CN110784405B - Robust energy efficiency routing method for teaching cloud computing network - Google Patents
Robust energy efficiency routing method for teaching cloud computing network Download PDFInfo
- Publication number
- CN110784405B CN110784405B CN201910982647.1A CN201910982647A CN110784405B CN 110784405 B CN110784405 B CN 110784405B CN 201910982647 A CN201910982647 A CN 201910982647A CN 110784405 B CN110784405 B CN 110784405B
- Authority
- CN
- China
- Prior art keywords
- link
- network
- criticality
- node
- equation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 66
- 230000007958 sleep Effects 0.000 claims abstract description 6
- 239000011159 matrix material Substances 0.000 claims description 50
- 238000005265 energy consumption Methods 0.000 claims description 38
- 238000004088 simulation Methods 0.000 claims description 15
- 238000010586 diagram Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 230000005059 dormancy Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000007423 decrease Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 101001120757 Streptococcus pyogenes serotype M49 (strain NZ131) Oleate hydratase Proteins 0.000 abstract description 3
- 229940083712 aldosterone antagonist Drugs 0.000 abstract description 3
- 238000004891 communication Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 6
- 238000009826 distribution Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/12—Shortest path evaluation
- H04L45/123—Evaluation of link metrics
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/14—Routing performance; Theoretical aspects
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/48—Routing tree calculation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a robust energy efficiency routing method for a teaching cloud computing network, and belongs to the field of green communication of cloud computing networks. The present invention applies SRLA to find the largest set of sleepable links and converts redundant links to a sleep state as much as possible to achieve an energy efficient network. The REERA can meet the network robustness requirement under the condition of ensuring normal network routing. The MCRA routing algorithm represents the stability of the whole network through network criticality, and finally, the dormant redundant link algorithm is combined with the minimum criticality routing algorithm to construct a robust, efficient and energy-saving route of the cloud computing network, so that the problem of operation of a cloud computing backbone network with low energy efficiency is solved. More importantly, the method can dynamically change and adjust the link weight, ensures that the links can be used uniformly to the maximum extent, avoids flow congestion when some links of the cloud computing network are used excessively, and can improve the performance of the whole network.
Description
Technical Field
The invention belongs to the field of green communication of cloud computing networks, and particularly relates to a robust energy efficiency routing method of a cloud computing network.
Background
Cloud computing services provide a novel collaborative and personalized learning style that has significant impact on learning and teaching due to its wider application offerings than other technologies. However, the tremendous growth in the number of network devices and service types makes it difficult to reduce energy consumption, especially in highly loaded cloud computing networks. In addition, the cost pressure of Internet Service Providers (ISPs) is increasing due to the rising price of energy, and the problem of energy consumption is urgently needed to be solved. Therefore, researchers have developed concepts of energy efficiency, aiming to improve the energy efficiency of the network, and to reduce the carbon footprint and energy consumption of the network. How to provide highly reliable and energy-efficient cloud computing services for learning and teaching is a major challenge in the industry and academia.
In general, a typical ISP network is based on an IP-over-WDM architecture; to meet Quality of Service (QoS) requirements, ISP network links typically use a redundant design to respond to network conditions of unexpected congestion and routing failures. The design scheme solves the QoS problem, but also causes the problem of low energy efficiency of the network, and how to design a routing algorithm which can meet the QoS requirement and realize a robust energy-saving network is the main purpose of the invention.
If the method of making the idle router and the link sleep, the utilization rate of network resources and network performance is low, so that the method is not suitable for the cloud computing network. The algorithm utilizes srla (sleeping Redundant Links algorithm) to make Redundant Links in the cloud computing backbone network sleep, and mcra (minimum critical Routing algorithm) to improve the robustness of the entire cloud computing network. The algorithm can dynamically change and adjust the link weight, ensures that the links can be used uniformly to the maximum extent, and avoids flow congestion when some links of the cloud computing network are excessively used.
Disclosure of Invention
Aiming at the defects of the existing method, the invention provides a Robust Energy Efficiency Routing method (REERA) for a teaching cloud computing network, so as to realize teaching and scientific research of a high-Energy Efficiency network for cloud computing service and simultaneously maintain the stability and robustness of the whole network.
The technical scheme of the invention is realized as follows: a robust energy efficiency routing method for a teaching cloud computing network comprises the following steps:
the method comprises the following steps: calculating link utilization
Establishing a link model and calculating the link utilization rate;
step two: dormancy of redundant links using SRLA algorithm
In the SRLA algorithm, a Dijkstra algorithm is used for calculating a shortest path tree of each router node, then a dispatch mechanism and an entry mechanism obtain a modified path tree, and a redundant link is calculated by comparing an SPT (shortest Path first) and an MPT (Multi-Path first) and is dormant;
step three: establishing REERA model
Establishing a REERA model by using the link weight as the input routing flow of the Dijkstra algorithm;
step four: determining SRLA algorithm parameters
Determining parameters, iteration orders and times which are suitable for an SRLA algorithm, triggering the frequency of the SRLA, and time interval T;
step five: calculating link criticality
Calculating link criticality based on the link capacity matrix;
step six: minimizing network criticality
Mapping the value of the link criticality to the link weight, minimizing the network criticality and realizing the most robust network;
step seven: obtaining an optimal routing path by the network link weight matrix;
the appropriate routing path is selected by the network link weight matrix using Dijkstra's algorithm.
