CN105721201A - Energy-saving virtual network migration method - Google Patents

Energy-saving virtual network migration method Download PDF

Info

Publication number
CN105721201A
CN105721201A CN201610044363.4A CN201610044363A CN105721201A CN 105721201 A CN105721201 A CN 105721201A CN 201610044363 A CN201610044363 A CN 201610044363A CN 105721201 A CN105721201 A CN 105721201A
Authority
CN
China
Prior art keywords
node
physical
dummy
dummy node
link
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
CN201610044363.4A
Other languages
Chinese (zh)
Other versions
CN105721201B (en
Inventor
苏森
张忠宝
李维天
叶丹娜
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.)
Beijing University of Posts and Telecommunications
Original Assignee
Beijing University of Posts and Telecommunications
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 Beijing University of Posts and Telecommunications filed Critical Beijing University of Posts and Telecommunications
Priority to CN201610044363.4A priority Critical patent/CN105721201B/en
Publication of CN105721201A publication Critical patent/CN105721201A/en
Application granted granted Critical
Publication of CN105721201B publication Critical patent/CN105721201B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08Configuration management of networks or network elements
    • H04L41/0803Configuration setting
    • H04L41/0823Configuration setting characterised by the purposes of a change of settings, e.g. optimising configuration for enhancing reliability
    • H04L41/0833Configuration setting characterised by the purposes of a change of settings, e.g. optimising configuration for enhancing reliability for reduction of network energy consumption
    • YGENERAL 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
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00Reducing energy consumption in communication networks
    • Y02D30/50Reducing energy consumption in communication networks in wire-line communication networks, e.g. low power modes or reduced link rate

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention discloses an energy-saving virtual network migration method and belongs to the technical field of network virtualization in the computer network field.The method comprises the steps of determining virtual nodes to be migrated, ranking the virtual nodes to be migrated to obtain a migration sequence, and conducting virtual node migration periodically.The migration technique is applied to virtual network mapping for the first time, a migration model is provided, and energy is saved by 20% by means of the algorithm compared with a previous optimal algorithm.

