Transcription
IEEEGLOBECOM2015,SanDiego,CAAHSN- ‐I- ession)TowardsMin- liangXue,XiangZhangTowardsMin- onorthepolicyofthefederalgovernment.
TowardsMin- 015,SanDiego,CAAHSN- ‐I- ession)3Introduction33 EtobealsoappliedtotheVIEproblem.4 1264 4VNR 24Virtual Network Embedding (VNE)543 One- ‐to- ‐onevirtual- irtual)linksPhysical)nodePhysical)node onfromrecently.33Tworelatedproblems: VirtualInfrastructureEmbedding(VIE)77 2 4VNR 1 3Embeddingvirtualnetworkedinfra- e,XiangZhang4 lityandreducesbandwidthVirtual Infrastructure Embedding (VIE)5 Serverconsolidation:many- ‐to- ‐onenodemapping7431546 Virtuallinkcanmaptophysicalpathorintra- yofthefederalgovernment.
TowardsMin- 015,SanDiego,CAAHSN- ‐I- ession)Existing algorithm: robability2.3.Form relaxed integer programformulation L;30.20.8Virtual Node Mapping:1.70.8X85Solve L and sort all variables;For each node mapping variable doa)Round the variable basedon its value;5 10 50.40.60.2Virtual Link Mapping:4.Case 1: node conflict happenswhen two nodes are bothpartially mapped to one host0.2 that has insufficient capacity.Solve Multi-Commodity-Flowfor link mapping.5X40.86Case 2: link conflict happenswhen two nodes are partiallymapped to two host that hasinsufficient bisectionalbandwidth between them.1ViNE AlgorithmViNEisarounding- nflictsduringrounding:1. Nodeembeddingconflict,and122. flectthepositionorthepolicyofthefederalgovernment.
TowardsMin- 015,SanDiego,CAAHSN- ‐I- ession)Case 1: node conflictsolved via re-solvingthe program androunding7RuozhouYu,GuoliangXue,XiangZhang3VIE Sequential RoundingVirtual Node Mapping:0.20.870.20.831.1.080.670.3355 10 582.a)Solve L and sort fractionalvariables;b)Round the first variablebased on its value;0.80.25 10 56While some node not mapped do50.40.6Form relaxed integer programformulation L;461271.0Case 2: link conflict solved viare-solving the program androunding6460.2c)Update L based on therounded variable;VIE- andre- gandroundingprocesstheSequentialRounding(VIE- ‐SR)algorithm.0.8Virtual Link Mapping:124.7Solve Multi-Commodity-Flow forlink mapping.Sequential nt.
IEEEGLOBECOM2015,SanDiego,CAAHSN- ‐I- ession)TowardsMin- etheVIE- ithbothdeterministic(D- ‐ViNE)andrandomized(R- ‐ViNE)rounding.ExperimentresultsshowthattheVIE- ementations.ThisvalidatesthatVIE- inalpaper.Evaluation & ResultsIn all scenarios, VIE-SR achieves better acceptance ratio than both R-ViNE and D-ViNE. The average costshows similar results, which can be found in our paper.DiscussionsTime ComplexityGeneralityOptimal SolutionThe linear factor induced by sequentialrounding can be reduced by constrainingthe amount of program re-solving in thealgorithm. In practice, the max amount ofprogram re-solving can be selected as aconstant that achieves the best trade-offbetween performance and running time.Since the proposed algorithm is based onthe program formulation of the problem,various objectives and constraints can beadopted to achieve different goals.The optimal solution can be achieved bycombining VIE-SR with backtracking. Infact, VIE-SR can be viewed as a goodtrade-off between the optimality ofbacktracking, and the polynomial time ofa simple rounding ment.
Towards(Min,Cost(Virtual(Infrastructure(Embedding Ruozhou(Yu,(Guoliang(Xue,(Xiang(Zhang Yu,# Xue# and# Zhang# ({ruozhouy,# xue,# xzhan229}@asu.edu)# are# all# with# Arizona State# University,# Tempe,# AZ# 85287.# All# . Server#consolidation# increases#success# probability#and# reduces#bandwidth 6 5 4 4 7 6 1 5 4 15 7 3. IEEE#GLOBECOM2015,#San .