Further, the first step calculates the link utilization, and the specific process is as follows:
the cloud computing backbone network is modeled as G (V, E, W), where V is a set of nodes, E is a set of links, and W represents a link weight matrix of the network. In the model, the link states are defined as follows:
wherein x isij∈{0,1},LijE represents a link from node i to node j;
the link energy consumption comprises a base energy consumptionAnd energy consumption related to flowThe basic energy consumption means that the constant energy consumption of the operation link is guaranteed to be expressed as:
wherein,represents a link LijBasic energy consumption of cijRepresenting the link capacity, δ being a basic energy related parameter, β representing the ratio of the link's energy consumption to its capacity, and the traffic related energy consumption is:
wherein f isij(t) denotes a pass-through link LijThe flow rate of (a);representing the relationship between link energy consumption and the traffic energy consumption carried by it;
link LijThe energy consumption of (a) is modeled as:
only the dormancy of redundant links of the network is considered in the network and does not involve the shutdown of network nodes. Thus, the energy efficiency of the entire networkCan be expressed as:
wherein,NsRepresenting a source node, NdRepresenting a destination node, flsd(t) denotes a slave source node NsE.g. V to destination node NdInstantaneous flow, x, over e VijIndicating the state of the link, NiRepresenting a node, p0Representing the power of nodes in the network, set to a constant value,representing a node NsTo node NdAverage flow rate of LijRepresents a link from node i to node j, V represents a set of nodes, and E represents a set of links;
Lijthe link utilization of (a) is:
wherein,indicating that the link is from LijAverage flow rate of above; c. CijRepresents the link capacity;
after T time interval, uijFor link utilization as:
wherein,indicating the average utilization of the initial link upon user request,representing the average utilization of the link.
Further, the specific method of the second step is as follows:
modeling a backbone network of cloud computing as G (V, E, W), wherein V is a node set, E is a link set, and N and L are sets respectivelyThe cardinality of V and E, W represents the link weight matrix of the network; calculating a shortest path tree SPT (v) of each router node by a Dijkstra algorithm, wherein v represents the router nodes in the network; set of active links EADefined as the set of links E involved in the shortest spanning treeA=Uv∈VSPT(v),Uv∈VRepresents a complete subgraph of diagram G; after a shortest path tree SPT (v) of each router node in a network model of cloud computing is obtained, a degree-out mechanism and a degree-in mechanism are utilized to carry out loop iteration to obtain a modified path tree, wherein the degree-out mechanism is to find a degree-out router node ER with the maximum degree in the network; routing router node IR among the ER's neighbors, computing the shortest path trees of ER and IR, comparing the shortest path trees of ER and IR to generate a modified path tree; and finally, comparing the MPT with the SPT to obtain a redundant link and converting the redundant link into a dormant state, and repeating the steps to realize the SRLA algorithm.
Further, the establishing of the REERA model in the third step:
s.t.
connection(Ns,Nd)=1 (9b)
equation (9a) represents a cost function that maximizes the energy efficiency of the network; equation (9b) shows that any traction node is always connected and the traffic between any two nodes is non-negative; equation (9c) indicates that the link weight is used as the link cost when constructing the route; equation (9d) means link LijIs equal to the load across link LijThe sum of the flow rates of all OD (origin destination) streams; equation (9e) indicates that the link utilization is not greater than the link capacity; costlijRepresents a link lijOverhead of traffic of wijA link weight value indicating the end of the request,representing (s, d) passing through the link lijFlow rate of (c)ijRepresents a link lijThe capacity of (c).
Further, the fourth step of determining parameters of the SRLA algorithm includes the following specific processes:
in order to maximize network energy efficiency, the number of iterations is set to 3, then the frequency of triggering the SRLA is determined, the time interval is set to T, the effect of the time interval T can be offset in the energy efficiency calculation process, and most user requests can be responded within a time interval.
Further, the link criticality is calculated in the fifth step, and the specific process is as follows:
the smaller the criticality of the network is, the stronger the robustness of the network is; therefore, the criticality τ of the network is:
τ=2nTr(L+) (10)
where n is the number of network nodes, Tr (-) represents the matrix trajectory,is the Moore-Penrose inverse of Laplace matrix L; in equation (6), the criticality τ of the network represents the sum of the resistive distances between all two nodes;
defining a linear relationship between link capacity and initial link weight value, i.e. cij=k·woij+ a, where the parameters k and a are constants, woijRepresenting an initial value of link weight, and calculating link criticality based on a link capacity matrix; the normalized form of the network criticality τ is expressed as:
the formula for calculating the criticality of the network is as follows:
d=diag(c1,c2,…cn) (12b)
Lc=d-W (12c)
wherein, A (r) represents the set of active nodes, and W represents the link weight matrix; equation (12a) represents the degree of fanout, c, of node rrjRepresents the link capacity from node r to node j; equation (12b) represents a matrix constructed from diagonal vectors of the link capacity matrix; equation (12c) represents a new matrix consisting of a diagonal vector matrix and a weight matrix of link capacities; equation (12d) represents how to compute the normalized network criticality.