Description

A kind of energy-conservation virtual network moving method
Technical field
The invention belongs to the network virtualization technical field of computer network field, be used for solving virtual network migration problem in network virtualization environment, specifically, refer to a kind of energy-conservation virtual network moving method.
Background technology
In recent years, network virtualization technology, as one of key characteristic that Future Internet application itself possesses, increasingly it is subject to the extensive concern of industrial quarters and academia.In the art, its core concept is that infrastructure provider and two roles of service provider are given decoupling: managed physical network by infrastructure provider, proposed the request of lease virtual network by service provider to infrastructure provider, create personalized application to be supplied to terminal use.This virtual network requests usually contains node demand and link requirements.This kind of virtual network requests is mapped to the problem on physical network, it is referred to as virtual network mapping problems (referring to list of references [1] M.Yu, Y.Yi, J.Rexford, andM.Chiang, " Rethinkingvirtualnetworkembedding:substratesupportforpat hsplittingandmigration; " ACMSIGCOMMComputerCommunicationReview, vol.38, no.2, pp.17 29,2008.).
Virtual network mapping problems causes at academia and industrial quarters and pays close attention to widely.But, it is frequent that virtual network requests arrives and departs from very, and the change over time of the resource of physical network is very big.After a period of time, mapping scheme possibility before can not effectively save energy.Therefore, one effective algorithm of necessary design optimizes the energy consumption problem of infrastructure provider further.
Summary of the invention
It is an object of the invention to after virtual network has mapped, for the situation that physical network nodes and link circuit resource dynamically change, design efficient virtual network moving method, energy-conservation further, so that while keeping the long-term running income of physical network, reducing the energy consumption expense of physical network.
The present invention devises virtual network moving method energy-conservation under a kind of dynamic need, and described method comprises the steps:
The first step, it is determined that dummy node to be migrated;
One CPU usage threshold value θ is set, the CPU usage physical node higher than threshold value θ, other dummy node to be migrated will be received as target physical node, and the physical node that CPU usage is lower than threshold value θ, on it, all dummy nodes will as dummy node to be migrated;
Second step, is ranked up dummy node to be migrated, obtains migration series;
3rd step, periodically carries out dummy node migration;
The migration cycle represents with MC, all dummy nodes to be migrated composition set U, and all target physical nodes composition set G, the two set will constitute a bipartite graph;Its limit weights W is defined for this bipartite graphugAs follows:
W u g = Σ l u v ∈ L v Σ l s t ∈ P s t f s t u v B ( l s t ) - Σ l u v ∈ L v Σ l t q ∈ P t q f t q u v B ( l t q ) - - - ( 8 )
Wherein,It is a binary number and if only if virtual link luvIt is mapped to physical pathway PstEqual to 1 time upper, otherwise equal to 0, B (lst) it is virtual link lstThe bandwidth value of request,It is a binary number and if only if virtual link ltqIt is mapped to physical pathway PtqEqual to 1 time upper, otherwise equal to 0, B (ltq) it is virtual link ltqThe bandwidth value of request;The difference that the link circuit resource of the link maps scheme after limit weight table being shown as original link mapping scheme corresponding to this dummy node and completing pre-matching consumes;The implication of definition bipartite graph matching is many-to-one coupling, the KM node weight matching algorithm improved is adopted progressively to find an augmenting path for each dummy node to be migrated, if augmenting path cannot be found in M time, then amendment top mark makes a new limit add, search again for augmenting path, until finding such augmenting path or algorithm beyond the maximum attempts preset;Then node migration, link re-establishment and time complexity analysis are carried out.
It is an advantage of the current invention that:
(1) migrating technology is applied in the problem that virtual network maps by the present invention first, it is proposed that migration models.
(2) compared to optimal algorithm before, the algorithm further energy-conservation 20% proposed in the present invention.
Accompanying drawing explanation
Figure 1A is physical network energy consumption (on a small scale virtual network);
Figure 1B is physical network energy consumption (normal scale virtual network);
Fig. 1 C is physical network energy consumption (large-scale virtual network);
Fig. 2 A is that physical network opens nodes (on a small scale virtual network);
Fig. 2 B is that physical network opens nodes (normal scale virtual network);
Fig. 2 C is that physical network opens nodes (large-scale virtual network).
Fig. 3 is that in embodiment, virtual network migrates schematic diagram.
Detailed description of the invention
Below in conjunction with drawings and Examples, the present invention is described in further detail.
The conceptual description that first present invention provides the physical network, virtual network and the virtual network mapping that relate in virtual network mapping problems is as follows.
