CN113742902A - Multi-parameter performance modeling evaluation method based on network calculation - Google Patents

Multi-parameter performance modeling evaluation method based on network calculation Download PDF

Info

Publication number
CN113742902A
CN113742902A CN202110955293.9A CN202110955293A CN113742902A CN 113742902 A CN113742902 A CN 113742902A CN 202110955293 A CN202110955293 A CN 202110955293A CN 113742902 A CN113742902 A CN 113742902A
Authority
CN
China
Prior art keywords
flow
flow table
curve
arrival
module
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.)
Granted
Application number
CN202110955293.9A
Other languages
Chinese (zh)
Other versions
CN113742902B (en
Inventor
李福亮
王皓
张欣书
王兴伟
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
CERNET Corp
Original Assignee
Northeastern University China
CERNET Corp
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, CERNET Corp filed Critical Northeastern University China
Priority to CN202110955293.9A priority Critical patent/CN113742902B/en
Publication of CN113742902A publication Critical patent/CN113742902A/en
Application granted granted Critical
Publication of CN113742902B publication Critical patent/CN113742902B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/50—Reducing energy consumption in communication networks in wire-line communication networks, e.g. low power modes or reduced link rate

Abstract

The invention discloses a multi-parameter performance modeling evaluation method based on network calculation, which mainly analyzes a typical scene of a software defined network and builds a model for a packet forwarding process of an OpenFlow switch based on the network calculation. Introducing a plurality of performance influence parameters including an arrival flow parameter, a competition flow parameter, a switch processing rate and a flow table hit rate, performing function derivation based on minimum additive algebra operation, and respectively calculating forwarding delay boundaries of flows hitting and not hitting the flow table. The method can evaluate the performance of the OpenFlow switch after being applied to a real network environment by presetting relevant network state parameters and substituting the relevant network state parameters into a relation function calculation mode, and can observe the influence degree of different parameters on the performance by controlling variables, thereby providing guidance for the performance optimization and the flow control of the switch.