Further, the network criticality is minimized in the sixth step, and the specific process is as follows:
due to the relationship between the network criticality and robustness, the objective equation for implementing the most robust network is:
Minτ (13a)
s.t.
∑(i,j)∈Lwij≤cij,cij is fixed (13b)
0≤wij≤cij (13c)
wherein, cijIs a link capacity matrix; w is aijIs a link LijWeights, which can obtain reasonable network link weight values by optimizing the routing algorithm, thereby improving the robustness of the whole network, equation (13a) represents an objective function, and equations (13b) and (13c) represent limits;
equation (10) indicates that the criticality τ of the network is a network-wide measure and can be used to describe network robustness, the smaller the criticality τ of the network, the more robust the network is; equation (13a) represents an objective function that minimizes the criticality τ of the network, the second and third equations showing the constraints satisfied by the weight of each link; equation (13) represents that by minimizing the criticality τ of the network, the optimal link weight satisfying a plurality of constraints is found, and a robust route is constructed;
equation (12) indicates that when the bandwidth capacity of a link changes, the criticality of the link also changes, and for a high-load link, the influence on the whole network in the routing process is large; thus, the value of link criticality is mapped to the link weight, which is updated as the bandwidth capacity of the network link decreases by the following formula:
τ=n(n-1)τ′ (14a)
Δτ=τ-τ0 (14b)
wij=b·Δτ+w0ij (14c)
an initial network criticality value for a time interval T; τ is the network criticality value at the end of time T, τ0Representing an initial network criticality within a particular time interval T; b is a constant, which is based on a simulation of a real network; n is the number of network nodes; w is a0ijIs the initial link weight value, w, of time TijIs the link weight value at the end of time interval T; equation (14a) represents the relationship between the criticality of the network and its normalized value; equation (14b) represents a change in the criticality value of the network; equation (14c) characterizes the relationship between the change in the criticality of the network and the link weight value.
The symbols involved in the present invention are defined as follows:
n denotes the collective V cardinal number (where N | | V |), L denotes the cardinal number of the collective E (where L | | | E |), Pij(t) denotes a link LijInstantaneous power, fij(t) representsLink LijInstantaneous flow of (fl)sd(t) denotes a source node NsE.v and destination node NdE.g. instantaneous flow between V, PijIndicating the link L within the period TijAverage power of fijIndicating the link L within the period TijAverage flow of, flsdRepresents N within the period TsAnd NdIs a constant, and po represents the power of the node in the network, cijRepresents a link LijThe capacity of the electric power transmission device is,represents a link LijBeta represents the link LijRatio of energy consumption to capacity, delta representing a parameter related to the base energy, xijIndicating the link state, ηijRepresents a link LijThe average energy efficiency over a period T, μ represents the change matrix of link utilization after a user request T period, ωijIndicating the link L at the end of the time interval TijWeight value, woijRepresents the initial link LijWeight values, A (k) represents the active node set, W represents the link weight matrix, τ0An initial network criticality value representing the time interval T, τ representing the network criticality value at the end of time T, b representing a constant, which is based on a simulation of a real network, and n representing the number of network nodes.
The invention has the beneficial effects that:
the method provided by the invention can meet the QoS requirement, realize a robust energy-saving network and provide highly reliable and energy-efficient cloud computing service for learning and teaching. The present invention applies SRLA to find the largest set of sleepable links and converts redundant links to a sleep state as much as possible to achieve an energy efficient network. The REERA can meet the network robustness requirement under the condition of ensuring normal network routing. The MCRA routing algorithm represents the stability of the whole network through network criticality, and finally, the dormant redundant link algorithm is combined with the minimum criticality routing algorithm to construct a robust, efficient and energy-saving route of the cloud computing network, so that the problem of operation of a cloud computing backbone network with low energy efficiency is solved. More importantly, the method can dynamically change and adjust the link weight, ensures that the links can be used uniformly to the maximum extent, avoids flow congestion when some links of the cloud computing network are used excessively, and can improve the performance of the whole network. Simulation analysis shows that the proposed REERA can save a large amount of energy consumption, and is far superior to the existing network routing algorithm based on OSPF (open Shortest Path first).
Drawings
FIG. 1 is a process flow diagram of the robust energy efficient routing method of the present invention;
FIG. 2 is a model of a cloud computing network connecting different data centers and accessing users;
FIG. 3 is a schematic diagram of a cloud computing backbone robust energy efficient routing algorithm;
fig. 4 is a variation process of a second order SRLA network topology;
FIG. 5 shows a network topology of data used in the simulation process;
fig. 6 is a histogram routing algorithm for link utilization for both. The horizontal axis represents the distribution of link utilization. The vertical axis represents the probability of link utilization. The blue bar corresponds to the proposed Algorithm and the red bar corresponds to the OSPFERA (OSPF-based Energy-influencing Routing Algorithm) Algorithm;
fig. 7 shows the distribution of the number of times of failure of user requests due to insufficient bandwidth capacity of the link, in which the horizontal axis represents the time interval and the vertical axis represents the number of user beg requests. The blue bar corresponds to the proposed REERA algorithm; red bar versus ospfara algorithm;
FIG. 8 is a graph of utilization profiles for each link in a network, where the x-axis represents time interval and the y-axis represents link utilization. The curve describes the usage of each link and the row is the average usage of both algorithms. The red curve represents the simulation of REERA. The blue curve represents a simulation of OSPFERA;
FIG. 9 is a distribution of sleep link counts, where the red line corresponds to REERA and the blue line corresponds to OSPFEA;
in fig. 10, the method runs 100 simulations for the same user request to calculate the average run time of both. Where the x-axis represents the number of user requests and the y-axis represents the average time in ms.
Detailed Description
The invention is described in further detail below with reference to the following figures and specific examples:
a robust energy-saving routing method for cloud computing network learning comprises the following specific steps:
step one, calculating the link utilization rate, specifically comprising the following steps:
the cloud computing backbone network is modeled as G (V, E, W), where V is a set of nodes, E is a set of links, and W represents a link weight matrix of the network. In the model, the link states are defined as follows:
wherein x isij∈{0,1},LijE represents the link from node i to node j.
The link energy consumption comprises a base energy consumptionAnd energy consumption related to flowThe basic energy consumption refers to the constant energy consumption of the guaranteed operation link, and is expressed as:
in the formula,represents a link LijBasic energy consumption of cijRepresenting the link capacity, δ being a basic energy related parameter, β representing the ratio of the link's energy consumption to its capacity, and the traffic related energy consumption is:
wherein f isij(t) denotes a pass-through link LijThe flow rate of (a);representing the relationship between the link energy consumption and the traffic energy consumption carried by it.
Link LijThe energy consumption of (a) is modeled as:
where B represents the base energy and R represents the flow related energy.
Therefore, the link L in the T periodijThe average energy efficiency of (d) can be expressed as:
only dormant network redundant links are considered herein, but not the shutdown node. The energy efficiency of the entire network can be expressed as:
wherein N issRepresenting a source node, NdRepresenting a destination node, flsd(t) denotes a slave source node NsE.g. V to destination node NdInstantaneous flow, x, over e VijIndicating the state of the link, NiRepresenting a node, p0Representing the power of nodes in the network, set to a constant value,representing a node NsTo node NdAverage flow rate of LijRepresenting a link from node i to node j, V representing a nodeThe set of points, E, represents the set of links.
LijThe link utilization of (a) is defined as:
wherein,indicating that the link is from LijAverage flow rate of above; c. CijIndicating link capacity
After T cycles requested by the user, u is set as a variable matrix of link utilization, and the mathematical formula is as follows:
wherein,andrespectively representing the average utilization rate of the initial link and the average utilization rate of the link after the user requests.
Step two: computing dormant redundant links using SRLA algorithm
Modeling a backbone network of cloud computing as G (V, E, W), wherein V is a node set, E is a link set, N and L are cardinality of the sets V and E respectively, and W represents a link weight matrix of the network. Calculating a shortest path tree SPT (v) of each router node by a Dijkstra algorithm, wherein v represents the router nodes in the network; set of active links EADefined as the set of links E involved in the shortest spanning treeA=Uv∈VSPT(v),Uv∈VRepresenting a complete sub-diagram of diagram G. After a shortest path tree SPT (v) of each router node in the network model of cloud computing is obtained, a degree-out mechanism and a degree-in mechanism are utilized to carry out loop iteration to obtain a Modified Path Tree (MPT), wherein,the basic idea of the out-degree mechanism is to find out the out-degree router node (ER) with the maximum degree in the network; an ingress Router node (IR) among the ER's neighbors computes a shortest path tree of ER and IR, which compares spt (IR) with spt (ER) to generate a Modified Path Tree (MPT). And finally, comparing the MPT with the SPT to obtain a redundant link and converting the redundant link into a dormant state, and repeating the steps to realize the SRLA algorithm.
Establishing a REERA model in the third step, which comprises the following specific processes:
the REERA can be seen as an integrated algorithm of SRLA and MCRA, according to a model of a robust energy routing algorithm in the backbone network. Through the above analysis, a REERA optimization model is established.
REERA can be expressed as the following mathematical expression:
s.t.
connection(Ns,Nd)=1 (9b)
equation (9a) represents a cost function that maximizes the energy efficiency of the network. Equation (9b) shows that any traction node is always connected and that the traffic between any two nodes is non-negative. Equation (9c) Indicating that the link weight is used as the link cost when constructing the route. Equation (9d) means link LijIs equal to the load across link LijThe sum of all the OD flows. Equation (9e) indicates that the link utilization is not greater than the link capacity.
Step four, determining the parameters of the SRLA algorithm, which comprises the following specific processes:
to maximize network energy efficiency, the order of the SRLA iteration needs to be determined. Based on the analysis, the number of iterations is set to 3, then the frequency of triggering the SRLA is determined, and the time interval is set to T. The effect of the time interval T can be offset in the energy efficiency calculation process. However, it must be determined that most users' requests can be responded to within a time interval.
Step five, calculating the link criticality, wherein the specific process is as follows:
the less critical the network, the more robust the network is. At the beginning of the initial network link capacity, the value of network criticality is minimal. The robustness of a communication network is reflected using the chain criticality, which is defined directly as:
where n is the number of network nodes, Tr (-) represents the matrix trajectory,is the inverse Moore-Penrose transform of the laplacian matrix L. In equation (10), the criticality τ of the network represents the sum of the resistive distances between all two nodes, which is a graph-theory based global metric.
Now define a linear relationship between link capacity and initial link weight value, i.e. cij=k·woij+ a (where the parameters k and a are constants, woijRepresenting link weight initial values) based on the link capacity matrix. Normalizing the criticality τ of the network to produce a variable, i.e.
The formula for calculating the criticality of the network is as follows:
d=diag(c1,c2,…cn) (12b)
Lc=d-W (12c)
wherein, A (r) represents the set of active nodes, and W represents the link weight matrix; equation (12a) represents the degree of fanout, c, of node rrjRepresents the link capacity from node r to node j; equation (12b) represents a matrix constructed from diagonal vectors of the link capacity matrix; equation (12c) represents a new matrix consisting of a diagonal vector matrix and a weight matrix of link capacities; equation (12d) represents how to compute the normalized network criticality.
And step six, minimizing the network criticality, which comprises the following specific processes:
due to the relationship between the network criticality and the robustness, the network criticality is minimized, and the network with the strongest robustness is realized, wherein the target equation is as follows:
Minτ (13a)
s.t.
∑(i,j)∈Lwij≤cij,cij is fixed (13b)
0≤wij≤cij (13c)
wherein, cijIs a link capacity matrix; w is aijIs a link LijAnd (4) weighting. Reasonable network link weight values can be obtained through the optimized routing algorithm, and therefore the robustness of the whole network is improved. Equation (13a) represents the objective function, while equations (13b) and (13c) represent the constraints.
Equation (10) indicates that the criticality τ of the network is a network-wide metric and can be used to describe network robustness. The smaller the criticality τ of the network, the more robust the network is. Equation (13a) represents an objective function that minimizes the criticality τ of the network; the second and third equations show the constraints that are satisfied by the weight of each link. Equation (13) indicates that finding optimal link weights that satisfy multiple constraints can help build robust routes by minimizing the criticality τ of the network.
Equation (12) indicates that as the bandwidth capacity of a link changes, the criticality of the link also changes. For a highly loaded link, the available bandwidth of the link capacity is relatively low, and the link has a large impact on the entire network during routing. Thus, the method can map the value of link criticality to link weights, which will vary with τ as the bandwidth capacity of the network link decreases. Updating the weight values of the links by the following formula:
τ=n(n-1)τ′ (14a)
Δτ=τ-τ0 (14b)
wij=b·Δτ+w0ij (14c)
wherein, a certain time interval T is a criticality value of the initial network; τ is the network criticality value at the end of time interval T; tau is0Representing an initial network criticality within a particular time interval T; b is a constant, which is based on a simulation of a real network; n is the number of network nodes; w is a0ijIs the initial link weight value, w, of time TijIs the link weight value at the end of time interval T.
Equation (14a) represents the relationship between the criticality of the network and its normalized value; equation (14b) represents a change in the criticality value of the network; equation (14c) characterizes the relationship between the change in the criticality of the network and the link weight value.
And seventhly, obtaining the optimal routing path by the network link weight matrix, wherein the specific process is as follows:
the appropriate routing path is selected by the network link weight matrix using the Dijkstra algorithm, which has been described in step one.
Examples
In the simulation of the present invention, the basic power consumption of each router was set to 10W/h. It is divided into each interface (link) according to the simulated topology. FIG. 5 shows a network topology of data used in the simulation process; there are 23 nodes and 74 links in the graph, that is, the positions of the nodes and the lengths of the links do not represent the specific meaning of the network, and there is a direct relationship between the traffic size and the link weight. In order to analyze the superior robustness of the proposed energy-efficient routing algorithm, REERA is compared with ospfara, and the specific implementation steps are as follows:
the method comprises the following steps: calculating link utilization
Establishing a link model and calculating the link utilization rate; FIG. 6 is a histogram routing algorithm for both link utilization, showing that the algorithm's link utilization is better in the range of 2-5, showing that the algorithm's routing structure can keep the link utilization between 20% and 50%; link utilization is neither too high, resulting in network congestion and link failure, nor too low resulting in considerable waste of resources.
Step two: dormancy of redundant links using SRLA algorithm
In the SRLA algorithm, a Dijkstra algorithm is used for calculating a Shortest Path Tree (SPT) of each router node, then an outbound mechanism and an inbound mechanism obtain a Modified Path Tree (MPT), and a redundant link is calculated and is dormant by comparing the SPT and the MPT;
step three: establishing REERA model
Using the link weight as the input routing flow of Dijkstra algorithm, and establishing a REERA model through analysis;
step four: determining parameters for SRLA algorithms
Determining parameters, iteration orders and times which are suitable for an SRLA algorithm, triggering the frequency of the SRLA, and time interval T;
step five: calculating link criticality
Calculating link criticality based on the link capacity matrix;
step six: minimizing network criticality
Mapping the value of the link criticality to the link weight, minimizing the network criticality and realizing the most robust network; fig. 9 shows the distribution of the number of dormant links, and the link weight of the ospfara algorithm remains unchanged after initialization, so that the network cannot dynamically change the distribution of the routes according to the state of the user request, and cannot realize load balancing, so that the robustness of the network is relatively weak. As shown in fig. 9, the REERA may shut down and wake up the link in a time-varying manner. Simulation experiments show that the REERA proposed herein is effective, not only to improve the energy efficiency of the network, but also to improve the robustness of the network.
Step seven: obtaining optimal routing path from network link weight matrix
Selecting a proper routing path by the network link weight matrix by using Dijkstra algorithm; further verifying the performance of both algorithms, run times for different numbers of user requirements were analyzed, as shown in fig. 10, where the method runs 100 simulations for the same user request to calculate the average run time of both. From the method of fig. 10, it can be seen that REERA and OSPFRA remain low on-time. The method uses a desktop computer with 2 cores 3.2ghz cpu and 16G memory for simulation. Less runtime is obtained if a special hardware platform is used. This indicates that rera is actually feasible, and can dynamically change the link weights, ensure that most links are uniformly used, and avoid traffic congestion when some links of the cloud computing network are excessively used, so that the performance of the whole network is improved. By comparing REERA with an Open Shortest Path First (OSPF) algorithm, simulation results show that REERA is superior to the OSPF algorithm, and REERA can realize an energy-saving routing algorithm under the premise of network robustness constraint.
Claims (2)
1. A robust energy efficiency routing method for a teaching cloud computing network comprises the following steps:
the method comprises the following steps: calculating link utilization
Establishing a link model and calculating the link utilization rate;
step two: dormancy of redundant links using SRLA algorithm
In the SRLA algorithm, a Dijkstra algorithm is used for calculating a shortest path tree of each router node, then a dispatching mechanism and an entering mechanism are used for obtaining a modified path tree, and a Shortest Path Tree (SPT) and a Modified Path Tree (MPT) are compared to calculate a redundant link and sleep is carried out;
step three: establishing REERA model
Establishing a REERA model by using the link weight as the input routing flow of the Dijkstra algorithm;
step four: determining SRLA algorithm parameters
Determining parameters of the SRLA algorithm: iteration order, iteration times, frequency of triggering SRLA, and time interval T;
step five: calculating link criticality
Calculating link criticality based on the link capacity matrix;
step six: minimizing network criticality
Mapping the value of the link criticality to the link weight, minimizing the network criticality and realizing the most robust network;
step seven: obtaining an optimal routing path by the network link weight matrix;
selecting a proper routing path by the network link weight matrix by using Dijkstra algorithm;
the first step is to calculate the link utilization rate, and the specific process is as follows:
modeling a cloud computing backbone network as G (V, E, W), wherein V is a node set, E is a link set, and W represents a link weight matrix of the network; in the model, the link states are defined as follows:
wherein x isij∈{0,1},LijE represents a link from node i to node j;
the link energy consumption comprises a base energy consumptionAnd energy consumption related to flowThe basic energy consumption means that the constant energy consumption of the operation link is guaranteed to be expressed as:
wherein,represents a link LijBasic energy consumption of cijRepresenting the link capacity, δ being a basic energy related parameter, β representing the ratio of the link's energy consumption to its capacity, and the traffic related energy consumption is:
wherein f isij(t) denotes a pass-through link LijThe flow rate of (a);representing the relationship between link energy consumption and the traffic energy consumption carried by it;
link LijThe energy consumption of (a) is modeled as:
only the dormancy of redundant links of the network is considered in the network, and the shutdown of the network nodes is not involved; thus, the energy efficiency of the entire networkCan be expressed as:
wherein N issRepresenting a source node, NdRepresenting a destination node, flsd(t) denotes a slave source node NsE.g. V to destination node NdInstantaneous flow, x, over e VijIndicating the state of the link, NiRepresenting a node, p0Representing the power of nodes in the network, set to a constant value,representing a node NsTo node NdAverage flow rate of LijRepresents a link from node i to node j, V represents a set of nodes, and E represents a set of links;
Lijthe link utilization of (a) is:
wherein,indicating that the link is from LijAverage flow rate of above; c. CijRepresents the link capacity;
after T time interval, uijFor link utilization as:
wherein,indicating the average utilization of the initial link upon user request,representing the average utilization of the link;
the specific method of the second step is as follows:
modeling a backbone network of cloud computing as G (V, E, W), wherein V is a node set, E is a link set, N and L are cardinality of the sets V and E respectively, and W represents a link weight matrix of the network; calculating a shortest path tree SPT (v) of each router node by a Dijkstra algorithm, wherein v represents the router nodes in the network; set of active links EADefined as the set of links E involved in the shortest spanning treeA=Uv∈VSPT(v),Uv∈VRepresents a complete subgraph of diagram G; after a shortest path tree SPT (v) of each router node in a network model of cloud computing is obtained, a degree-out mechanism and a degree-in mechanism are utilized to carry out loop iteration to obtain a modified path tree, wherein the degree-out mechanism is to find a degree-out router node ER with the maximum degree in the network; routing router node IR among the ER's neighbors, computing the shortest path trees of ER and IR, comparing the shortest path trees of ER and IR to generate a modified path tree; finally, comparing the MPT with the SPT to obtain a redundant link and converting the redundant link into a dormant state, and repeating the steps to realize the SRLA algorithm;
establishing a REERA model in the third step:
s.t.
connection(Ns,Nd)=1 (9b)
equation (9a) represents a cost function that maximizes the energy efficiency of the network; equation (9b) shows that any tow node is always connected and the traffic between any two nodes is non-negative; equation (9c) represents the use of link weights as link costs in constructing routes; equation (9d) means that link LijIs equal to the load across link LijThe sum of the flow rates of all Origin Destination streams; equation (9e) indicates that link utilization is not greater than link capacity; costLijRepresents a link LijOverhead of traffic of wijA link weight value indicating the end of the request,indicating that (s, d) passes through the link LijFlow rate of (c)ijRepresents a link LijThe capacity of (a);
calculating the criticality of the link in the step five, wherein the specific process is as follows:
the smaller the criticality of the network is, the stronger the robustness of the network is; therefore, the criticality τ of the network is:
τ=2nTr(L+) (10)
where n is the number of network nodes, Tr (-) represents the matrix trajectory,is the Moore-Penrose inverse of Laplace matrix L; in equation (6), the criticality τ of the network represents the sum of the resistive distances between all two nodes;
defining a linear relationship between link capacity and initial link weight value, i.e. cij=k·woij+ a, where the parameters k and a are constants, woijRepresenting an initial value of link weight, and calculating link criticality based on a link capacity matrix;
the formula for calculating the criticality of the network is as follows:
d=diag(c1,c2,…cn) (12b)
Lc=d-W (12c)
wherein, A (r) represents the set of active nodes, and W represents the link weight matrix; equation (12a) represents the degree of fanout, c, of node rrjRepresents the link capacity from node r to node j; equation (12b) represents a matrix constructed from diagonal vectors of the link capacity matrix; equation (12c) represents a new matrix consisting of a diagonal vector matrix and a weight matrix of link capacities; equation (12d) represents how to calculate the normalized network criticality;
in the sixth step, the network criticality is minimized, and the specific process is as follows:
due to the relationship between the network criticality and robustness, the objective equation for implementing the most robust network is:
Minτ (13a)
s.t.
∑(i,j)∈Lwij≤cij,cij is fixed (13b)
0≤wij≤cij (13c)
wherein, cijIs a link capacity matrix; w is aijIs a link LijWeighting, wherein a reasonable network link weight value can be obtained by optimizing a routing algorithm, so that the robustness of the whole network is improved, a formula (13a) represents an objective function, and formulas (13b) and (13c) represent limitations;
equation (10) shows that the criticality τ of the network is a network-wide measure and can be used to describe the robustness of the network, the smaller the criticality τ of the network, the more robust the network is; equation (13a) represents an objective function that minimizes the criticality τ of the network, the second and third equations showing the constraints satisfied by the weight of each link; formulas (13a) to (13c) show that the optimal link weight meeting a plurality of constraints is found by minimizing the criticality tau of the network, and a robust route is constructed;
formulas (12a) to (12d) show that when the bandwidth capacity of a link changes, the criticality of the link also changes, and for a high-load link, the influence on the whole network in the routing process is large; thus, the value of link criticality is mapped to the link weight, which is updated as the bandwidth capacity of the network link decreases by the following formula:
τ=n(n-1)τ′ (14a)
Δτ=τ-τ0 (14b)
wij=b·Δτ+w0ij (14c)
τ is the network criticality value at the end of the time interval T, τ0Representing an initial network criticality within a particular time interval T; b is a constant, which is based on a simulation of a real network; n is the number of network nodes; w is a0ijIs the initial link weight value, w, of the time interval TijIs the link weight value at the end of time interval T; equation (14a) represents the relationship between the criticality of the network and its normalized value; equation (14b) represents the change in the criticality value of the network; equation (14c) characterizes the relationship between the change in network criticality and the link weight value.
2. The robust energy efficiency routing method oriented to the teaching cloud computing network according to claim 1, wherein the fourth step determines parameters of the SRLA algorithm, and the specific process is as follows:
in order to maximize network energy efficiency, the number of iterations is set to 3, then the frequency of triggering the SRLA is determined, the time interval is set to T, the effect of the time interval T can be offset in the energy efficiency calculation process, and most user requests can be responded within a time interval.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910982647.1A CN110784405B (en) | 2019-10-16 | 2019-10-16 | Robust energy efficiency routing method for teaching cloud computing network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910982647.1A CN110784405B (en) | 2019-10-16 | 2019-10-16 | Robust energy efficiency routing method for teaching cloud computing network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110784405A CN110784405A (en) | 2020-02-11 |
CN110784405B true CN110784405B (en) | 2021-11-02 |
Family
ID=69385544
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910982647.1A Active CN110784405B (en) | 2019-10-16 | 2019-10-16 | Robust energy efficiency routing method for teaching cloud computing network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110784405B (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7593341B1 (en) * | 2006-06-08 | 2009-09-22 | At&T Corp. | Method and apparatus for updating a shortest path graph |
CN105323166A (en) * | 2015-11-17 | 2016-02-10 | 东北大学 | Cloud computing-oriented routing method based on network efficiency priority |
CN105337861A (en) * | 2015-11-18 | 2016-02-17 | 东北大学 | Routing method based on energy efficiency priority and cognitive theory |
-
2019
- 2019-10-16 CN CN201910982647.1A patent/CN110784405B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7593341B1 (en) * | 2006-06-08 | 2009-09-22 | At&T Corp. | Method and apparatus for updating a shortest path graph |
CN105323166A (en) * | 2015-11-17 | 2016-02-10 | 东北大学 | Cloud computing-oriented routing method based on network efficiency priority |
CN105337861A (en) * | 2015-11-18 | 2016-02-17 | 东北大学 | Routing method based on energy efficiency priority and cognitive theory |
Also Published As
Publication number | Publication date |
---|---|
CN110784405A (en) | 2020-02-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Jiang et al. | An optimization-based robust routing algorithm to energy-efficient networks for cloud computing | |
CN106411770A (en) | Data center network energy-saving routing algorithm based on software defined network (SDN) architecture | |
CN113810233B (en) | Distributed computation unloading method based on computation network cooperation in random network | |
CN111953758A (en) | Method and device for computing unloading and task migration of edge network | |
CN105515987B (en) | A kind of mapping method based on SDN framework Virtual optical-fiber networks | |
CN106936645B (en) | The optimization method of tree network topology structure based on queueing theory | |
CN105550159A (en) | Power distributing method for network-on-chip of multi-core processor | |
CN108599172A (en) | A kind of transmission & distribution net overall situation tidal current computing method based on artificial neural network | |
CN111726854A (en) | Method for reducing calculation unloading energy consumption of Internet of things | |
CN103297341A (en) | Method for configuring flow on nodes of internal router | |
CN104317646A (en) | Cloud data central virtual machine scheduling method based on OpenFlow frame | |
CN114126066A (en) | MEC-oriented server resource allocation and address selection joint optimization decision method | |
CN114357676A (en) | Aggregation frequency control method for hierarchical model training framework | |
CN113992259A (en) | Method for constructing time slot resource expansion diagram | |
Nace et al. | Computing optimal max-min fair resource allocation for elastic flows | |
CN111506431A (en) | Method for optimizing perception load performance of cloud server under energy consumption constraint | |
CN111404815A (en) | Constrained routing method based on deep learning | |
CN105406517B (en) | Economic Dispatch method based on finite time average homogeneity algorithm | |
CN108810109B (en) | Virtual data center placement method based on energy consumption optimization | |
CN110784405B (en) | Robust energy efficiency routing method for teaching cloud computing network | |
CN107528731B (en) | Network segmentation optimization algorithm applied to NS3 parallel simulation | |
CN107707471A (en) | The energy-saving scheme of minimum conversion link number based on extensive SDN | |
CN110781352B (en) | Method for optimizing topological structure to realize network structure controllability at lowest cost | |
Abdelghany et al. | Wholesale electricity pricing in the presence of geographical load balancing | |
Peng et al. | Edge-cloud collaborative computation offloading for federated learning in smart city |
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 |