Physical network: physical network is labeled as weighted-graph G by the present inventionp=(Np,Lp), wherein NpRepresent the set of all physical nodes, L in physical networkpRepresent the set of all physical links in physical network.For physical node, it can provide CPU computing capability and Memory storage capacity;For physical link, it can provide bandwidth ability.
Virtual network: virtual network is labeled as weighted-graph G by the present inventionv=(Nv,Lv), wherein NvRepresent the set of all dummy nodes, LvRepresent the set of all virtual links.On each dummy node u, there is the constraint of CPU computing capability and Memory storage demand constraint, use C respectivelyuAnd MemuRepresent;On virtual link, there is the constraint of bandwidth ability demand, use BuvRepresent.According to existing research (list of references [2]: G.Sun, H.Yu, L.Li, V.Anand, Y.Cai, andH.Di, " Exploringonlinevirtualnetworksmappingwithstochasticbandw idthdemandinmulti-datacenter, " PhotonicNetworkCommunications, vol.23, no.2, pp.109 122,2012;List of references [3]: B.Heller, S.Seetharaman, P.Mahadevan, Y.Yiakoumis, P.Sharma, S.Banerjee, andN.McKeown, " ElasticTree:Savingenergyindatacenternetworks, " inProceedingsofthe7thNSDI.USENIXAssociation, 2010, pp.249-264;List of references [4]: X.Wang, Y.Yao, X.Wang, K.Lu, andQ.Cao, " Carpo:Correlation-awarepoweroptimizationindatacenternetw orks, " inINFOCOM.IEEE, 2012, pp.1125 1133;List of references [5]: L.Wang, F.Zhang, J.A.Aroca, A.V.Vasilakos, K.Zheng, C.Hou, D.Li, andZ.Liu, " Greendcn:ageneralframeworkforachievingenergyefficiencyin datacenternetworks; " SelectedAreasinCommunications, IEEEJournalon, vol.32, no.1, pp.4 15,2014;List of references [6]: M.Wang, X.Meng, andL.Zhang, " Consolidatingvirtualmachineswithdynamicbandwidthdemandin datacenters, " inINFOCOM.IEEE, 2011, pp.71 75.), in real world applications, the dynamic change often of the demand of CPU computing capability and the demand of bandwidth ability, and meet Gauss distribution, it may be assumed that Wherein, μu,Represent the constraint C of CPU capability requirement respectivelyuAverage and variance, μuv,Represent the constraint B of broadband ability demand respectivelyuvAverage and variance.Owing to virtual network can dynamically arrive and depart from, present invention taAnd teRepresent that virtual network requests arrives and departs from the time of physical network respectively.
Virtual network maps: the weighted-graph G of a given virtual networkvWeighted-graph G with physical networkp, map M:G in virtual networkv→GpIn mapping, comprise two mapping: MnAnd Ml
MnRefer to the set N from dummy nodevSet N to physical nodepInjective function.In this injective function, for any dummy node u ∈ NvWith physics routing node Mn(u)∈Np, MnU () needs to meet following two condition: (1) meets the CPU computing capability constraint C of physical node uu: (2) meet the Memory storage demand constraint Mem of physical node uu:
MlRepresent the set L from virtual linkvSet P to loop free pathpInjective function, wherein, PpRepresent by all physical link LpThe set of the loop free path of composition.In this injective function, for any virtual link luv∈Lv, Ml(luv)∈PpOn path, all of physical link is required for meeting luvBandwidth demand.
Owing to the energy consumption of physical node almost presents linear relationship (referring to list of references [7]: S.Rivoire with the load of CPU, P.Ranganathan, andC.Kozyrakis, " Acomparisonofhigh-levelfull-systempowermodels; " inProceedingsofthe2008conferenceonPowerawarecomputingand systems.USENIXAssociation, 2008, pp.3 3.;List of references [8]: X.Fan, W.Weber, andL.Barroso, " Powerprovisioningforawarehouse-sizedcomputer, " inProceedingsofthe34thannualinternationalsymposiumonComp uterarchitecture.ACM, 2007, pp.13 23.), and the energy consumption including internal memory and hard disk changes less (referring to list of references [9]: D.Economou, S.Rivoire, C.Kozyrakis, andP.Ranganathan, " Full-systempoweranalysisandmodelingforserverenvironments, " inInProceedingsofWorkshoponModeling, Benchmarking, andSimulation, 2006, pp.70 77.).Therefore, the energy consumption PN of present invention approximate representation physics multihome node in the following manner:
PN=Pb+Pl·m(1)
Wherein, PbRepresent the power of physical node during idling load, be referred to as baseline energy consumption power;M represents the load of current CPU;PlRepresent that physical node power is with the m linear dimensions changed.
Based on the energy consumption PN of above-mentioned physics multihome node, for maps virtual node u (CPU capability requirement CuRepresent), the extra energy consumption that physics multihome node j needs is:
ΔPN j u = P b + P l · C u PS j = 0 P l · C u PS j ≠ 0 - - - ( 2 )
Wherein, PSjRepresent the on off state of physics multihome node j: if PSj=0, represent that this physics multihome node j is in inactive (dormancy) state, be otherwise in active (work) state.
Physical node is divided into two categories below by the present invention: (1) multihome node, namely provides the physical node of CPU computing capability in node mapping process;(2) forward node, the i.e. transit node forwarding packet in physical pathway in link maps process.NH (t) and NF (t) is utilized to be illustrated respectively in t the quantity of multihome node and the transit node (being also forward node) opened.Based on above-mentioned physics multihome node energy consumption PN, the node energy consumption that Computational Physics network consumes in time T is as follows:
E ( T ) = ∫ 0 T [ P b ( N H ( t ) + N F ( t ) ) + P l Σ i = 1 G v ( t ) Σ u ∈ N v i C u ( t ) ] d t - - - ( 3 )
Wherein, GvT () represents the virtual request quantity being complete mapping in t;Represent and be complete dummy node u, C in the i-th virtual request of mapping in tuT () represents the dummy node u constraint size in the CPU computing capability request amount of t.Because a series of virtual network arrives over time and leaves, bottom-layer network resource dynamically changes.When a request arrives, even if the optimum mapping scheme under have found present case, after cannot guarantee that a period of time, the program is still optimum.In order to solve this problem, present invention migrating technology optimize further before mapping scheme (referring to list of references [10] C.Clark, K.Fraser, S.Hand, J.G.Hansen, E.Jul, C.Limpach, I.Pratt, andA.Warfield, " Livemigrationofvirtualmachines, " inProceedingsofthe2ndconferenceonSymposiumonNetworkedSys temsDesign&Implementation-Volume2.USENIXAssociation, 2005, pp.273 286).But, this technology has two-sidedness.
The advantage migrated: because dummy node can be moved to another physical node from a physical node, therefore also just can all dummy nodes on underloading physical node be moved on the physical node of one or several heavy duty, it is then shut off underloading physical node, thus reaching energy-conservation target.Because only that could be closed by this physical node after all being moved out by all dummy nodes on certain physical node, therefore, the physical node number N following formula closed in transition process is represented by the present invention:
N = Σ g = 0 | G | Π u = 0 | U | M s t u - - - ( 4 )
Wherein | G | is the sum that there is physical node to be migrated, and | U | is the sum of all dummy nodes to be migrated,Being a binary number, if successfully dummy node u having been moved to physical node g from physical node s, then it is equal to 1, otherwise equal to 0.
The shortcoming migrated: when migrating, it is necessary to the memory of dummy node is moved to target physical node from source physical node, it is necessary to find a paths for the duplication of memory.The physical node that this paths relates to will be opened, and is used for forwarding memory data, and the new physical node opened is by consumed energy.Migration simultaneously can cause service disruption, if using the method for pre-copy to migrate, the break period will be 1/10th Milliseconds, and user will be substantially unaffected.The cost of migration can be represented by below equation.
C = Σ u = 0 | U | Mem u m i n l s t ∈ P s t B s t - - - ( 5 )
Wherein MemuRepresent the Memory storage demand constraint of dummy node u, PstRepresent from source physical node to the path of target physical node, lstRepresent PstOn each paths, | U | be all dummy nodes to be migrated sum, BstRepresent the remaining bandwidth size in path.
The primary evaluation index that virtual network maps includes the long-term average power expense of network average running income and the physical network for a long time of physics.
The long-term average running income of described physical network can represent with following formula:
lim T → ∞ Σ t = 0 T R i ( G v ) T , - - - ( 6 )
WhereinRepresent the income that infrastructure provider produces when successfully mapping i-th virtual network requests, BstRepresent the remaining bandwidth size in path.
Correspondingly, the long-term average power expense of physical network can be defined as:
lim T → ∞ Σ t = 0 T E ( T ) T , - - - ( 7 )
Wherein, the node energy consumption that E (T) consumes in time T for physical network.
First the present invention introduces a kind of energy-conservation virtual network moving method, EnergyAwareVirtualNetworkEmbeddingwithMigration, is called for short EA-VNM.EA-VNM is a kind of triphasic virtual network moving method, migrates three phases including stage that remaps of preparatory stage, node and link and node and link.Specifically comprise the following steps that
The first step, it is determined that dummy node to be migrated;
One CPU usage threshold value θ is set, the CPU usage physical node higher than threshold value θ, other dummy node to be migrated will be received as target physical node, and the physical node that CPU usage is lower than threshold value θ, on it, all dummy nodes will as dummy node to be migrated.
Second step, is ranked up dummy node to be migrated, obtains migration series;
The design of migration series is by strong influence energy-saving effect, and the present invention arranges the purpose of this migration series and is in that closedown underloading physical node as much as possible, and dummy node distribution is concentrated more.Ordering rule is as follows, first by rule 1, is then ensureing that regular 1 uses rule 2 when satisfied:
Rule 1: for certain physical node p, if its CPU makes consumption more big, then dummy node to be migrated thereon is by more high for the priority obtained.
Rule 2: for certain dummy node u, the cpu resource of its request is more many, by more high for the priority obtained.
3rd step, periodically carries out dummy node migration.
The migration cycle represents with MC, and the concrete numerical value of MC will rule of thumb data be arranged.
All dummy nodes to be migrated composition set U, all target physical nodes composition set G, the two set will constitute a bipartite graph.Its limit weights W is defined for this bipartite graphugAs follows:
W u g = Σ l u v ∈ L v Σ l s t ∈ P s t f s t u v B ( l s t ) - Σ l u v ∈ L v Σ l t q ∈ P t q f t q u v B ( l t q ) - - - ( 8 )
Wherein,It is a binary number and if only if virtual link luvIt is mapped to physical pathway PstEqual to 1 time upper, otherwise equal to 0, B (lst) it is virtual link lstThe bandwidth value of request,It is a binary number and if only if virtual link ltqIt is mapped to physical pathway PtqEqual to 1 time upper, otherwise equal to 0, B (ltq) it is virtual link ltqThe bandwidth value of request.The difference that the link circuit resource of the link maps scheme after limit weight table being shown as original link mapping scheme corresponding to this dummy node and completing pre-matching consumes.
The implication of definition bipartite graph matching is many-to-one coupling, rather than man-to-man coupling in original bipartite graph matching.
This stage needs to reach following two target: with energy-conservation for primary goal, under the premise maximizing energy-saving effect, it is ensured that bigger income.
Based on the Optimum Matching method KM algorithm in bipartite graph, the present invention have devised the KM node weight matching algorithm of improvement, specifically comprises the following steps that
Step 1: for each dummy node u to be migrated,
Step 2:a=0 (a is counting variable);
Step 3: if a is < M (M be default threshold value);
Step 4: call algorithm 2, if returning mistake, performing step 5 and 6, otherwise returning to step 1;
Step 5: call function update_label ();
Step 6:a adds 1;
Step 7: by the state before physical network recovery to step 6.
The KM node weight matching algorithm of above-mentioned improvement progressively will find an augmenting path for each dummy node to be migrated, if augmenting path cannot be found in M time, then amendment top mark makes a new limit add, search again for augmenting path, until finding such augmenting path or algorithm beyond the maximum attempts preset.
Algorithm 2: find augmenting path;
Input: be currently needed for the dummy node u, target physical node queue L of searchg,
Output: record which dummy node and match the vector M of which physical nodeGU, record which physical node and match the vector M of which dummy nodeGU
Step 1:visitU [u]=1,
Step 2: for each at LgIn physical node g to be migrated carry out,
Step 3: if (Wu+Wg=WugAnd MUG[u]!=g) then,
Step 4:visitG [g]=1;
Step 5: if (node_constrain_check () and link_constrain_check ()), carry out step 6-9, otherwise carry out step 10-20,
Step 6: update PN resource situation;
Step 7:MUG[u]=g;
Step 8: u is placed on MGU[g] tail of the queue;
Step 9: return correct;
Step 10: for each at MGUDummy node i in [g],
Step 11: discharge the resource PN about dummy node i in PN;
Step 12: at MGU[g] deletes dummy node i;
Step 13: if (node_constrain_check () &&link_constrain_check ()) carries out step 14-20,
Step 14: discharge the resource PN about dummy node i in PN;
Step 15: u is placed on MGUThe tail of the queue of [g];
Step 16: if calling algorithm 2 return mistake to carry out step 17-19, otherwise carry out step 20,
Step 17: update PN resource;
Step 18:MUG[u]=g;
Step 19: return correct;
Step 20: recover PN to the state before step 13;
Step 3 at algorithm 2 arrives step 9, and first time attempts dummy node u is matched physical node g.If successful match, algorithm returns correct.If it fails, certain dummy node matching physical node g will be discharged, and again find new coupling physical node for this dummy node.If can not find, algorithm returns mistake.
Being explained as follows of letter in algorithm 2 and function:
It is 1 that 1.visitU [u] this array have recorded the dummy node which dummy node is accessed, and is otherwise 0.
It is 1 that 2.visitG [g] this array have recorded the physical node which physical node is accessed, and is otherwise 0.
3.node_constrain_check () (node restriction checks) will check whether that this physical node has enough CPU and memory resources, and whether has the dummy node from same request on it.Only above 2 satisfied just returns simultaneously are correct, otherwise return mistake.
Can link_constrain_check () (link restriction checks) be that the dummy node migrated finds new link maps scheme by checking, if success, returns correct, otherwise returns mistake.
4.update_label () (renewal top mark) function will first look at which physical node and dummy node in algorithm 2 and be accessed.For LuThe middle point accessed and LgIn the point that has not visited, calculate slack=Wu+Wg-WugMinima, wherein WuAnd WtIt is L respectivelyuAnd LgThe top mark at midpoint.For LuIn point, their top mark value will deduct slack, for LgIn point, their top mark value will plus slack.
Node migrates: determining after which dummy node will migrate into which physical node, will find a paths to be used for memory duplication.In order to reduce the cost of migration, the path allocation of big bandwidth will be had to having the dummy node of big memory as much as possible, and first selected by the dummy node having big memory when distributing path.
Link re-establishment: completing after memory replicates, target physical node will be rebuild and from the link between other physical node of same request.First find the path of minimum hop count with shortest path first, then select from these paths and have turned on the maximum path of nodes as link re-establishment scheme.
Time complexity is analyzed
KM node weight matching algorithm according to above-mentioned improvement and algorithm 2, the time complexity that can draw algorithm 2 is O (| V | | P |), improve KM node weight matching algorithm time complexity be O (| V | | P |) therefore, the time complexity of whole node migration algorithm is O (| V |2|P|2)。
The present invention uses GT-ITM instrument (referring to list of references [12]: E.Zegura, K.Calvert, andS.Bhattacharjee, " Howtomodelaninternetwork; " inINFOCOM ' 96.FifteenthAnnualJointConferenceoftheIEEEComputerSociet ies.NetworkingtheNextGeneration, IEEE, 1996, vol.2, pp.594 602.) produce the topology of physical network and virtual network.With list of references [11, 13, 14] (list of references [11]: X.Cheng, S.Su, Z.Zhang, H.Wang, F.Yang, Y.Luo, andJ.Wang, " VirtualNetworkEmbeddingThroughTopology-AwareNodeRanking, " ACMSIGCOMMComputerCommunicationReview, vol.41, no.2, pp.39 47, 2011. list of references [13]: S.Su, Z.Zhang, X.Cheng, Y.Wang, Y.Luo, andJ.Wang, " Energy-awarevirtualnetworkembeddingthroughconsolidation, " inIEEEINFOCOMWS-CCSES:GreenNetworkingandSmartGrids, 2012, pp.2708 2713. list of references [14]: S.Su, Z.Zhang, A.X.Liu, X.Cheng, Y.Wang, andX.Zhao, " Energy-awarevirtualnetworkembedding, " IEEETransactionsonNetworking, 2014, vol.22, no.5, pp.1607-1620.) similar.Form 1 gives detailed parameter and arranges.In order to portray the dynamic of node and bandwidth request, the present invention creates all requests based on Gauss distribution.The present invention is by adjusting the scale asked from carrying out evaluation algorithms design to Standard capacity on a small scale.When being set on a small scale, the number of the dummy node in each request is evenly distributed between 2 to 5;When being set to Standard capacity, the number of the dummy node in each request is evenly distributed between 2 to 10.Present invention assumes that the time of advent of VN virtual request meets Poisson distribution: every 100 unit of time on average have 4 requests to arrive.And assume that the persistent period of each request meets the exponential that average is 500 unit of time.In an experiment embodiment, one has 50000 unit of time, comprises 2000 virtual network requests.Experiment runs example 10 groups different, finally calculates the meansigma methods of 10 groups.The present invention arrange migration cycle MC be 200 unit of time/time, CPU thresholding θ is 0.3.P is setbAnd PlRespectively 165 watts/CPU and 15 watts/CPU.Identical with list of references [14], E is setSFor the energy consumption under full load state.
Table 1 simulation parameter is arranged
Topology Physical network Virtual network
Node 50 2-5&2-10&2-15
Link probability 50% 50%
CPU 50-100 N (μ, σ2) (μ ∈ [0,20], σ ∈ [0,10])
Bandwidth 50-100 N (μ, σ2) (μ ∈ [0,50], σ ∈ [0,10])
The present invention utilize C++ realize algorithm and with up-to-date mapping algorithm EA-VNE[13-14]Carry out detailed comparison.Virtual request is used as constant by this algorithm, have ignored the dynamic of CPU and bandwidth.Owing to the algorithm of meta-heuristic and the algorithm of the present invention adhere to separately different classes of, thus the present invention not with meta-heuristic method comparison.Experiment conclusion and interpretation of result are as follows:
Compared with EA-VNE, the algorithm that the present invention proposes can significantly decrease energy consumption: Figure 1A shows that EA-VNM algorithm significantly decreases energy consumption under input on a small scale.Such as, at the 48000th unit of time, the observable index EA-VNE of EA-VNM reduces 25%.Reason is in that, EA-VNM utilizes migrating technology to be assembled by dummy node so that the physical node number of unlatching reduces, and this point can be observed from Fig. 2 A~Fig. 2 C and be obtained.By Figure 1A, 1B and 1C it is observed that along with the reduction of requesting node scale, the energy consumption of saving also becomes apparent from.This is because along with the scale of request increases, the resource residual amount of physical network declines, and this makes the virtual network quantity of Successful migration reduce.
Compared with EA-VNE, this algorithm achieves and is close to the same income.This is because when migrating, we impart bipartite graph limit weights the limit that prioritizing selection weights are big the bandwidth resources difference migrating front and back consumption, so that utilization of resources.The operation time of this algorithm is essentially the same with EA-VNE: the present invention is as follows for the server record emulated, Intel double-core 3GHzCPU, 2GB internal memory, 160G hard disk, Linux2.6 operating system.According to experimental result, algorithm EAD-VNE and EA-VNE maps the average performance times of a virtual network requests all in a few tens of milliseconds rank.But the operation time that algorithm EA-VNM is than EA-VNE slightly increases.This is because algorithm EAD-VNE needs additionally a small amount of time to be used for migrating dummy node, this is acceptable.
Embodiment
Present invention could apply to, in the backbone network of support network virtualization technology or data center network environment, utilizing the dynamic of virtual network requests, reducing the physical node opened in virtual network process, thus reducing the purpose of physical network energy consumption expense.EA-VNM algorithm is using the topological structure of physical network and virtual network requests and resource capability situation as input, using the virtual network mapping scheme of preferably energy consumption perception as output.Such as, Fig. 3 gives an embodiment.In the figure, Fig. 3 (a) represents a physical network, represents the average of this node cpu ability near nodal rectangle, the average of this link bandwidth ability of the digitized representation on link.Fig. 3 (b) represents a virtual network, represents the average of this node cpu ability need near nodal rectangle, the average of this link bandwidth ability need of the digitized representation on link.Fig. 3 (c) represents this virtual network mapping scheme on physical network, and { a → A, b → C, c → F}, link maps scheme is { ab → ABC, bc → CBF, ac → AF}.Fig. 3 (d) gives a feasible migration scheme, and by dummy node b is moved to physical node B from physical node C, new virtual network node mapping scheme is { a → A, b → C, c → B}, link maps scheme is { ab → AB, bc → BF, ac → AF}.After completing migration, physical node C will be closed, thus saving energy consumption.

