest in a server to run software, the server, that performs various network functions. Network function virtualization provides flexibility and agility in deploying network functions, while Software Defined Networking (SDN) consists of a series of automated network objects and components such as firewalls and routers. It provides an easy and centralized way to manage networks, including transport networks, data center networks, etc. Both SDN and NFV can provide the functionality required for network slicing, and the main idea of 5G network slicing is to create and partition different services on the network, enabling operators to provide optimal support for these services. As a basis for network slicing, in a 5G network slicing scenario, a software defined network will be associated with server virtualization to connect virtual network functions.
Existing online and offline scheduling methods for networks consider optimization of single-dimensional resource deployment, and in fact, each network function needs to consume multiple resources on a server during operation, such as consuming CPU clock cycles, server memory resources, and so on. Other research works prove that under the condition that available resources of a server are limited, the operation of network functions can become network bottlenecks, so that packet loss and performance reduction are caused.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide a network function deployment method considering multidimensional resource constraints, aiming at the deficiencies of the prior art, and minimizing the deployment overhead of network functions in the whole network on the premise of multidimensional resource constraints. The invention is realized on the premise that: the server nodes through which each flow needs to pass are known, the set of network functions required by each flow is also predetermined, the overhead and resource requirements for each network function to deploy on each server are known, and the resource capacity on each server is known. The above are all reasonable assumptions. Based on the method, the optimal deployment decision of the network function can be obtained, the deployment cost is reduced to the maximum degree, and meanwhile, resources are utilized reasonably.
In order to achieve the technical purpose, the technical scheme adopted by the invention is as follows:
a network function deployment method considering multidimensional resource constraints comprises the following steps:
1) abstracting a network function Deployment Problem into a Multi-resource network function Deployment Problem (MVDP) mathematical model and giving a linear programming expression;
2) the problem in the demonstration model is the NP-hard problem;
3) and giving off-line and on-line approximate scheduling algorithm solution.
In order to optimize the technical scheme, the specific measures adopted further comprise:
in the step 1), the linear programming expression is as follows:
wherein M is the set of running network function platform, F is the set of flow in network, Ri＝(r1，r2，…，rk，…，rd)，RiFor resource vectors on network function platform i, VNFfNeed to operate for flow fThe set of VNFs of (a) is,
in the t-th time slot, if the network function j is deployed on the platform i and needs to consume the kth resource, k satisfies k is less than or equal to d, d is the resource dimension,
lease costs, x, deployed on platform i for network functions j that need to be traversed during the t-th time slot stream ff，i，j(t), at the t-th time slot, if the network function j that the flow f needs to pass through finally needs to be installed on the platform i, the value is 1, otherwise, it is 0.
In the step 3), the offline approximate scheduling algorithm includes the following steps:
a) listing targets and constraints of MVDP, and solving linear programming of the MVDP of the relaxation version to obtain a fractional solution;
b) constructing a bipartite graph;
c) and performing minimum weight matching based on the bipartite graph.
In the step b), the construction of the bipartite graph specifically includes: the vertices of a graph are partitioned into two disjoint sets U and V such that each edge connects U, V the vertices, respectively.
In the step c), the minimum weight complete matching of the bipartite graph specifically includes: and taking the inverse numbers of all the edge weights, solving the maximum weight matching, and then taking the inverse numbers of the matched values.
The maximum weight matching is solved by adopting a KM algorithm.
The above online approximate scheduling algorithm specifically comprises: the server deployment requiring the least overhead is selected in turn for each network function required for each arriving stream.
The invention has the following beneficial effects:
1) based on the service chain of the network function which each flow needs to pass through and the resources required by each flow, an optimal service chain deployment scheme can be obtained, so that the total service chain deployment cost is minimum, and the resources consumed by the network function do not exceed the maximum resource capacity of the server;
2) the deployment decision of the invention can ensure that the network function required to be operated by each flow is deployed on the server node through which the flow passes.
3) The method is suitable for being realized by combining network function virtualization and software defined network technology under a 5G network slicing scene; not only an offline solution is provided, but also an online solution is provided, so that different requirements are met; the method can balance server resources while remarkably reducing the network function deployment overhead.
Drawings
FIG. 1 is an embodiment of the present invention;
FIG. 2 is a process of least weight matching for bipartite graphs for an off-line algorithm of the invention.
Detailed Description
Embodiments of the present invention are described in further detail below with reference to the accompanying drawings.
The invention relates to a network function deployment method considering multidimensional resource constraint, which comprises the following steps:
1) abstracting a network function deployment problem into an MVDP mathematical model and providing a linear programming expression;
in an embodiment, the abstraction process is as follows:
the MVDP mathematical model is represented by a binary set (M, F), where M represents the set of running network function platforms and F represents the set of traffic in the network.
Each flow should go through a set of network functions (VNFs), and given that the set of network functions that each flow needs to go through to run is known, it needs to be determined on which server platform each network function needs to be installed to run. Each network function platform is assumed to have some limited resources, with dimension d, which is expressed as resources such as CPU, memory, etc. Expressed by the following expression: ri＝(r1，r2，…，rk，…，rd) Wherein R isiRepresenting a resource vector on the network function platform i.
The aim of the invention is to find a suitable network function for each network function without exceeding these resourcesThe deployment platform of (1). The present invention should have the following inputs: f, representing a set of flow; m, representing a set of network function platforms; VNFfRepresents the set of VNFs that flow f needs to run;
it means that at the time of the t-th time slot, if the network function j is deployed on the platform i, the k should satisfy k ≦ d if the size of the k-th resource that needs to be consumed.
Representing the lease cost of the network function j which needs to pass through the t time slot flow f and is deployed on the platform i; x is the number off，i，j(t), which is a variable of 0 and 1, and at the t-th time slot, if the network function j that the flow f needs to pass through finally needs to be installed on the platform i, the value is 1, otherwise, the value is 0.
The linear programming expression from the abstract results is as follows:
this problem is the MVDP problem.
2) The problem in the demonstration model is the NP-hard problem;
3) and giving off-line and on-line approximate scheduling algorithm solution.
The invention can obtain a deployment scheme by running the program under any condition, and then deploy the actual network function according to the program result. The deployment scheme obtained according to the algorithm in the invention can ensure that the deployment cost of the network function is minimized on the premise of meeting the requirement of multidimensional resources without exceeding the maximum load of the server.
On one hand, the invention aims at the scene problem that the network function deployment difficulty and the cost are increased due to the limitation of various network resources (bandwidth and the like) and computing resources (CPU, memory and the like) along with the blowout type increasing of the core network access equipment in the 5G network slice scene; on the other hand, when the network function runs on a virtual machine or a container, multiple types of resources, such as CPU clock cycles and memory resources, are often consumed.
Fig. 1 is an application example of the present invention, which shows the network architecture of the cellular core network of the present day. The dotted line and the solid line in the figure represent signal traffic (Signaling traffic) and Data traffic (Data traffic), respectively, and the cellular Network architecture mainly consists of a Radio Access Network (RAN) and a Packet Core Network (EPC), wherein the RAN is used to connect a terminal equipment (UE) and a base station (eNodeB) of a user, and once the traffic from the terminal equipment reaches the base station, the RAN is forwarded to the Packet Core Network (EPC), and the EPC is composed of a set of hardware including a Mobility Management Entity (MME), a Serving Gateway (SGW), a Packet Data Network Gateway (Packet Gateway, PGW), a Home Subscriber Server (Home Subscriber Server, HSS), and a Policy and Charging Rule function Entity (Policy and Charging Rule function, PCRF). The MME is responsible for handling all signal traffic from the user terminals and base stations, and for user identity authentication, mobility management and session management. The SGW and PGW handle all data traffic. The SGW forwards the data traffic from the base station to the PGW, which then queries the PCRF for confirming the charging rules. The PGW will then forward the message to the particular switch connected to the internet.
The problem is firstly abstracted into MVDP, and the problem requires that the optimal deployment decision is sought on the premise that various types of resources meet the requirement that the load of a server is not exceeded when the network function is deployed, so that the deployment cost of the network function is minimized. It is also necessary to ensure that the network functions that a flow needs to run must be deployed on the server platform that the flow's path passes through. The MVDP problem is proved to be an NP-hard problem, and a series of approximate algorithm solutions are given. The approximate algorithm for solving MVDP proposed by the invention mainly has two, one is an off-line approximate algorithm, and the other is an on-line approximate algorithm.
The off-line approximate scheduling algorithm comprises the following steps:
a) listing targets and constraints of MVDP, and solving linear programming of the MVDP of the relaxation version to obtain a fractional solution;
b) constructing a bipartite graph;
c) and performing minimum weight matching based on the bipartite graph. The method comprises the following specific steps:
in the offline approximation algorithm, the targets and constraints of the MVDP problem are listed first, and by solving the linear programming of the relaxed version of the problem, a fractional solution for the network function deployment scenario can be obtained. Of course, this solution cannot be used as the final solution of the algorithm, because for each service chain, either deployment on a certain server or non-deployment is chosen. The result can only be an integer solution of 0 or 1. In the following, only how to know on which server the deployment is for each network function needs to be considered. Therefore, a mode of constructing a bipartite graph is adopted, and a minimum weight matching algorithm of the bipartite graph is carried out according to the constructed bipartite graph, so that a final output solution is obtained. And actually deploying each service chain to a deployment platform of the corresponding network function according to the result. The specific method for constructing the bipartite graph is as follows:
FIG. 2 is a least-squares complete matching process for bipartite graphs for a first level deployment of the invention.
Bipartite graph means, in brief, that if there is a graph where points can be divided into two groups and all edges cross the group boundaries, the graph is a bipartite graph. To be precise: the vertices of a graph are partitioned into two disjoint sets U and V such that each edge connects U, V the vertices, respectively, and if such a partition exists, the graph is a bipartite graph.
Complete matching means that all points in the X point set have corresponding matching in the bipartite graph, or all points in the Y point set have corresponding matching, and the matching is called complete matching. The specific minimum weight complete matching method for the bipartite graph is as follows: and taking the opposite numbers of all the edge weights, solving the maximum weight matching, and then taking the opposite numbers of the matched values.
The most weight matching algorithm for solving bipartite graphs is commonly used by the Kuhn Munkras (KM) algorithm. The KM algorithm transforms the problem of maximum weight matching into the problem of perfect matching by giving each vertex a label (called a superscript).
Suppose vertex XiIs denoted by A [ i ]]Vertex YiIs denoted by the top B [ i ]]Vertex XiAnd YiThe side weight between W [ i, j]At any time during the algorithm execution, for any edge (i, j), A [ i]+B[j]>＝W[i，j]This is always true. The correctness of the KM algorithm is based on the following theorem: if all of the bipartite graphs satisfy A [ i ]]+B[j]＝W[i，j]The sub-graph formed by the edges of (1) has a perfect match, and then the perfect match is the maximum weight match of the bipartite graph.
In online MVDP problem solving, the time for any traffic to reach the network is unpredictable, and therefore the set of network functions traversed by each flow cannot be known in advance. In this case, deployment decisions for network functions will be made online as traffic arrives. In the algorithm of the invention, when traffic arrives, a server node with low deployment cost and resource requirement is selected for the network function required by the traffic, and the path of the traffic passes through the server node.
By operating an offline or online approximation algorithm, an optimal deployment strategy of the network function on the premise of meeting the requirements of the multidimensional resources can be obtained, and the strategy can ensure that the deployment cost is the lowest compared with other deployment strategies.
The novelty of the invention lies in the comprehensive exploration and design of the deployment problem of network functions based on multi-dimensional resource constraints. Cellular traffic within the framework of the invention may be uniformly managed by a logically centralized central controller. Through the virtualization mode, an operator can accelerate innovation by shortening the updating period of service chain deployment, reduce capital cost and improve expansibility. The implementation of the invention is based on a deployment scenario derived after a series of feasible programs are run.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may be made by those skilled in the art without departing from the principle of the invention.
Claims (6)
1. A network function deployment method considering multidimensional resource constraints is characterized in that: the method comprises the following steps:
1) abstracting a network function deployment problem into an MVDP mathematical model and providing a linear programming expression;
2) the problem in the demonstration model is the NP-hard problem;
3) giving off-line and on-line approximate scheduling algorithm solution;
step 3) the off-line approximate scheduling algorithm comprises the following steps:
a) listing targets and constraints of MVDP, and solving linear programming of the MVDP of the relaxation version to obtain a fractional solution;
b) constructing a bipartite graph;
c) and performing minimum weight matching based on the bipartite graph.
2. The method according to claim 1, wherein the network function deployment method takes multidimensional resource constraints into consideration is characterized in that: the linear programming expression in step 1) is as follows:
wherein M is a set for operating a network function platform, and F is a set R of traffic in the networki＝(r1，r2，…,rk,…,rd)，RiFor resource vectors on network function platform i, VNFfThe set of VNFs that need to be run for flow f,
(t) when the network function j is deployed on the platform i in the tth time slot, if the kth type of resources required to be consumed are the size of the network function j, k is less than or equal to d, d is the resource dimension,
lease costs, x, deployed on platform i for network functions j that need to be traversed during the t-th time slot stream ff,i,j(t), in the t-th time slot, if the network function j that the flow f needs to pass through needs to be installed on the platform i finally, the value is 1, otherwise, the value is 0.
3. The method according to claim 1, wherein the network function deployment method takes multidimensional resource constraints into consideration is characterized in that: the step b) of constructing the bipartite graph specifically comprises the following steps: the vertices of a graph are partitioned into two disjoint sets U and V such that each edge connects U, V the vertices, respectively.
4. The method according to claim 1, wherein the network function deployment method takes multidimensional resource constraints into consideration is characterized in that: the minimum weight complete matching of the bipartite graph in the step c) is specifically as follows: and taking the inverse numbers of all the edge weights, solving the maximum weight matching, and then taking the inverse numbers of the matched values.
5. The method of claim 4, wherein the method comprises: and solving the maximum weight matching by adopting a KM algorithm.
6. The method according to claim 1, wherein the network function deployment method takes multidimensional resource constraints into consideration is characterized in that: the online approximate scheduling algorithm specifically comprises the following steps: the server deployment requiring the least overhead is selected in turn for each network function required for each arriving stream.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811620365.9A CN109831319B (en) | 2018-12-27 | 2018-12-27 | Network function deployment method considering multidimensional resource constraints |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811620365.9A CN109831319B (en) | 2018-12-27 | 2018-12-27 | Network function deployment method considering multidimensional resource constraints |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109831319A CN109831319A (en) | 2019-05-31 |
| CN109831319B true CN109831319B (en) | 2021-10-26 |
Family
ID=66861402
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811620365.9A Active CN109831319B (en) | 2018-12-27 | 2018-12-27 | Network function deployment method considering multidimensional resource constraints |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109831319B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113242560B (en) * | 2021-04-21 | 2022-11-01 | 南通大学 | Multi-stage planning method, device and equipment for UPF deployment and UE scheduling |
| CN115622889B (en) * | 2022-12-19 | 2023-05-09 | 湖北省楚天云有限公司 | Containerized network architecture and network function deployment method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102932279A (en) * | 2012-10-30 | 2013-02-13 | 北京邮电大学 | Multidimensional resource scheduling system and method for cloud environment data center |
| CN105656999A (en) * | 2015-12-25 | 2016-06-08 | 清华大学 | Energy consumption optimized collaborative task migration method in mobile cloud computing environment |
| CN107332913A (en) * | 2017-07-04 | 2017-11-07 | 电子科技大学 | A kind of Optimization deployment method of service function chain in 5G mobile networks |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10491528B2 (en) * | 2016-10-27 | 2019-11-26 | Hewlett Packard Enterprise Development Lp | Selectively monitoring a network of network function chains based on probability of service level agreement violation |
-
2018
- 2018-12-27 CN CN201811620365.9A patent/CN109831319B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102932279A (en) * | 2012-10-30 | 2013-02-13 | 北京邮电大学 | Multidimensional resource scheduling system and method for cloud environment data center |
| CN105656999A (en) * | 2015-12-25 | 2016-06-08 | 清华大学 | Energy consumption optimized collaborative task migration method in mobile cloud computing environment |
| CN107332913A (en) * | 2017-07-04 | 2017-11-07 | 电子科技大学 | A kind of Optimization deployment method of service function chain in 5G mobile networks |
Non-Patent Citations (3)
| Title |
|---|
| "5G网络切片资源编排与故障管理技术研究";李凌书;《中国优秀硕士学位论文全文数据库信息科技辑》;20180615;I136-458 * |
| "一种基于最优匹配的测试资源动态配置方法";金天 等;《信息工程大学学报》;20100615;第322-325页 * |
| "基于元能力的可重构网络功能编排技术研究";谢立军;《中国博士学位论文全文数据库信息科技辑》;20180615;正文第3-5章 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109831319A (en) | 2019-05-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bagaa et al. | Service-aware network function placement for efficient traffic handling in carrier cloud | |
| Taleb et al. | User mobility-aware virtual network function placement for virtual 5G network infrastructure | |
| Di Mauro et al. | Availability evaluation of multi-tenant service function chaining infrastructures by multidimensional universal generating function | |
| Mechtri et al. | VNF placement and chaining in distributed cloud | |
| Al-Tam et al. | Fractional switch migration in multi-controller software-defined networking | |
| CN111538587A (en) | Service function chain reconfiguration method based on load balancing | |
| CN102055667A (en) | Methods and apparatus for configuring a virtual network switch | |
| Globa et al. | Quality control for mobile communication management services in hybrid environment | |
| CN109831319B (en) | Network function deployment method considering multidimensional resource constraints | |
| Bayrakdar et al. | Artificial bee colony–based spectrum handoff algorithm in wireless cognitive radio networks | |
| CN108833227B (en) | Intelligent home communication optimal scheduling system and method based on edge calculation | |
| Skulysh et al. | Managing the process of servicing hybrid telecommunications services. Quality control and interaction procedure of service subsystems | |
| Tsvetkov et al. | Model of restoration of the communication network using the technology of software defined networks | |
| Churyumov et al. | Method for ensuring survivability of flying ad-hoc network based on structural and functional reconfiguration | |
| Huang et al. | Distributed resource allocation for network slicing of bandwidth and computational resource | |
| Amiri et al. | Optimizing virtual network function splitting in open-RAN environments | |
| Do et al. | Latency-aware placement for state management functions in service-based 5G mobile core network | |
| Tinnaluri et al. | Edge-Cloud Computing Systems for Unmanned Aerial Vehicles Capable of Optimal Work Offloading with Delay | |
| Xiao et al. | DMSSM: A Decision-Making Scheme of Switch Migration for SDN Control Plane | |
| CN108923961B (en) | Multi-entry network service function chain optimization method | |
| WO2018194888A1 (en) | Dynamic computer network classification using machine learning | |
| CN110224873B (en) | NFV (network virtual function) arranging method and device based on VNF (virtual network context) instance multiplexing | |
| Mudvari et al. | ML-driven scaling of 5G Cloud-Native RANs | |
| CN106789289B (en) | Method and apparatus for virtual network mapping | |
| US11622322B1 (en) | Systems and methods for providing satellite backhaul management over terrestrial fiber |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |