CN113676909A - Virtual network function universal scheduling method under 5G/B5G environment - Google Patents

Virtual network function universal scheduling method under 5G/B5G environment Download PDF

Info

Publication number
CN113676909A
CN113676909A CN202110816504.0A CN202110816504A CN113676909A CN 113676909 A CN113676909 A CN 113676909A CN 202110816504 A CN202110816504 A CN 202110816504A CN 113676909 A CN113676909 A CN 113676909A
Authority
CN
China
Prior art keywords
vnf
service
equation
network
chain
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.)
Pending
Application number
CN202110816504.0A
Other languages
Chinese (zh)
Inventor
王兴伟
李康玲
易波
黄敏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northeastern University China
Original Assignee
Northeastern University China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Northeastern University China filed Critical Northeastern University China
Priority to CN202110816504.0A priority Critical patent/CN113676909A/en
Publication of CN113676909A publication Critical patent/CN113676909A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
    • H04W16/02—Resource partitioning among network components, e.g. reuse partitioning
    • H04W16/10—Dynamic resource partitioning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/15—Correlation function computation including computation of convolution operations
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
    • H04W16/22—Traffic simulation tools or models

Abstract

The invention provides a general virtual network function scheduling method under a 5G/B5G environment, and provides a general VNF scheduling method for reasonably scheduling different flows using the same VNF instance object, so as to maximize the utilization rate of VNF, wherein the general VNF scheduling method comprises the following steps: firstly, establishing a new service function chain model based on a minimum additive algebraic theory, and supporting VNF sharing among different service chains; secondly, on the basis of a service function chain model, a general scheduling method VNF-S for VNF is provided, which is used for allocating reasonable execution time slices for simultaneously arriving traffic belonging to different service chains and supporting reallocation of allocated resources in an idle state, thereby maximizing the utilization rate of the VNF. The method of the invention not only can improve the utilization rate of VNF and reduce the quantity of VNF instances deployed in the network, but also can improve the utilization rate of network resources to a great extent, and meets the application requirements of huge connection and large flow under the environment of 5G/B5G.

Description

Virtual network function universal scheduling method under 5G/B5G environment
Technical Field
The invention belongs to the technical field of traffic scheduling under the condition of virtual network function sharing in a 5G/B5G environment, and particularly relates to a software defined network, a network function virtualization technology, a VNF sharing concept, a service function chain, a minimum additive algebra theory and a VNF scheduling method.
Background
In the large environment of 5G/B5G, each Network element assumes a certain Network Function (NF) in the communication Network. However, in the face of the situation of large connection and large traffic, the processing of Network functions is weak, and this problem exists for each Network Function, so a Virtual Network Function (VNF) is formed by virtualizing these Network elements by a virtualization technology, and a common method is proposed to solve the problem of various Network functions. Under the large background of 5G/B5G nowadays, Software Defined Networking (SDN) and Network Function Virtualization (NFV) technologies bring numerous advantages and simultaneously improve the diversity and dynamics of traffic in a network, thereby bringing a series of challenges to optimization of service Function chains. By introducing the concept of VNF sharing, the utilization rate of the VNF is improved, the number of VNF instances deployed in the network is reduced, and therefore the utilization rate of network resources is improved to the great extent. However, since a single VNF instance object may be used by multiple service function chains at the same time, the following problems arise: how should the VNF common to the service chains handle the different traffic when the traffic of the service function chains arrives at the same time at a certain time?
Aiming at the problem, the invention provides a general scheduling method for the VNF, and different flows using the same VNF instance object are reasonably scheduled, so that the utilization rate of the VNF is maximized. Specifically, the corresponding time slice is allocated for the traffic that uses the current VNF object and belongs to different service chains, so as to perform corresponding processing on the arriving traffic. Aiming at a service function chain of a shared VNF instance object, a new service chain performance model is provided based on a minimum additive algebraic Theory (Min-plus Algebra Theory, MAT), and a VNF sequence is integrated into a serial service system by the model; on the basis, a VNF-S based general scheduling method is provided, which performs resource allocation according to the size of a service chain and supports the reallocation of allocated resources in an idle state, thereby maximizing the utilization rate of network resources.
Disclosure of Invention
The invention aims to solve the problem of how a VNF shared by a plurality of service function chains can handle different traffics when the traffics of the service function chains arrive at the same time at a certain moment in a 5G/B5G environment, namely the traffic scheduling problem under the VNF sharing condition in a 5G/B5G environment.
The technical scheme adopted by the invention is as follows:
a general scheduling method of virtual network function under 5G/B5G environment, taking the service capability of VNF as a resource, allocating reasonable execution time slices for the arriving traffic belonging to different service chains, realizing effective allocation of resource, maximizing the utilization rate of VNF, including establishing a new service function chain model and the general scheduling method of VNF proposed on the basis of the model, the main steps include:
step 1, establishing a network topology model;
step 2, virtual network function VNF;
step 3, establishing a service function chain model based on a minimum additive algebra theory;
step 4, establishing a performance index model, wherein the performance index model comprises data backlog and time delay;
and 5, establishing a new VNF scheduling method based on the service function chain model.
The network topology model in step 1 is represented by using a graph G (N, L), where N represents a set of virtual machines for running a VNF in a network, and L represents a virtual link between the virtual machines. Where each virtual machine supports running one or more different VNF instances simultaneously.
The step 2 comprises the following steps:
step 2.1 Using Unicode FVTo represent multiple corresponding instance objects simultaneously existing in any one VNF in the network, using
Figure RE-GDA0003265277540000021
Representing a networkThe i-th VNF present in (1), use
Figure RE-GDA0003265277540000022
To represent
Figure RE-GDA0003265277540000023
The jth instance object of (1);
step 2.2 establishment
Figure RE-GDA0003265277540000024
The flow model of (2); the statistics at time intervals (τ, t) are calculated using equation (1)]Internal cumulative arrival
Figure RE-GDA0003265277540000025
Flow of (2) by symbols
Figure RE-GDA0003265277540000026
To represent; at time intervals [0, t]Internal cumulative departure
Figure RE-GDA0003265277540000027
Traffic usage symbol of
Figure RE-GDA0003265277540000028
It shows that since the cumulative flow leaving the VNF cannot exceed the cumulative total flow arriving, its relationship to the inflow flow is shown in equation (2); at a time interval (tau, t)]In the interior of said container body,
Figure RE-GDA0003265277540000029
efficient service capability usage notation that can be provided
Figure RE-GDA00032652775400000210
A representation whose correlation model with the outflow is shown in equation (3);
Figure RE-GDA00032652775400000211
Figure RE-GDA00032652775400000212
Figure RE-GDA00032652775400000213
wherein
Figure RE-GDA00032652775400000214
Is shown at time interval [0, t]Internal running to instance object
Figure RE-GDA00032652775400000215
The flow (in bit) of (1), the value of which is non-negative;
step 2.3 according to the relationship between the inflow flow and the outflow flow, the method can obtain
Figure RE-GDA00032652775400000216
The relationship between the incoming traffic, the outgoing traffic, and the effective service capability, as shown in equation (4):
Figure RE-GDA00032652775400000217
step 2.4 introduces the concept of MAT based on equation (4), converting equation (4) to equation (5):
Figure RE-GDA00032652775400000218
wherein
Figure RE-GDA00032652775400000219
Representing convolution operations in the MAT.
The step 3 comprises the following steps:
step 3.1, defining a representation symbol of a service function chain set in the network, as shown in formula (6); defining arbitrary service function chain requests ΨpThe VNF requirement of (2), as shown in equation (7):
Ψ＝{Ψp|p∈[1，|Ψ|]} (6)
Figure RE-GDA00032652775400000220
wherein
Figure RE-GDA00032652775400000221
As described in step 2
Figure RE-GDA00032652775400000222
The method comprises the steps of having a certain mapping relation, wherein the former represents VNF requirements of a service chain, and the latter represents VNF instance objects which are already deployed in a network;
Figure RE-GDA00032652775400000223
denotes ΨpThe required qth VNF;
step 3.2 focuses on analyzing VNF performance, so for any service chain, it is assumed that no route blocking exists between two adjacent VNFs, and based on this assumption and the explanation of step 2, formula (8) can be obtained:
Figure RE-GDA00032652775400000224
step 3.3 because the VNF is shared, VNF instance objects already deployed in the network
Figure RE-GDA00032652775400000225
Can be used by a plurality of different service chains, and therefore also needs to be matched
Figure RE-GDA00032652775400000226
Is allocated, the allocated service capability uses symbols
Figure RE-GDA00032652775400000227
And (4) showing. Based on MAT, the concrete expression is shown as formula (9):
Figure RE-GDA00032652775400000228
wherein
Figure RE-GDA0003265277540000031
The representation is any service function chain ΨpAssigned service capabilities of
Figure RE-GDA0003265277540000032
And (4) forming.
The step 4 comprises the following steps:
step 4.1 defines the cumulative arrival at the first VNF
Figure RE-GDA0003265277540000033
The flow symbol of
Figure RE-GDA0003265277540000034
Leave the last VNF i.e.
Figure RE-GDA0003265277540000035
The flow symbol of
Figure RE-GDA0003265277540000036
Introduction of a widely used constrained envelope function to define
Figure RE-GDA0003265277540000037
The upper boundary of (c) is specifically shown in formula (10):
Figure RE-GDA0003265277540000038
where ρ (> 0) represents the arrival rate of data traffic, σ is the data burst parameter;
step 4.2, based on the boundary function in step 4.1, defining two important performance index models, namely data backlog and time delay; data backlog indicator
Figure RE-GDA0003265277540000039
Is described as shown in equation (11), the delay index of the service function chain
Figure RE-GDA00032652775400000310
As shown in equation (12):
Figure RE-GDA00032652775400000311
Figure RE-GDA00032652775400000312
wherein ε represents an offset time;
step 4.3 converts the data backlog and the time delay index into the maximum supremum form according to the formula (4) in step 2.3, as shown in formula (13) and formula (14), respectively:
Figure RE-GDA00032652775400000313
Figure RE-GDA00032652775400000314
wherein, formula (13) represents the maximum upper bound form of data backlog, and formula (14) represents the maximum upper bound form of time delay.
The step 5 comprises the following steps:
step 5.1, in order to reasonably and fairly allocate the service processing capability of the VNF to each service function chain using the VNF, a weight is allocated to each service function chain according to the size of the service function chain, and ω is used for each service function chain1，ω2，…，ω|Ψ|Represents;
step 5.2 instantiates any given VNF
Figure RE-GDA00032652775400000315
Service capability (i.e. of
Figure RE-GDA00032652775400000316
) Discretized in time series, where this service capability can be considered as a resource and assigned to a service chain that uses the VNF at the same time, so
Figure RE-GDA00032652775400000317
Is shown in equation (15):
Figure RE-GDA00032652775400000318
step 5.3 if backlog occurs continuously in an instance object, two service chains Ψ are defined that use the VNF instance object togetherpAnd Ψp′The relationship between service capabilities of (a) is shown in equation (16):
Figure RE-GDA00032652775400000319
step 5.4, as a plurality of service function chains can not fully utilize the distributed service capacity, the resource which is not fully utilized is redistributed and used for serving other service function chains with backlog phenomenon to maximize the service/resource utilization rate;
step 5.5, according to the weight of the service chain, the service capability of the VNF is equally allocated or re-allocated to each service chain requiring the service capability.
Compared with the prior art, the invention has the beneficial effects that: by introducing the concept of VNF sharing, the utilization rate of the VNF can be improved, the number of VNF instances deployed in a network can be reduced, and the utilization rate of network resources can be improved to a great extent; and a general scheduling method is provided for the VNF to handle how the VNF handles the different traffic problems when the traffic of the service function chain arrives at the same time at a certain time, and reasonably schedule the different traffic using the same VNF instance object, thereby maximizing the utilization rate of the VNF and meeting the application requirements of macro-connection and large traffic in the 5G/B5G environment.
Drawings
FIG. 1 is an architectural diagram of the present invention.
FIG. 2 is an exemplary graph of latency and data backlog according to the present invention.
Fig. 3 is an exemplary diagram of average service chain delay under different topologies of the present invention.
FIG. 4 is an exemplary graph of data backlog and throughput in accordance with the present invention.
Fig. 5 is a flow chart of the service chain usage conclusion of the present invention.
Detailed Description
In order to meet the requirement of huge connection and large flow in the 5G/B5G environment, one of the problems that must be solved is the scheduling problem under the VNF sharing condition. In order to solve the problem, the invention firstly establishes a new service function chain model based on MAT, and the model supports VNF sharing among different service chains; then, on the basis of a service function chain model, a general scheduling method for the VNF is provided, which is used for allocating reasonable execution time slices for arriving traffic belonging to different service chains, so as to maximize the utilization rate of the VNF.
Referring to the architecture diagram shown in fig. 1, the method includes the following aspects:
1. virtual network function VNF
1.1 establishing a flow model
1.1.1 inflow Rate
Given an arbitrary VNF instance object
Figure RE-GDA0003265277540000041
At time intervals [0, t]In this example, the traffic (in bits) usage symbols that reach the instance object are accumulated
Figure RE-GDA0003265277540000042
And (4) showing. It is clear that,
Figure RE-GDA0003265277540000043
the function is not negative and there is no decrement. On the basis, in order to describe the arriving traffic more finely, the pair is needed
Figure RE-GDA0003265277540000044
And (5) completing. Thus, a symbol is introduced
Figure RE-GDA0003265277540000045
To represent the time interval (tau, t)]Internal cumulative arrival
Figure RE-GDA0003265277540000046
The flow rate of (c). Wherein t is more than tau and is more than or equal to 0. Taking these two functions together, equation (1) is derived:
Figure RE-GDA0003265277540000047
1.1.2 outflow Rate
Similar to the inflow rate, for the time interval 0, t]Internal cumulative departure
Figure RE-GDA0003265277540000048
Traffic usage symbol of
Figure RE-GDA0003265277540000049
And (4) showing. Since the cumulative flow leaving the VNF cannot exceed the cumulative total flow arriving, it has the conclusion as equation (2):
Figure RE-GDA00032652775400000410
for equation (2), on the one hand, when
Figure RE-GDA00032652775400000411
When the flow reaches
Figure RE-GDA00032652775400000412
Faster than the speed of departureTo thereby cause
Figure RE-GDA00032652775400000413
The redundant data cannot be processed in time, so that the data backlog is caused; on the other hand, when
Figure RE-GDA00032652775400000414
Then, it means
Figure RE-GDA00032652775400000415
The backlog of data in (1) continues to remain balanced.
1.1.3 efficient service capability of VNF
To be provided with
Figure RE-GDA0003265277540000051
And
Figure RE-GDA0003265277540000052
the association is performed assuming that each VNF is in a continuous operation state, that is, the VNFs will operate all the time as long as there is data traffic waiting for processing. For the
Figure RE-GDA0003265277540000053
At a time interval (tau, t)]In which the effective service capability provided by it uses symbols
Figure RE-GDA0003265277540000054
And (4) showing. From equation (1), the same can be found
Figure RE-GDA0003265277540000055
Figure RE-GDA0003265277540000056
1.1.4 correlation model
For the
Figure RE-GDA0003265277540000057
Given arbitrary two time points τ and t (> τ), and falseLet these two points in time belong to the same busy period. Then during this time it is possible to,
Figure RE-GDA0003265277540000058
will operate at full load. In other words,
Figure RE-GDA0003265277540000059
all available service capabilities will be used to handle the arriving data traffic. Based on this situation, establish
Figure RE-GDA00032652775400000510
And
Figure RE-GDA00032652775400000511
the correlation model between them, as shown in formula (3); in formula (3), assuming τ and t as the starting and ending time points of the last busy period, respectively, then there is a formula (4) conclusion; substituting equation (4) into equation (3) yields equation (5):
Figure RE-GDA00032652775400000512
Figure RE-GDA00032652775400000513
Figure RE-GDA00032652775400000514
since τ is generally unknown, equation (5) is converted to a general form as shown in equation (6):
Figure RE-GDA00032652775400000515
in addition, due to leaving
Figure RE-GDA00032652775400000516
Does not exceed the total amount of data reached, i.e.
Figure RE-GDA00032652775400000517
Substituting it into the formula (3), and generalizing it to obtain the following formula (7):
Figure RE-GDA00032652775400000518
by combining equation (6) and equation (7), the conclusion is reached as equation (8):
Figure RE-GDA00032652775400000519
1.2 minimum additive algebraic theory MAT
Based on equation (8), the concept of MAT is introduced. Specifically, MAT replaces the addition operation with the minimization operation and the multiplication operation with the addition operation, respectively, and then based on MAT, equation (8) can be converted into equation (9):
Figure RE-GDA00032652775400000520
in the formula (9), the reaction mixture,
Figure RE-GDA00032652775400000521
representing convolution operations in the MAT.
2. MAT-based service function chain model establishment
Any number of subsystem sequences can be easily integrated into a complete series system based on the convolution operation in the MAT. From the service function chain point of view it is composed of several VNFs in a certain order. If each VNF is considered as a separate subsystem, the service function chain composed of them can be considered as a complete tandem system. Therefore, the present application builds a service function chain model based on MAT.
Service function chain set in network uses Ψ ═ Ψ { Ψ }p|p∈[1，|Ψ|]Represents it. Request Ψ for any service function chainpIts VNF requirements are expressed as follows:
Figure RE-GDA00032652775400000522
wherein the content of the first and second substances,
Figure RE-GDA00032652775400000523
with respect to flow models
Figure RE-GDA00032652775400000524
There is a certain mapping relationship, the former represents VNF requirements of the service chain, and the latter represents VNF instance objects already deployed in the network. In the formula (10), the first and second groups,
Figure RE-GDA00032652775400000525
denotes ΨpThe q-th VNF needed. Are used separately
Figure RE-GDA00032652775400000526
And
Figure RE-GDA00032652775400000527
indicating arrival and departure within a certain time interval
Figure RE-GDA00032652775400000528
And these flows belong to ΨpSince the present application focuses on analyzing the performance of VNFs, for any service chain, assuming that no route blocking exists between every two adjacent VNFs, the conclusion as shown in equation (11) can be obtained:
Figure RE-GDA0003265277540000061
VNF instance objects already deployed in the network may be according to a specific VNF deployment policyIs used by a plurality of different service chains, so that the requirement is that
Figure RE-GDA0003265277540000062
To each service chain using it, the following conclusions can be drawn: given an arbitrary service function chain ΨpThe flow rate of which is to be passed through in sequence
Figure RE-GDA0003265277540000063
Let ΨpThe allocated service capability is
Figure RE-GDA0003265277540000064
It is composed of
Figure RE-GDA0003265277540000065
The composition is specifically expressed as formula (12):
Figure RE-GDA0003265277540000066
the process of the overall conclusion is shown in the flow chart of fig. 5, wherein athrough(t) represents traffic belonging to the current service chain, and AcrossAnd (t) represents the traffic belonging to other service chains.
3. Establishing a performance index model including data backlog and time delay
Connecting service function chain ΨpViewed as a series system combined by VNFs, then in the time interval (τ, t)]In, the first VNF is reached cumulatively (i.e.
Figure RE-GDA0003265277540000067
) And leave the last VNF (i.e. the
Figure RE-GDA0003265277540000068
) Respectively using the flow rates of
Figure RE-GDA0003265277540000069
And
Figure RE-GDA00032652775400000610
and (4) showing. Next, a definition is made based on a widely used constrained envelope function
Figure RE-GDA00032652775400000611
As shown in formula (13):
Figure RE-GDA00032652775400000612
where ρ (> 0) represents the arrival rate of data traffic and σ (> 0) is the data burst parameter.
Based on the boundary function, two important performance index models are defined, namely data backlog and time delay. Symbol for data backlog indicator
Figure RE-GDA00032652775400000613
Indicating, while time delays use symbols
Figure RE-GDA00032652775400000614
And (4) showing. Fig. 2 shows a graph of arrival traffic versus departure traffic over time for a service chain, wherein the horizontal arrows indicate arrival traffic and the vertical arrows indicate departure traffic.
3.1 data backlog
The backlog of data represents the sum of traffic waiting in the queue and in the process. In fig. 2, given time t (corresponding to coordinate X axis), the difference between the longitudinal directions of the two curves is the total amount of data currently retained. Thus, the index is described as equation (14), which is extended to equation (15) according to equation (8) in 1.1.4:
Figure RE-GDA00032652775400000615
Figure RE-GDA00032652775400000616
wherein equation (15) gives the maximum supremum of the backlog index, the function
Figure RE-GDA00032652775400000617
Is given by equation (13), function
Figure RE-GDA00032652775400000618
The calculation of (c) will be described later.
3.2 time delay
For the delay of the service function chain, given the data volume (corresponding to the Y-axis of the coordinates) in fig. 2, the difference between the two curves in the horizontal direction is the time required for these data volumes to move from entry to exit. Therefore, the delay metric of the service function chain is formulated as equation (16), where ε represents the offset time. For a certain traffic, the time it takes from entering to completely leaving the VNF is considered as the time delay required for the VNF to process. Also, equation (16) can be extended according to equation (8) in 1.1.4, as shown in equation (17):
Figure RE-GDA00032652775400000619
Figure RE-GDA00032652775400000620
4. VNF scheduling method based on service function chain model
For any in the network
Figure RE-GDA00032652775400000621
It may be shared by multiple different service function chains, in other words, there may be multiple
Figure RE-GDA00032652775400000622
Is mapped to
Figure RE-GDA00032652775400000623
The above.To establish this mapping relationship, the following variables are defined:
Figure RE-GDA00032652775400000624
wherein the value 1 represents
Figure RE-GDA00032652775400000625
Is mapped to
Figure RE-GDA00032652775400000626
And 0 indicates none. Then, the following constraints exist:
Figure RE-GDA0003265277540000071
in order to reasonably and fairly allocate the service processing capability of the VNF to each service function chain using the VNF, each service function chain is allocated a weight (in a proportional relationship) according to the size of the service function chain, and ω is used for each service function chain1，ω2，…，ω|Ψ|And (4) showing. Given any VNF instance
Figure RE-GDA0003265277540000073
Will have its service capability (i.e. the
Figure RE-GDA0003265277540000074
) Discretizing according to a time sequence. This service capability can then be considered as a resource and allocated to a service chain that uses the VNF at the same time. From this, it is derived
Figure RE-GDA0003265277540000075
Is calculated as follows:
Figure RE-GDA0003265277540000076
given arbitrary
Figure RE-GDA0003265277540000077
It is used by two service function chains, respectively psipOf (i.e. the qth VNF)
Figure RE-GDA0003265277540000078
) And Ψp′Of (i.e. the q' th VNF)
Figure RE-GDA0003265277540000079
). If ΨpAt a time interval (tau, t)]In and at
Figure RE-GDA00032652775400000710
In which backlog occurs continuously, then the assignment to ΨpAnd Ψp′The relationship between service capabilities of (c) is described as follows:
Figure RE-GDA00032652775400000711
to avoid loss of generality, assume all uses
Figure RE-GDA00032652775400000712
Are in a set, using symbols
Figure RE-GDA00032652775400000713
And (4) showing. Thus, for all
Figure RE-GDA00032652775400000714
Generalizing equation (21) as follows:
Figure RE-GDA00032652775400000715
since each VNF can be considered as a continuously operating server, then, assuming it is running continuously at a fixed rate, then:
Figure RE-GDA00032652775400000716
substituting equation (20) and equation (23) into equation (22), and expanding them yields equation (24):
Figure RE-GDA00032652775400000717
however, equation (24) assumes that all service function chains are able to fully utilize the assigned service capabilities, i.e., they are continuously backlogged. In fact, many service function chains cannot fully utilize the service capabilities allocated to them. Therefore, the underutilized resources can be reallocated for serving other service function chains with backlog phenomena, thereby achieving the purpose of maximizing service/resource utilization rate. Given any such considerations
Figure RE-GDA00032652775400000718
It is linked by multiple different service functions
Figure RE-GDA00032652775400000719
Simultaneous use, using sets, provided that there is a part of a continuously backlogged chain of services among them
Figure RE-GDA00032652775400000720
To express, then, by perfecting the formula (24), the formula (25) can be obtained:
Figure RE-GDA00032652775400000721
based on the above planning, the service capability of the VNF can be evenly allocated (or re-allocated) to each service chain using it according to the weight of the service chain.
5. Evaluation of example
The VNF scheduling method VNF-S is realized by adopting Python language.
5.1 simulation topology
In order to evaluate and verify the VNF-S, the present application uses three network topologies in the simulation: the topology is a small-scale topology, and the topology comprises 5 virtual machines and 2 service chains which are expressed by using a Network I; the medium-scale topology comprises 10 virtual machines and 6 service chains, and is represented by a Network II; large-scale topology, with 20 virtual machines and 12 service chains, is represented using Network III.
5.2 comparison of references
When the performance of the VNF-S method is evaluated, two reference methods are selected for comparison: the First is First Come First Serve (FCFS) method, which processes strictly according to the arrival order of data traffic; the second utilizes Genetic Algorithms (GA) to solve VNF scheduling problems.
5.3 evaluation index
Three performance evaluation indexes are adopted: scheduling time, average service delay, data backlog, and throughput.
The average service delay consists of three parts, namely traffic scheduling delay, data processing delay and propagation delay, and the propagation delays calculated by the three methods are consistent as the VNF instance is assumed to be deployed in the network and the service function path is set, so that the scheduling delay and the processing delay only need to be compared; since the data backlog and throughput result trend obtained in the three topologies with different scales are approximately the same, the VNF-S is evaluated by taking Network III as a simulation topology, and the relationship between the data backlog and throughput is discussed.
The first two indexes are smaller in value and better in VNF-S performance, and the last index is required to have higher throughput and lower data backlog to explain the superiority of the VNF-S performance.
5.4 evaluation results
5.4.1 scheduling time
Please refer to table 1 for the scheduling time calculated by the three methods. The FCFS processes the corresponding data traffic according to the arrival order of the service requests. However, once the size of the pending request is particularly large (e.g., elephant flow), FCFS takes a long time to process the request, and FCFS can only service other requests after processing the request. For smaller streams, such as rat streams, they may need to wait for a long period of time. And the frequent occurrence of this will eventually result in an increase in the total scheduling time, so the FCFS scheduling time is longest. The GA-based scheduling algorithm aims to minimize the latency by iterative computation, and its complexity is less than the FCFS, so it can be estimated that the GA scheduling time is less than the FCFS. The scheduling complexity of the VNF-S is the same as the GA, however the VNF-S allows sharing of the same VNF instance between different service chains, and the scheduling time of the GA is also affected by the number of iterations and initialization population size, whereas the VNF-S does not have such constraints. Also, too few iterations or too small a population size may result in the GA not obtaining the expected results. VNF-S is superior to GA.
TABLE 1 scheduling time (ms) in different networks for three methods
Figure RE-GDA0003265277540000081
5.4.2 average service delay
The results of the averaging delay are shown in fig. 3. The average time delay consists of three parts, namely traffic scheduling time delay, data processing time delay and propagation time delay, wherein the propagation time delays of the three methods are consistent, the scheduling time delay shows the result in 5.4.1, and the data processing time delay is mainly determined by respective adopted strategies. FCFS simply serves according to the order of arrival of data, and thus takes a long time; the GA preferentially services the small-scale data streams, thus reducing their latency and thus the overall processing latency, which is nonetheless affected by its inherent characteristics (e.g., based on encoding and decoding); the VNF-S supports sharing the same VNF instance between different service chains, and may allocate the service capabilities of the same VNF to the service function chain using it, and in addition, if one or more service function chains do not fully use the allocated service resources, the VNF-S reallocates these idle resources for processing backlogged data. Based on this consideration, the VNF-S can reduce the data processing latency to a large extent. Thus, the average delay achieved by the VNF-S is minimal.
5.4.3 data backlog and throughput
The results of data backlog and throughput are shown in figure 4. In fig. 4, the first observed phenomenon is that the data backlog grows slower as throughput increases faster (e.g., before time point 70) and vice versa (e.g., after time point 80). If there is a backlog of data, this means that the VNF is in a state of full load operation. This naturally leads to a steady increase in throughput, and therefore the first phenomenon is reasonable; the second observed phenomenon is that VNF-S has higher throughput and lower data backlog compared to GA and FCFS. On the one hand, it is reasonable to have both higher throughput and lower data backlog; VNF-S, on the other hand, proposes sharing deployed VNF instances. Based on this consideration, the VNF-S can reclaim and reallocate idle resources to a service chain with data backlog, thereby further improving the utilization of network resources, and on the contrary, the FCFS and GA do not consider the reclamation and reallocation of resources, so their network resource utilization and throughput are lower than those of the VNF-S.
The invention is superior to the reference method in the aspects of scheduling time, average service delay, data backlog and throughput, and the superiority of the invention in performance is illustrated.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.

Claims (6)

1. A virtual network function universal scheduling method under 5G/B5G environment is characterized by mainly comprising the following steps:
step 1, establishing a network topology model;
step 2, virtual network function VNF;
step 3, establishing a service function chain model based on a minimum additive algebra theory;
step 4, establishing a performance index model, wherein the performance index model comprises data backlog and time delay;
and 5, establishing a new VNF scheduling method based on the service function chain model.
2. The method of claim 1, wherein the network topology model in step 1 is represented by a graph G (N, L), where N represents a set of virtual machines running VNF in the network, and L represents a virtual link between the virtual machines; where each virtual machine supports running one or more different VNF instances simultaneously.
3. The method for universal scheduling of virtual network functions in a 5G/B5G environment according to claim 1, wherein the step 2 comprises the steps of:
step 2.1 Using Unicode FVTo represent multiple corresponding instance objects simultaneously existing in any one VNF in the network, using
Figure FDA0003170304190000011
Indicating the presence of a VNF of the i-th kind in the network, using
Figure FDA0003170304190000012
To represent
Figure FDA0003170304190000013
The jth instance object of (1);
step 2.2 establishment
Figure FDA0003170304190000014
The flow model of (2); the statistics at time intervals (τ, t) are calculated using equation (1)]Internal cumulative arrival
Figure FDA0003170304190000015
Flow of (2) by symbols
Figure FDA0003170304190000016
To represent; at time intervals [0, t]Internal cumulative departure
Figure FDA0003170304190000017
Traffic usage symbol of
Figure FDA0003170304190000018
It shows that since the cumulative flow leaving the VNF cannot exceed the cumulative total flow arriving, its relationship to the inflow flow is shown in equation (2); at a time interval (tau, t)]In the interior of said container body,
Figure FDA0003170304190000019
efficient service capability usage notation that can be provided
Figure FDA00031703041900000110
A representation whose correlation model with the outflow is shown in equation (3);
Figure FDA00031703041900000111
Figure FDA00031703041900000112
Figure FDA00031703041900000113
wherein
Figure FDA00031703041900000114
Is shown at time interval [0, t]Internal running to instance object
Figure FDA00031703041900000115
The flow rate of (a) to (b),
Figure FDA00031703041900000116
the unit of (1) is bit, the value of which is not negative;
step 2.3 according to the relationship between the inflow flow and the outflow flow, the method can obtain
Figure FDA00031703041900000117
The relationship between the incoming traffic, the outgoing traffic, and the effective service capability, as shown in equation (4):
Figure FDA00031703041900000118
step 2.4 introduces the concept of MAT based on equation (4), converting equation (4) to equation (5):
Figure FDA00031703041900000119
wherein
Figure FDA00031703041900000120
Representing convolution operations in the MAT.
4. The method for universal scheduling of virtual network functions in a 5G/B5G environment according to claim 1, wherein the step 3 comprises the steps of:
step 3.1, defining a representation symbol of a service function chain set in the network, as shown in formula (6); defining arbitrary service function chain requests ΨpThe VNF requirement of (2), as shown in equation (7):
Ψ＝{Ψp|p∈[1，|Ψ|]} (6)
Figure FDA0003170304190000021
wherein
Figure FDA0003170304190000022
As described in step 2
Figure FDA0003170304190000023
The method comprises the steps of having a certain mapping relation, wherein the former represents VNF requirements of a service chain, and the latter represents VNF instance objects which are already deployed in a network;
Figure FDA0003170304190000024
denotes ΨpThe required qth VNF;
step 3.2 focuses on analyzing VNF performance, so for any service chain, it is assumed that no route blocking exists between two adjacent VNFs, and based on this assumption and the explanation of step 2, formula (8) can be obtained:
Figure FDA0003170304190000025
step 3.3 because the VNF is shared, VNF instance objects already deployed in the network
Figure FDA0003170304190000026
Can be used by a plurality of different service chains, and therefore also needs to be matched
Figure FDA0003170304190000027
Is allocated, the allocated service capability uses symbols
Figure FDA0003170304190000028
Represents; based on MAT, the concrete expression is shown as formula (9):
Figure FDA0003170304190000029
wherein
Figure FDA00031703041900000210
The representation is any service function chain ΨpAssigned service capabilities of
Figure FDA00031703041900000211
And (4) forming.
5. The method for universal scheduling of virtual network functions in a 5G/B5G environment according to claim 1, wherein the step 4 comprises the steps of:
step 4.1 defines the cumulative arrival at the first VNF
Figure FDA00031703041900000212
The flow symbol of
Figure FDA00031703041900000213
Leave the last VNF i.e.
Figure FDA00031703041900000214
The flow symbol of
Figure FDA00031703041900000215
A widely used constrained Envelope Function (defined Envelope Function) is introduced to define
Figure FDA00031703041900000216
The upper boundary of (c) is specifically shown in formula (10):
Figure FDA00031703041900000217
where ρ (> 0) represents the arrival rate of data traffic, σ is the data burst parameter;
step 4.2, based on the boundary function in step 4.1, defining two important performance index models, namely data backlog and time delay; data backlog indicator
Figure FDA00031703041900000218
Is described as shown in equation (11), the delay index of the service function chain
Figure FDA00031703041900000219
As shown in equation (12):
Figure FDA00031703041900000220
Figure FDA00031703041900000221
wherein ε represents an offset time;
step 4.3 converts the data backlog and the time delay index into the maximum supremum form according to the formula (4) in step 2.3, as shown in formula (13) and formula (14), respectively:
Figure FDA00031703041900000222
Figure FDA00031703041900000223
wherein, formula (13) represents the maximum upper bound form of data backlog, and formula (14) represents the maximum upper bound form of time delay.
6. The method for universal scheduling of virtual network functions in a 5G/B5G environment according to claim 1, wherein the step 5 comprises the steps of:
step 5.1, in order to reasonably and fairly allocate the service processing capability of the VNF to each service function chain using the VNF, a weight is allocated to each service function chain according to the size of the service function chain, and ω is used for each service function chain1，ω2，…，ω|Ψ|Represents;
step 5.2 instantiates any given VNF
Figure FDA0003170304190000031
Service capability of
Figure FDA0003170304190000032
Discretized in time series, where this service capability can be considered as a resource and assigned to a service chain that uses the VNF at the same time, so
Figure FDA0003170304190000033
Is shown in equation (15):
Figure FDA0003170304190000034
step 5.3 if backlog occurs continuously in an instance object, two service chains Ψ are defined that use the VNF instance object togetherpAnd Ψp′The relationship between service capabilities of (a) is shown in equation (16):
Figure FDA0003170304190000035
step 5.4, as a plurality of service function chains can not fully utilize the distributed service capacity, the resource which is not fully utilized is redistributed and used for serving other service function chains with backlog phenomenon to maximize the service/resource utilization rate;
step 5.5, according to the weight of the service chain, the service capability of the VNF is equally allocated or re-allocated to each service chain requiring the service capability.
CN202110816504.0A 2021-07-20 2021-07-20 Virtual network function universal scheduling method under 5G/B5G environment Pending CN113676909A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110816504.0A CN113676909A (en) 2021-07-20 2021-07-20 Virtual network function universal scheduling method under 5G/B5G environment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110816504.0A CN113676909A (en) 2021-07-20 2021-07-20 Virtual network function universal scheduling method under 5G/B5G environment

Publications (1)

Publication Number Publication Date
CN113676909A true CN113676909A (en) 2021-11-19

Family

ID=78539857

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110816504.0A Pending CN113676909A (en) 2021-07-20 2021-07-20 Virtual network function universal scheduling method under 5G/B5G environment

Country Status (1)

Country Link
CN (1) CN113676909A (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108965024A (en) * 2018-08-01 2018-12-07 重庆邮电大学 A kind of virtual network function dispatching method of the 5G network slice based on prediction
CN110505099A (en) * 2019-08-28 2019-11-26 重庆邮电大学 A kind of service function chain dispositions method based on migration A-C study
US20200236038A1 (en) * 2019-01-18 2020-07-23 Rise Research Institutes of Sweden AB Dynamic Deployment of Network Applications Having Performance and Reliability Guarantees in Large Computing Networks

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108965024A (en) * 2018-08-01 2018-12-07 重庆邮电大学 A kind of virtual network function dispatching method of the 5G network slice based on prediction
US20200236038A1 (en) * 2019-01-18 2020-07-23 Rise Research Institutes of Sweden AB Dynamic Deployment of Network Applications Having Performance and Reliability Guarantees in Large Computing Networks
CN110505099A (en) * 2019-08-28 2019-11-26 重庆邮电大学 A kind of service function chain dispositions method based on migration A-C study

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
BO YI: "A Generalized VNF Sharing Approach for Service Scheduling", 《IEEE COMMUNICATIONS LETTERS》, pages 1 - 4 *
孙士清;彭建华;游伟;: "一种时延约束下可靠性保证的服务功能链部署方法", 计算机应用研究, no. 08 *
易波: "基于SDN和NFV的网络服务编排机制研究", 《博士电子期刊》 *

Similar Documents

Publication Publication Date Title
CN110275758B (en) Intelligent migration method for virtual network function
CN108566659B (en) 5G network slice online mapping method based on reliability
CN103605567B (en) Cloud computing task scheduling method facing real-time demand change
CN111475252B (en) Virtual network function deployment optimization method based on deep reinforcement learning
CN109660376B (en) Virtual network mapping method, equipment and storage medium
CN109104373B (en) Method, device and system for processing network congestion
CN104540234B (en) A kind of associated task scheduling mechanism synchronously constrained based on CoMP under C RAN frameworks
CN102780581B (en) AFDX (Avionics Full Duplex Switched Ethernet) end-to-end delay bound claculation method based on random network calculus
Hanczewski et al. A queueing model of a multi-service system with state-dependent distribution of resources for each class of calls
Li Resource optimization scheduling and allocation for hierarchical distributed cloud service system in smart city
CN111010295B (en) SDN-MEC-based power distribution and utilization communication network task migration method
CN106775949B (en) Virtual machine online migration optimization method capable of sensing composite application characteristics and network bandwidth
CN109697122A (en) Task processing method, equipment and computer storage medium
CN110191155B (en) Parallel job scheduling method, system and storage medium for fat tree interconnection network
Shi et al. Real-time communication analysis with a priority share policy in on-chip networks
CN113535393B (en) Computing resource allocation method for unloading DAG task in heterogeneous edge computing
CN106598727B (en) A kind of computational resource allocation method and system of communication system
CN104079502A (en) Multi-user multi-queue scheduling method
CN113938930B (en) Construction method of virtual network function forwarding graph adapting to 5G network multi-service scene
CN100593312C (en) Scheduling using quantum and deficit values
CN112291791A (en) Power communication mesh bandwidth resource allocation method based on 5G slice
CN101827446B (en) Radio bearer scheduling method and device
CN110048966B (en) Coflow scheduling method for minimizing system overhead based on deadline
CN113946455A (en) Multi-stage feedback queue flow scheduling method based on bottleneck perception
Wided et al. A novel agent based load balancing model for maximizing resource utilization in grid computing

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