Description

Multi-parameter performance modeling evaluation method based on network calculation
Technical Field
The invention belongs to the technical field of network performance evaluation, and mainly aims at a multi-parameter forwarding delay performance modeling evaluation method based on network calculation designed for an OpenFlow switch in a software defined network; in particular to a flow analysis method of a typical scene in a software defined network, a model construction method based on network calculation and a multi-parameter derivation method.
Background
The network performance evaluation is an important work for guiding network structure optimization and equipment management, and can verify the correctness, validity and reliability of a network optimization strategy. In the traditional performance evaluation scheme, a simulation platform is set up to simulate a real network environment and obtain a measurement result, but the problems of high experiment overhead, slow progress, poor feasibility and the like exist, and the simulation experiment cannot deeply analyze the influence of different optimization strategies on specific performance parameters and cannot explain the influence relationship among different parameters, so that a scientific network performance modeling evaluation method is particularly important.
The software defined network separates the control plane and the data plane, improves the flexibility of network management and configuration, and simultaneously causes the performance bottleneck of the network. Performance evaluation for OpenFlow switches in software-defined networks is an important research direction for optimizing software-defined networks. The OpenFlow switch carries out packet forwarding according to the flow table, and competition of forwarding resources exists in the switch for different flows. The new mode of centralized management of multiple switches by a single controller further aggravates the resource competition, introduces more network performance influencing factors, and causes difficulty in performance evaluation work. Aiming at the problem, most of the existing modeling evaluation methods are modeled based on queuing theory, and the distribution of packet arrival interval time, the distribution of forwarding service time and the forwarding service supply number are represented by using an 'X/Y/Z' symbol. However, the queuing theory model is highly dependent on a specific distribution hypothesis, the evaluation result only reflects the average state under the obedience probability distribution, and the uncertainty of packet arrival in the real network and the dynamic characteristics of the network are diluted, so that the application range of the corresponding evaluation model and the reliability of the evaluation result are greatly reduced.
The network algorithm is a network performance analysis tool based on a minimum additive algebra structure, and the arrival characteristics of traffic and the service characteristics of a system are extracted by recording the input and output cumulative functions of a network system in a period of time. The network algorithm is applied to the forwarding delay analysis of a Software Defined network for the first time in document 1(Azodolmolky S, Nejabati R, Pazouki M, et al. an analytical model for Software Defined Networking: A network calulus-based approach [ C ]. GLOBECOM, 2013: 1397-. However, researchers cannot deeply analyze the influence of the controller on packet forwarding, and simply define the influence as a linear function, and the differentiated services provided by the switch on the traffic of different flow table hit conditions cannot be embodied in the proposed model. Document 2(Lin C T, Wu C M, Huang M, et al. Performance Evaluation for SDN delivery: an applied basic on storage Network calculation [ J ]. China Communications,2016, Supplement (1):98-106.) proposes a Stochastic Network Calculus model, in which an arrival flow function expression under normal exponential and logarithmic distributions is derived. Researchers have failed to teach how to use this function to further analyze packet forwarding delays and have limited practical implications. Document 3 (koohanesi a K, Osgouei a G, Saidi H, et al. an Analytical Model for Delay Bound of OpenFlow Based SDN using Network call [ J ]. Journal of Network and Computer Applications,2017,96:31-38.) analyzes the influence of the switch flow table cache space on packet forwarding Delay in detail using a Network algorithm method to build a Model. However, there are many delay-influencing parameters besides the buffer space of the flow table, and the analysis of the influencing parameters by the model is not perfect.
The invention defines the packet forwarding action of a typical scene of a network by analyzing software, introduces a plurality of performance influence parameters to establish a network calculation model, introduces parameters including arrival flow parameters, competition flow parameters, switch processing speed and flow table hit rate, and further establishes a relation function between each parameter and forwarding delay through operation deduction. Compared with the existing method, the method has the advantages that the analysis of the time delay influence parameters is more comprehensive, the hit condition of the flow is refined and distinguished, the more accurate performance boundary value can be obtained, and the reliable reference basis is provided for the application and optimization of the OpenFlow switch.
Disclosure of Invention
The invention aims to provide a method for evaluating the performance of an OpenFlow switch and accurately calculating the forwarding delay boundary of the switch according to a plurality of state parameters in a software defined network. The method can evaluate the performance of the equipment before the switch is applied to the actual network, improve the management efficiency, realize the control of the network influence parameters and guide the network optimization.
The technical scheme of the invention is as follows: a multi-parameter performance modeling evaluation method based on network calculation comprises the steps of OpenFlow switch grouping forwarding scene analysis, network calculation model construction and multi-parameter derivation based on minimum additive arithmetic, and mainly comprises the following steps:
step 1, analyzing a packet forwarding flow in a typical scene, wherein the packet forwarding flow comprises a packet processing link in an OpenFlow switch, a flow table hit condition, the influence of a controller on packet forwarding and the influence of a QoS (quality of service) queue on packet output, establishing a network calculation model, and determining parameters of the switch;
step 2, defining characteristic attributes of the arriving flow, including a flow table hit state and an output queue, so as to distinguish the flow arriving at the OpenFlow switch; setting target flow and defining flow parameters;
step 3, according to the flow hypothesis and the characteristic distinction, deducing an arrival curve of the flow arriving at the switch, wherein the arrival curve comprises flows hitting and missing a flow table;
and 4, according to a packet forwarding flow under a typical scene, stripping forwarding resource competition interference of different flows in the OpenFlow switch, deriving equivalent service curves of flows hitting the flow table and flows not hitting the flow table in each module of the switch, and calculating the forwarding delay of the flows by combining the arrival curves.
The specific steps of the step 1 comprise:
step 1.1, analyzing a packet forwarding flow, dividing an OpenFlow switch into a flow table query module, a packet scheduling module and a QoS output queue, and establishing a controller and an OpenFlow switch model by using an abstract form of a waiting queue-server;
step 1.2, adding a flow trend line into a network calculation model according to a packet forwarding flow so as to represent the processing process of packets and the resource competition conditions of different flows;
step 1.3, defining parameters of the switch, including processing rates of modules of the OpenFlow switch: flow table query module processing rate d1Packet scheduling module processing rate d2Processing rate of output queue i
Figure BDA0003220251750000031
And controller processing rate dC。
The specific steps of the step 2 comprise:
step 2.1, the characteristic attributes for flow differentiation include a flow table hit state and an output queue, the flow table hit state determines whether the packet needs to be processed by the controller, and the output queue determines the QoS service obtained by the packet. According to the flow quantity, the flow of the output queue i is selected as the target flow quantity without loss of generality, the parts of the flow table which hit and miss are analyzed respectively, and the output queues except i are j;
step 2.2, defining parameters of the stream, including the arrival rate and the burst size of each stream of the data plane: arrival rate ρ of trafficfBurst size σ of flow ffThe arrival rate and burst size of the control plane contention flow: rate of arrival ρ of competing streamsOAmount of bursts of competing streams σO。
The specific steps of the step 3 comprise:
and 3.1, calculating an arrival curve alpha (t) of the flow according to an input accumulation function A (t) of each flow arriving at the OpenFlow switch. Where t represents the time span and A (t) represents the cumulative amount of arriving traffic over time, measured using the number of packets arriving. According to the formula (1), [0, t]Any time point s within the interval should satisfy this equation, thereby obtaining the arrival curve α (t) of the corresponding stream. Arrival curve alpha of flow ff(t) can be simplified as represented by formula (2):
A(t)-A(s)≤α(t-s)，0≤s≤t (1)
αf(t)＝σf+ρft，t≥0 (2)
step 3.2, respectively summing the arrival curves of the flows hitting and missing the flow table according to the arrival curve composite rule of formula (3), wherein alphatotalfArrival curve of the finger complex:
αf～(σf，ρf)→αtotalf～(∑σf，∑ρf) (3)
the specific steps of the step 4 comprise:
step 4.1, accumulating function A according to the output of each module*(t), a service curve β (t) is calculated. Wherein A is*(t) represents the cumulative amount of output traffic over time, measured using the number of packets output. Calculating a service curve beta (t) of the corresponding module according to equation (4), including the service curve beta of the flow table query module1Packet scheduling module service curve beta2Output queue i service curve
Figure BDA0003220251750000042
And controller service curve betaC. The service curve can be simplified as expressed by equation (5), where r represents the maximum output rate of the corresponding module, and b represents the maximum time interval for packet waiting:
Figure BDA0003220251750000041
β(t)＝r(t-b)+，t≥0 (5)
wherein if t > b, (t-b)+Not, t-b, otherwise (t-b)+＝0；
Step 4.2, according to the FCFS equivalent service curve derivation method of the formula (6), the resource competition influence of the flow of the non-output queue i and the flow of the missed flow table at the flow table query module is stripped, and the equivalent service curve of the flow of the hit flow table at the module is calculated
Figure BDA0003220251750000051
βeq，1(t，s)＝{β(t)-α2(t-s)}+，0≤s＜t (6)
Figure BDA0003220251750000052
Step 4.3, calculating the arrival curve alpha of the flow sent back by the controller at the packet scheduling module according to the characteristic limit rule of the output flow by combining the formula (8)R：
Figure BDA0003220251750000053
Step 4.4, calculating the arrival curve of the hit flow of the non-output queue i at the packet scheduling module according to the formula (9)
Figure BDA0003220251750000054
Figure BDA0003220251750000055
Step 4.5, according to the FCFS equivalent service curve derivation method of the formula (6), the resource competition influence of the flow sent back by the controller and the hit flow of the non-output queue i at the packet scheduling module is stripped, and the equivalent service curve of the flow hitting the flow table at the module is calculated
Figure BDA0003220251750000056
Figure BDA0003220251750000057
Step 4.6, bandwidth respectively allocated to flow of hit flow table and flow of miss flow table according to QoS output queue
Figure BDA0003220251750000058
Figure BDA0003220251750000059
Calculating equivalent service curve of hit flow in the module
Figure BDA00032202517500000510
Figure BDA00032202517500000511
Step 4.7, based on the derivation results of the formulas (7), (10) and (11), converting the service curves of the three modules into equivalent service curves of the flow table hitting flow table and passing through the whole OpenFlow switch system through minimum convolution operation
Figure BDA0003220251750000061
Figure BDA0003220251750000062
Step 4.8, calculating hit flow arrival curve
Figure BDA00032202517500000612
And equivalent service curves
Figure BDA00032202517500000613
Maximum horizontal distance between them, i.e. Delay bound Delay of a flow through a switch hitting the flow tableH：
Figure BDA0003220251750000063
The latency bound for a flow that misses the flow table is derived as follows:
step 4.9, according to the FCFS equivalent service curve derivation method of the formula (6), the resource competition influence of the flow of the output queue j and the flow of the hit flow table at the flow table query module is stripped, and the equivalent service curve of the flow of the missed flow table at the module is calculated
Figure BDA0003220251750000064
Figure BDA0003220251750000065
Step 4.10, strip the resource competition influence of the competition flow of the control plane at the controller, calculate the equivalent service curve of the flow which does not hit the flow table at the controller
Figure BDA0003220251750000066
Figure BDA0003220251750000067
wherein ,
Figure BDA00032202517500000614
the equivalent service curve at the flow table lookup module for the part of the missing flow table of the other output queue j can be calculated by equation (16);
Figure BDA00032202517500000615
the arrival curve burst characteristic parameter for the part of the stream arriving at the controller can be calculated by the formula (17):
Figure BDA0003220251750000068
Figure BDA0003220251750000069
step 4.11, stripping the resource competition influence of the flow hitting the flow table and the flow of the output queue j sent back by the controller at the packet scheduling module, and calculating the service curve of the flow not hitting the flow table at the local module
Figure BDA00032202517500000610
Figure BDA00032202517500000611
wherein ,
Figure BDA0003220251750000071
an equivalent service curve at the controller for the portion of the missing flow table of output queue j, in the same way as equation (15), can be obtained:
Figure BDA0003220251750000072
step 4.12, according to the bandwidth allocation of the QoS output queue, calculating the equivalent service curve of the non-hit flow in the module
Figure BDA0003220251750000073
Figure BDA0003220251750000074
Step 4.13, based on the derivation results of equations (14), (15), (18) and (20), calculating an equivalent service curve of the flow missing the flow table passing through the whole OpenFlow switch system
Figure BDA0003220251750000075
Figure BDA0003220251750000076
Step 4.14, similar to equation (13), calculate the miss flow arrival curve
Figure BDA0003220251750000077
And equivalent service curves
Figure BDA0003220251750000078
The maximum horizontal distance between the flow tables, and the time delay boundary of the flow which misses the flow table and passes through the switch is obtained.
The invention has the main beneficial effects that: the packet forwarding delay boundary of the OpenFlow switch can be calculated by constructing a network calculation model by combining the flow parameters and the switch parameters in the network. The flows of the hit flow table and the miss flow table are calculated differently, and the dimensionality of performance evaluation is enriched. The invention can evaluate the performance before the switch is applied to a real network environment, is beneficial to judging whether the switch deployment strategy is reasonable, objectively reflects the influence of a plurality of parameters on the performance of the OpenFlow switch, and provides guidance for the performance optimization and the flow control of the OpenFlow switch.
Drawings
Fig. 1 is a "waiting queue-server" abstraction example of a network operation.
Fig. 2 is an OpenFlow switch model based on network calculus.
Fig. 3A and 3B are diagrams illustrating the impact of the arrival rate and burst size of a flow hitting the flow table on latency bounds, where: fig. 3A is a diagram illustrating the delay boundary effect of a stream hitting a stream table; fig. 3B illustrates the latency boundary effect of a flow that misses in the flow table.
Fig. 4A and 4B are diagrams illustrating the impact of the arrival rate and burst size of a flow that misses in the flow table on latency bounds, where: FIG. 4A is a diagram of latency bound impact of a flow hitting a flow table; fig. 4B shows the latency boundary effect of a flow that misses in the flow table.
Fig. 5A and 5B are diagrams illustrating the impact of the arrival rate and burst size of a flow on the delay bound for a control plane, wherein: fig. 5A is a diagram illustrating the delay boundary effect of a stream hitting a stream table; fig. 5B illustrates the latency boundary effect of a flow that misses in the flow table.
Detailed Description
The invention provides a multi-parameter performance modeling evaluation method based on network calculation. The method utilizes a network calculus principle to establish an OpenFlow switch model, clears the resource competition relationship of different flows, defines a target flow and a plurality of parameters influencing the packet forwarding performance of the switch, further deduces an arrival curve and an equivalent service curve of the target flow through minimum additive operation, and obtains a time delay boundary of the flow passing through the whole switch system. The key steps involved in the process of the present invention are described in detail below.
1. Abstract modeling
As shown in fig. 1, a network algorithm may abstract a study object as a "waiting queue-server" combination. Wherein, the 'waiting queue' represents the buffer space of the system, records the arrival condition of the packet, and generates an input accumulation function A (t); "Server" represents the processing function of the system, records the processing condition of the packet, and generates the output accumulation function A*(t)。
Because the OpenFlow switch provides different forwarding services for the flow of the hit flow table and the flow of the miss flow table, the invention selects to respectively calculate the delay boundaries of the two flows for enriching the performance evaluation dimensionality of the switch. Thus, OpenFlow switches cannot be easily abstracted as a set of "waiting queue-servers". In a typical scene, the forwarding process inside the switch is analyzed in detail, each functional module is abstracted into a waiting queue-server, the resource competition relationship of different flows is clarified, parameters influencing the performance are further clarified, and the target flow is established. The specific implementation steps are as follows:
1.1 typical scene analysis
In the software defined network, forwarding paths of packets are uniformly calculated by a controller and are issued to related OpenFlow switches. When the packet arrives at the switch, firstly, the flow table is inquired, and the flow table entry is matched according to the message information of the packet header. In general, the matching can be successful and the forwarding operation can be obtained. And then, executing a forwarding operation, sending the packet to a corresponding output queue, and finally executing an output operation. And the unsuccessfully matched Packet is packaged into a Packet-in message and sent to the controller, and the controller performs routing calculation. After the calculation is completed, the controller sends back the Flow-mod message and the Packet-out message again to guide the switch to update the Flow table, and the successful matching is realized.
Based on the packet forwarding processing flow of the OpenFlow switch, the interior of the switch can be divided into three modules: a flow table query module, a grouping scheduling module and a QoS output queue. The three modules and the controller are respectively abstracted to be a waiting queue-server, and an OpenFlow switch model as shown in FIG. 2 is constructed. The relevant parameters of the switch include: the processing rate of the three modules and the processing rate of the controller.
1.2 establishing a target flow
The characteristic attributes for differentiation of flows include a flow table hit status, which determines whether a packet needs to be processed by the controller, and an output queue, which determines a QoS service that the packet acquires, by which the flow is differentiated. The contention of different flows at each module of the switch is shown in fig. 2, and the flow of the output queue i is selected as the target flow without loss of generality, and the portions of the hit and miss flow tables are analyzed respectively. The relevant parameters of the flow include the arrival rate and burst size of each flow of the data plane, and the arrival rate and burst size of the contention flow of the control plane.
2. Minimum additive algebra operation
The minimum additive algebra operation specific to network calculus simplifies the function derivation process. The core operation rule used in the function derivation process of the invention is as follows:
minimum convolution addition operation:
Figure BDA0003220251750000091
minimum add deconvolution operation:
Figure BDA0003220251750000092
3. function derivation
In order to enrich the performance evaluation dimensionality of the switch, the invention respectively carries out function derivation on the flows of hit and miss flow tables, derives respective arrival curves and equivalent service curves, and further calculates the forwarding delay boundary of the two flows. The specific implementation steps are as follows:
3.1 arrival Curve
And calculating an arrival curve alpha (t) of the flow according to an input accumulation function A (t) of each flow arriving at the OpenFlow switch. Where t represents the time span and A (t) represents the cumulative amount of arriving traffic over time, measured using the number of packets arriving. According to the formula (1), [0, t]At any time s within the interval, the equation is satisfied, and the arrival curve α (t) of the corresponding flow is obtained). Arrival curve alpha of flow ff(t) can be simplified as represented by formula (2):
A(t)-A(s)≤α(t-s)，0≤s≤t (1)
αf(t)＝σf+ρft，t≥0 (2)
step 3.2 sum the arrival curves of flows hitting and missing flow tables, respectively, according to the arrival curve composite rule of equation (3), where αtotalfArrival curve of the finger complex:
αf～(σf，ρf)→αtotalf～(∑σf，∑ρf) (3)
3.2 latency bound values for flows hitting the flow table
Accumulating function A according to the output of each module*(t), a service curve β (t) is calculated. Wherein A is*(t) represents the cumulative amount of output traffic over time, measured using the number of packets output. Calculating a service curve beta (t) of the corresponding module according to equation (4), including the service curve beta of the flow table query module1Packet scheduling module service curve beta2Output queue i service curve
Figure BDA0003220251750000102
And controller service curve betaC. The service curve can be simplified as expressed by equation (5), where r represents the maximum output rate of the corresponding module, and b represents the maximum time interval for packet waiting:
Figure BDA0003220251750000101
β(t)＝r(t-b)+，t≥0 (5)
wherein if t > b, (t-b)+Not, t-b, otherwise (t-b)+＝0；
According to the FCFS equivalent service curve derivation method of the formula (6), the resource competition influence of the flow of the non-output queue i and the flow of the missed flow table at the flow table query module is stripped, and the equivalent service curve of the flow of the hit flow table at the module is calculated
Figure BDA0003220251750000111
βeq，1(t，s)＝{β(t)-α2(t-s)}+，0≤s＜t (6)
Figure BDA0003220251750000112
According to the characteristic limit rule of the output stream, the combined formula (8) calculates the arrival curve alpha of the stream sent back by the controller at the packet scheduling moduleR：
Figure BDA0003220251750000113
Calculating the arrival curve of the hit flow of the non-output queue i at the packet scheduling module according to the formula (9)
Figure BDA0003220251750000114
Figure BDA0003220251750000115
According to the FCFS equivalent service curve derivation method of the formula (6), the resource competition influence of the flow sent back by the stripping controller and the hit flow of the non-output queue i at the packet scheduling module is calculated, and the equivalent service curve of the flow hitting the flow table at the module is calculated
Figure BDA0003220251750000116
Figure BDA0003220251750000117
Bandwidth allocated to flows hitting and missing flow tables respectively according to QoS output queue
Figure BDA0003220251750000118
Calculating equivalent service curve of hit flow in the module
Figure BDA0003220251750000119
Figure BDA00032202517500001110
Based on the derivation results of the formulas (7), (10) and (11), converting the service curves of the three modules into equivalent service curves of flows which hit the flow table and pass through the whole OpenFlow switch system through the minimum convolution operation
Figure BDA00032202517500001111
Figure BDA00032202517500001112
Calculating hit flow arrival curves
Figure BDA00032202517500001213
And equivalent service curves
Figure BDA00032202517500001214
Maximum horizontal distance between them, i.e. Delay bound Delay of a flow through a switch hitting the flow tableH：
Figure BDA0003220251750000121
3.3 latency bound values for flows that miss the flow table
According to the FCFS equivalent service curve derivation method of the formula (6), the resource competition influence of the flow of the output queue j and the flow of the hit flow table at the flow table query module is stripped, and the equivalent service curve of the flow of the missed flow table at the module is calculated
Figure BDA0003220251750000122
Figure BDA0003220251750000123
Stripping resource competition influence of competition flows of the control plane at the controller, and calculating an equivalent service curve of the flow which does not hit the flow table at the controller
Figure BDA0003220251750000124
Figure BDA0003220251750000125
wherein ,
Figure BDA0003220251750000126
the equivalent service curve at the flow table lookup module for the part of the missing flow table of the other output queue j can be calculated by equation (16);
Figure BDA00032202517500001215
the arrival curve burst characteristic parameter for the part of the stream arriving at the controller can be calculated by the formula (17):
Figure BDA0003220251750000127
Figure BDA0003220251750000128
the resource competition influence of the flow hitting the flow table and the flow of the output queue j sent back by the controller at the packet scheduling module is stripped, and the service curve of the flow not hitting the flow table at the local module is calculated
Figure BDA0003220251750000129
Figure BDA00032202517500001210
wherein ,
Figure BDA00032202517500001211
an equivalent service curve at the controller for the portion of the missing flow table of output queue j, in the same way as equation (15), can be obtained:
Figure BDA00032202517500001212
calculating the equivalent service curve of the missed flow in the module according to the bandwidth allocation of the QoS output queue
Figure BDA0003220251750000131
Figure BDA0003220251750000132
Calculating an equivalent service curve of a flow missing the flow table through the whole OpenFlow switch system based on the derivation results of the equations (14), (15), (18) and (20)
Figure BDA0003220251750000133
Figure BDA0003220251750000134
The miss flow arrival curve is calculated in the same manner as equation (13)
Figure BDA0003220251750000136
And equivalent service curves
Figure BDA0003220251750000137
The maximum horizontal distance between the flow tables, and the time delay boundary of the flow which misses the flow table and passes through the switch is obtained.
4. Evaluation of example
In order to show the specific implementation effect of the patent, the patent uses the performance parameters of a Beacon OpenFlow controller and a standard virtual switch OpenvSwitch to carry out a test experiment.
4.1 parameter Range of variation
This patent has adopted the flow parameter variation range of universal setting among the evaluation of OpenFlow switch performance, includes: the arrival rate of the flows hitting and missing the flow table, the burst amount of the flows hitting and missing the flow table, the arrival rate of the control plane contention flow, the burst amount of the control plane contention flow, the processing rate of each module of the switch, and the processing rate of the controller. The specific parameter values can be referred to the following table:
table 1 simulation experiment parameter set-up
Figure BDA0003220251750000135
Figure BDA0003220251750000141
Some parameter settings are referred to in document [1 ]. And substituting the derived function under different flow parameter values, and generating a time delay boundary calculation result by using Matlab. The calculation results of this experiment are shown in fig. 3A to 5B.
4.2 evaluation results
Experiments are performed on the beacon openflow controller and the standard virtual switch OpenvSwitch, and a hit flow, a miss flow and a control plane contention flow are respectively tested in a method, and the test results refer to fig. 3A to 5B.
The test result accurately quantifies the influence degree of different parameters on the forwarding delay boundary of the flow which hits and misses the flow table, which shows that the invention can comprehensively evaluate the forwarding performance of the OpenFlow switch and provide guidance for the further optimization and flow control of the switch.
The above description of exemplary embodiments has been presented only to illustrate the technical solution of the invention and is not intended to be exhaustive or to limit the invention to the precise form described. Obviously, many modifications and variations are possible in light of the above teaching to those skilled in the art. The exemplary embodiments were chosen and described in order to explain certain principles of the invention and its practical application to thereby enable others skilled in the art to understand, implement and utilize the invention in various exemplary embodiments and with various alternatives and modifications. It is intended that the scope of the invention be defined by the following claims and their equivalents.

Claims (5)

1. A multi-parameter forwarding delay performance modeling evaluation method based on network calculation is characterized by comprising the following steps:
step 1, analyzing a packet forwarding flow in a typical scene, wherein the packet forwarding flow comprises a packet processing link in an OpenFlow switch, a flow table hit condition, the influence of a controller on packet forwarding and the influence of a QoS (quality of service) queue on packet output, establishing a network calculation model, and determining parameters of the switch;
step 2, defining characteristic attributes of the arriving flow, including a flow table hit state and an output queue, so as to distinguish the flow arriving at the OpenFlow switch; setting target flow and defining flow parameters;
step 3, according to the flow hypothesis and the characteristic distinction, deducing an arrival curve of the flow arriving at the switch, wherein the arrival curve comprises flows hitting and missing a flow table;
and 4, according to a packet forwarding flow under a typical scene, stripping forwarding resource competition interference of different flows in the OpenFlow switch, deriving equivalent service curves of flows hitting the flow table and flows not hitting the flow table in each module of the switch, and calculating the forwarding delay of the flows by combining the arrival curves.
2. The network-calculus-based multi-parameter forwarding delay performance modeling evaluation method according to claim 1, wherein the specific steps of step 1 include:
step 1.1, analyzing a packet forwarding flow, dividing an OpenFlow switch into a flow table query module, a packet scheduling module and a QoS output queue, and establishing a controller and an OpenFlow switch model by using an abstract form of a waiting queue-server;
step 1.2, adding a flow trend line into a network calculation model according to a packet forwarding flow so as to represent the processing process of packets and the resource competition conditions of different flows;
step 1.3, defining parameters of the switch, including processing rates of modules of the OpenFlow switch: flow table query module processing rate d1Packet scheduling module processing rate d2Processing rate of output queue i
Figure FDA0003220251740000011
And controller processing rate dC。
3. The network-calculus-based multi-parameter forwarding delay performance modeling evaluation method according to claim 1 or 2, wherein the specific steps of step 2 include:
step 2.1, the characteristic attribute used for distinguishing the flow includes the hit state of the flow table and output queue, the hit state of the flow table determines whether the grouping needs to be processed by the controller, the output queue determines the QoS service obtained by the grouping; according to the flow quantity, the flow of the output queue i is selected as the target flow quantity without loss of generality, the parts of the flow table which hit and miss are analyzed respectively, and the output queues except i are j;
step 2.2, defining parameters of the stream, including the arrival rate and the burst size of each stream of the data plane: arrival rate ρ of flow ffBurst size σ of flow ffThe arrival rate and burst size of the control plane contention flow: rate of arrival ρ of competing streamsoAmount of bursts of competing streams σo。
4. The network-calculus-based multi-parameter forwarding delay performance modeling evaluation method according to claim 3, wherein the specific steps of step 3 include:
step 3.1, reaching OpenFlow switch according to each flowInputting an accumulation function A (t), and calculating an arrival curve alpha (t) of the flow; where t represents the time span, A (t) represents the cumulative amount of arrival traffic over time, measured using the number of packets arriving; according to the formula (1), [0, t]Any time point s in the interval should satisfy the formula, so as to obtain an arrival curve alpha (t) of the corresponding flow; arrival curve alpha of flow ff(t) is represented by the formula (2) in a simplified manner:
A(t)-A(s)≤α(t-s)，0≤s≤t (1)
αf(t)＝σf+ρft，t≥0 (2)
step 3.2, respectively summing the arrival curves of the flows hitting and missing the flow table according to the arrival curve composite rule of formula (3), wherein alphatotalfArrival curve of the finger complex:
αf～(σf，ρf)→αtotalf～(∑σf，∑ρf) (3)。
5. the network-calculus-based multi-parameter forwarding delay performance modeling evaluation method according to claim 4, wherein the specific steps of step 4 include:
step 4.1, accumulating function A according to the output of each module*(t), calculating a service curve β (t); wherein A is*(t) represents the cumulative amount of output traffic over time, measured using the number of packets output; calculating a service curve beta (t) of the corresponding module according to equation (4), including the service curve beta of the flow table query module1Packet scheduling module service curve beta2Output queue i service curve
Figure FDA0003220251740000021
And controller service curve betaC(ii) a The service curve is simplified as equation (5), where r represents the maximum output rate of the corresponding module, and b represents the maximum time interval for packet waiting:
Figure FDA0003220251740000022
β(t)＝r(t-b)+，t≥0 (5)
wherein if t > b, (t-b)+Not, t-b, otherwise (t-b)+＝0；
Step 4.2, according to the FCFS equivalent service curve derivation method of the formula (6), the resource competition influence of the flow of the non-output queue i and the flow of the missed flow table at the flow table query module is stripped, and the equivalent service curve of the flow of the hit flow table at the module is calculated
Figure FDA0003220251740000031
βeq，1(t，s)＝{β(t)-α2(t-s)}+，0≤s＜t (6)
Figure FDA0003220251740000032
Step 4.3, calculating the arrival curve alpha of the flow sent back by the controller at the packet scheduling module according to the characteristic limit rule of the output flow by combining the formula (8)R：
Figure FDA0003220251740000033
Step 4.4, calculating the arrival curve of the hit flow of the non-output queue i at the packet scheduling module according to the formula (9)
Figure FDA0003220251740000034
Figure FDA0003220251740000035
Step 4.5, according to the FCFS equivalent service curve derivation method of the formula (6), resource competition influence of the flow sent back by the stripping controller and the hit flow of the non-output queue i at the packet scheduling module is eliminated, and the flow of the hit flow table is calculated in the local moduleEquivalent service curve of block
Figure FDA0003220251740000036
Figure FDA0003220251740000037
Step 4.6, bandwidth respectively allocated to flow of hit flow table and flow of miss flow table according to QoS output queue
Figure FDA0003220251740000038
Figure FDA0003220251740000039
Calculating equivalent service curve of hit flow in the module
Figure FDA00032202517400000310
Figure FDA00032202517400000311
Step 4.7, based on the derivation results of the formulas (7), (10) and (11), converting the service curves of the three modules into equivalent service curves of the flow table hitting flow table and passing through the whole OpenFlow switch system through minimum convolution operation
Figure FDA0003220251740000041
Figure FDA0003220251740000042
Step 4.8, calculating hit flow arrival curve
Figure FDA0003220251740000043
And equivalent service curves
Figure FDA0003220251740000044
Maximum horizontal distance between them, i.e. Delay bound Delay of a flow through a switch hitting the flow tableH：
Figure FDA0003220251740000045
Step 4.9, according to the FCFS equivalent service curve derivation method of the formula (6), the resource competition influence of the flow of the output queue j and the flow of the hit flow table at the flow table query module is stripped, and the equivalent service curve of the flow of the missed flow table at the module is calculated
Figure FDA0003220251740000046
Figure FDA0003220251740000047
Step 4.10, strip the resource competition influence of the competition flow of the control plane at the controller, calculate the equivalent service curve of the flow which does not hit the flow table at the controller
Figure FDA0003220251740000048
Figure FDA0003220251740000049
wherein ,
Figure FDA00032202517400000410
calculating an equivalent service curve at the flow table query module for the part of the missed flow table of the other output queue j by using an equation (16);
Figure FDA00032202517400000411
for the arrival curve of the part of the flow at the controllerThe emission characteristic parameter is calculated by the formula (17):
Figure FDA00032202517400000412
Figure FDA00032202517400000413
step 4.11, stripping the resource competition influence of the flow hitting the flow table and the flow of the output queue j sent back by the controller at the packet scheduling module, and calculating the service curve of the flow not hitting the flow table at the local module
Figure FDA00032202517400000414
Figure FDA00032202517400000415
wherein ,
Figure FDA00032202517400000416
an equivalent service curve at the controller for the portion of the missing flow table of output queue j, in the same way as equation (15), can be obtained:
Figure FDA0003220251740000051
step 4.12, according to the bandwidth allocation of the QoS output queue, calculating the equivalent service curve of the non-hit flow in the module
Figure FDA0003220251740000052
Figure FDA0003220251740000053
Step 4.13, based on the derivation results of equations (14), (15), (18) and (20), calculating an equivalent service curve of the flow missing the flow table passing through the whole OpenFlow switch system
Figure FDA0003220251740000054
Figure FDA0003220251740000055
Step 4.14, similar to equation (13), calculate the miss flow arrival curve
Figure FDA0003220251740000056
And equivalent service curves
Figure FDA0003220251740000057
The maximum horizontal distance between the flow tables, and the time delay boundary of the flow which misses the flow table and passes through the switch is obtained.
CN202110955293.9A 2021-08-19 2021-08-19 Multi-parameter performance modeling evaluation method based on network algorithm Active CN113742902B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110955293.9A CN113742902B (en) 2021-08-19 2021-08-19 Multi-parameter performance modeling evaluation method based on network algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110955293.9A CN113742902B (en) 2021-08-19 2021-08-19 Multi-parameter performance modeling evaluation method based on network algorithm

Publications (2)

Publication Number Publication Date
CN113742902A true CN113742902A (en) 2021-12-03
CN113742902B CN113742902B (en) 2023-08-04

Family

ID=78731800

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110955293.9A Active CN113742902B (en) 2021-08-19 2021-08-19 Multi-parameter performance modeling evaluation method based on network algorithm

Country Status (1)

Country Link
CN (1) CN113742902B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114244680A (en) * 2021-12-17 2022-03-25 网络通信与安全紫金山实验室 End-to-end deterministic transmission control method, device, equipment and medium

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105553845A (en) * 2016-01-19 2016-05-04 重庆邮电大学 Software defined network route selection method
US20160191306A1 (en) * 2014-12-27 2016-06-30 Iosif Gasparakis Programmable protocol parser for nic classification and queue assignments
CN106330555A (en) * 2016-08-29 2017-01-11 南京市产品质量监督检验院 OpenFlow switch performance parameter measurement method based on KPLS algorithm
CN106571883A (en) * 2016-07-04 2017-04-19 长春理工大学 Random network calculation method for satellite network performance evaluation
CN109714275A (en) * 2019-01-04 2019-05-03 电子科技大学 A kind of SDN controller and its control method for access service transmission
CN112491619A (en) * 2020-11-25 2021-03-12 东北大学 Self-adaptive distribution technology for service customized network resources based on SDN
CN112671673A (en) * 2020-12-28 2021-04-16 广州西麦科技股份有限公司 SDN-based flow control system and method

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160191306A1 (en) * 2014-12-27 2016-06-30 Iosif Gasparakis Programmable protocol parser for nic classification and queue assignments
CN105553845A (en) * 2016-01-19 2016-05-04 重庆邮电大学 Software defined network route selection method
CN106571883A (en) * 2016-07-04 2017-04-19 长春理工大学 Random network calculation method for satellite network performance evaluation
CN106330555A (en) * 2016-08-29 2017-01-11 南京市产品质量监督检验院 OpenFlow switch performance parameter measurement method based on KPLS algorithm
CN109714275A (en) * 2019-01-04 2019-05-03 电子科技大学 A kind of SDN controller and its control method for access service transmission
CN112491619A (en) * 2020-11-25 2021-03-12 东北大学 Self-adaptive distribution technology for service customized network resources based on SDN
CN112671673A (en) * 2020-12-28 2021-04-16 广州西麦科技股份有限公司 SDN-based flow control system and method

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
JANFIZZA BUKHARI等: "Simulated view of SDN based multicasting over D2D enabled heterogeneous cellular networks", pages 1 - 4, Retrieved from the Internet <URL:《网页在线公开：https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7979908》> *
熊兵等: "面向软件定义核心网的OpenFlow分组转发优先制排队模型研究", 《电子学报》, vol. 47, no. 10, pages 1969 - 1982 *
鲁垚光等: "软件定义网络中的动态负载均衡与节能机制", 《计算机学报》, vol. 43, no. 10, pages 1969 - 1982 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114244680A (en) * 2021-12-17 2022-03-25 网络通信与安全紫金山实验室 End-to-end deterministic transmission control method, device, equipment and medium
CN114244680B (en) * 2021-12-17 2023-06-27 网络通信与安全紫金山实验室 End-to-end deterministic transmission control method, device, equipment and medium

Also Published As

Publication number Publication date
CN113742902B (en) 2023-08-04

Similar Documents

Publication Publication Date Title
Stiliadis et al. Design and analysis of frame-based fair queueing: A new traffic scheduling algorithm for packet-switched networks
Scharbarg et al. A probabilistic analysis of end-to-end delays on an AFDX avionic network
Bauer et al. Applying and optimizing trajectory approach for performance evaluation of AFDX avionics network
CN112532409A (en) Network parameter configuration method and device, computer equipment and storage medium
CN108847969B (en) Network service reliability analysis method based on information flow
CN110708259A (en) Information-agnostic Coflow scheduling system capable of automatically adjusting queue threshold and scheduling method thereof
CN110290077B (en) Industrial SDN resource allocation method based on real-time service configuration
CN113742902B (en) Multi-parameter performance modeling evaluation method based on network algorithm
US7469203B2 (en) Wireless network hybrid simulation
CN112202644B (en) Collaborative network measurement method and system oriented to hybrid programmable network environment
CN104734907A (en) Method for actively measuring end-to-end path performance of OpenFlow network and system adopted by the same
CN109714795A (en) A kind of method for managing resource, resource management system and device based on SDN network slice
CN112995036A (en) Network traffic scheduling method and device
CN117082008B (en) Virtual elastic network data transmission scheduling method, computer device and storage medium
CN111416779B (en) Internet service queue scheduling method based on time limit
EP1842332B1 (en) A method and apparatus for calculating bandwidth requirements
Gao et al. Modeling and parameter optimization of statistical priority-based multiple access protocol
JP2001285344A (en) Method for control of accepting connection
CN115811489A (en) Exchange chip verification system and method based on UVM and storage medium
Li et al. A data forwarding mechanism based on deep reinforcement learning for deterministic networks
CN107743077A (en) A kind of method and device for assessing information physical emerging system network performance
CN115378881A (en) Federal learning-based home router data flow identification method and identification framework
Liu et al. A stochastic network calculus
Siddikov et al. Hybrid Model management of Information messages in multiservice networks
EP4131869A1 (en) Communication method, apparatus, and system

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