Claims (4)

1. virtual network moving method one kind energy-conservation, it is characterised in that: comprise the following steps that,
The first step, it is determined that dummy node to be migrated;
One CPU usage threshold value θ is set, the CPU usage physical node higher than threshold value θ, other dummy node to be migrated will be received as target physical node, and the physical node that CPU usage is lower than threshold value θ, on it, all dummy nodes will as dummy node to be migrated;
Second step, is ranked up dummy node to be migrated, obtains migration series;
3rd step, periodically carries out dummy node migration;
The migration cycle represents with MC, all dummy nodes to be migrated composition set U, and all target physical nodes composition set G, the two set will constitute a bipartite graph;Its limit weights W is defined for this bipartite graphugAs follows:
W u g = &Sigma; l u v &Element; L v &Sigma; l s t &Element; P s t f s t u v B ( l s t ) - &Sigma; l u v &Element; L v &Sigma; l t q &Element; P t q f t q u v B ( l t q ) - - - ( 8 )
Wherein,It is a binary number and if only if virtual link luvIt is mapped to physical pathway PstEqual to 1 time upper, otherwise equal to 0, B (lst) it is virtual link lstThe bandwidth value of request,It is a binary number and if only if virtual link ltqIt is mapped to physical pathway PtqEqual to 1 time upper, otherwise equal to 0, B (ltq) it is virtual link ltqThe bandwidth value of request;The difference that the link circuit resource of the link maps scheme after limit weight table being shown as original link mapping scheme corresponding to this dummy node and completing pre-matching consumes;The implication of definition bipartite graph matching is many-to-one coupling, the KM node weight matching algorithm improved is adopted progressively to find an augmenting path for each dummy node to be migrated, if augmenting path cannot be found in M time, then amendment top mark makes a new limit add, search again for augmenting path, until finding such augmenting path or algorithm beyond the maximum attempts preset;Then node migration, link re-establishment and time complexity analysis are carried out.
2. a kind of energy-conservation virtual network moving method according to claim 1, it is characterised in that: the ordering rule that the sequence described in second step adopts is, first by rule 1, then uses rule 2 when guarantee rule 1 meets;
Rule 1: for certain physical node p, if its CPU makes consumption more big, then dummy node to be migrated thereon is by more high for the priority obtained;
Rule 2: for certain dummy node u, the cpu resource of its request is more many, by more high for the priority obtained.
3. a kind of energy-conservation virtual network moving method according to claim 1, it is characterised in that: the KM node weight matching algorithm of described improvement, specifically comprise the following steps that
Step 1: for each dummy node u to be migrated,
Step 2:a=0, a are counting variable;
Step 3: if a < M, M are default threshold value;
Step 4: call algorithm 2, if returning mistake, performing step 5 and 6, otherwise returning to step 1;
Step 5: call function update_label ();
Step 6:a adds 1;
Step 7: by the state before physical network recovery to step 6;
Algorithm 2: find augmenting path;
Input: be currently needed for the dummy node u, target physical node queue L of searchg,
Output: record which dummy node and match the vector M of which physical nodeUG, record which physical node and match the vector M of which dummy nodeGU
Step (1): visitU [u]=1,
Step (2): for each at LgIn physical node g to be migrated carry out,
Step (3): if (Wu+Wg=WugAnd MUG[u]!=g) then,
Step (4): visitG [g]=1;
Step (5): if (node_constrain_check () and link_constrain_check ()), carry out step 6-9, otherwise carry out step 10-20,
Step (6): update PN resource situation;
Step (7): MUG[u]=g;
Step (8): u is placed on MGU[g] tail of the queue;
Step (9): return correct;
Step (10): for each at MGUDummy node i in [g],
Step (11): discharge the resource PN about dummy node i in PN;
Step (12): at MGU[g] deletes dummy node i;
Step (13): if (node_constrain_check () &&link_constrain_check ()), carries out step 14-20,
Step (14): discharge the resource PN about dummy node i in PN;
Step (15): u is placed on MGUThe tail of the queue of [g];
Step (16): if calling algorithm 2 return mistake to carry out step 17-19, otherwise carry out step 20,
Step (17): update PN resource;
Step (18): MUG[u]=g;
Step (19): return correct;
Step (20): recover PN to the state before step 13;
In above-mentioned algorithm:
It is 1 that visitU [u] this array have recorded the dummy node which dummy node is accessed, and is otherwise 0;
It is 1 that visitG [g] this array have recorded the physical node which physical node is accessed, and is otherwise 0;
Node_constrain_check () function checks whether that this physical node has enough CPU and memory resources, and whether has the dummy node from same request on it, and only above 2 satisfied just returns simultaneously are correct, otherwise return mistake;
Can the inspection of link_constrain_check () function be that the dummy node migrated finds new link maps scheme, if success, returns correct, otherwise returns mistake;
Update_label () function will first look at which physical node and dummy node in algorithm 2 and be accessed, for LuThe middle point accessed and LgIn the point that has not visited, calculate slack=Wu+Wg-WugMinima, wherein WuAnd WtIt is L respectivelyuAnd LgThe top mark at midpoint, for LuIn point, their top mark value will deduct slack, for LgIn point, their top mark value will plus slack.
4. a kind of energy-conservation virtual network moving method according to claim 1, it is characterized in that: described node migrates and refers to, which dummy node determining after will migrate into which physical node, a paths will be found to be used for memory replicate, in order to reduce the cost of migration, the path allocation of big bandwidth will be had to having the dummy node of big memory as much as possible, and first selected by the dummy node having big memory when distributing path;
Described link re-establishment refers to, completing after memory replicates, will rebuild target physical node and from the link between other physical node of same request.
CN201610044363.4A 2016-01-22 2016-01-22 A kind of energy-efficient virtual network moving method Active CN105721201B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610044363.4A CN105721201B (en) 2016-01-22 2016-01-22 A kind of energy-efficient virtual network moving method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610044363.4A CN105721201B (en) 2016-01-22 2016-01-22 A kind of energy-efficient virtual network moving method

Publications (2)

Publication Number Publication Date
CN105721201A true CN105721201A (en) 2016-06-29
CN105721201B CN105721201B (en) 2018-12-18

Family

ID=56153853

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610044363.4A Active CN105721201B (en) 2016-01-22 2016-01-22 A kind of energy-efficient virtual network moving method

Country Status (1)

Country Link
CN (1) CN105721201B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107819662A (en) * 2017-11-23 2018-03-20 电子科技大学 A kind of service function chain real-time migration method applied to cloud and mist calculating network
CN112688848A (en) * 2020-12-17 2021-04-20 北京邮电大学 Heterogeneous virtual network mapping method and system

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102868733A (en) * 2012-08-29 2013-01-09 北京邮电大学 Method for remapping virtual network resources
CN102904794A (en) * 2012-09-27 2013-01-30 北京邮电大学 Method and device for mapping virtual network
CN103684960A (en) * 2012-08-29 2014-03-26 株式会社日立制作所 Network system and virtual node migration method
CN103856355A (en) * 2014-03-12 2014-06-11 湖州师范学院 Energy consumption perceptive virtual network mapping transportation model and algorithm

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102868733A (en) * 2012-08-29 2013-01-09 北京邮电大学 Method for remapping virtual network resources
CN103684960A (en) * 2012-08-29 2014-03-26 株式会社日立制作所 Network system and virtual node migration method
CN102904794A (en) * 2012-09-27 2013-01-30 北京邮电大学 Method and device for mapping virtual network
CN103856355A (en) * 2014-03-12 2014-06-11 湖州师范学院 Energy consumption perceptive virtual network mapping transportation model and algorithm

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
SEN SU,ZHONGBAO ZHANG,ALEX X. LIU等: "Energy-Aware Virtual Network Embedding", 《IEEE/ACM TRANSACTIONS ON NETWORKING》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107819662A (en) * 2017-11-23 2018-03-20 电子科技大学 A kind of service function chain real-time migration method applied to cloud and mist calculating network
CN107819662B (en) * 2017-11-23 2019-10-18 电子科技大学 A kind of service function chain real-time migration method calculating network applied to cloud and mist
CN112688848A (en) * 2020-12-17 2021-04-20 北京邮电大学 Heterogeneous virtual network mapping method and system

Also Published As

Publication number Publication date
CN105721201B (en) 2018-12-18

Similar Documents

Publication Publication Date Title
Aujla et al. MEnSuS: An efficient scheme for energy management with sustainability of cloud data centers in edge–cloud environment
CN103179052B (en) A kind of based on the central virtual resource allocation method and system of the degree of approach
CN103076870B (en) Scheduling and dynamic resource allocation method are merged in the application that in data center, energy consumption drives
CN103457752A (en) Virtual network mapping method
CN103595780A (en) Cloud computing resource scheduling method based on repeat removing
CN104104621B (en) A kind of virtual network resource dynamic self-adapting adjusting method based on Nonlinear Dimension Reduction
CN103455363B (en) Command processing method, device and physical host of virtual machine
CN105577806B (en) A kind of distributed caching method and system
CN104932996A (en) A method, apparatus, and system for controlling power consumption of unused hardware of a link interface
Tseng et al. Service-oriented virtual machine placement optimization for green data center
Wang et al. Achieving energy efficiency in data centers using an artificial intelligence abstraction model
CN104778112A (en) System and method of adaptive voltage frequency scaling
Wu et al. Towards collaborative storage scheduling using alternating direction method of multipliers for mobile edge cloud
Liao et al. Energy consumption optimization scheme of cloud data center based on SDN
Fan et al. Node essentiality assessment and distributed collaborative virtual network embedding in datacenters
Sadegh et al. A two-phase virtual machine placement policy for data-intensive applications in cloud
CN105721201A (en) Energy-saving virtual network migration method
CN105429843A (en) Energy-saving virtual network mapping method under dynamic requirements
Qin et al. An improved Top‐K algorithm for edge servers deployment in smart city
CN113157383A (en) Method for dynamically adjusting super-proportion in OpenStack environment
CN104348695A (en) Artificial immune system-based virtual network mapping method and system thereof
CN106789289B (en) Method and apparatus for virtual network mapping
CN105306547A (en) Data placing and node scheduling method for increasing energy efficiency of cloud computing system
Zhang et al. Repeatable multi-dimensional virtual network embedding in cloud service platform
Zhao et al. Virtual machine allocation method for cloud computing based on multi-objective evolutionary algorithm

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant