optimal selection of dynamic routing protocol with real time case studies kalyan2012

5
Optimal Selection of Dynamic Routing Protocol with Real Time Case Studies G. P. Sai Kalyan, D.Venkata Vara Prasad Department of Computer Science and Engineering SSN College of Engineering, OMR Road, Chennai, India E-mail: [email protected], [email protected] Abstract – Selection of a Dynamic Routing Protocol has great impact on the utilization of various network resources. The factors like usage of bandwidth and behaviour towards growing entries in the routing table, determine the suitability of a protocol. In this paper, we have measured and compared the bandwidths utilised by the three IPv4 dynamic routing protocols, RIPv2 (Routing Information Protocol), EIGRP (Enhanced Interior Gateway Routing Protocol) and OSPFv2 (Open Shortest Path First) with real time case studies. We have also addressed the issue of choosing the optimal protocol as per the network specifications. Keywords- Bandwidth, Dynamic Routing Protocols, RIP, EIGRP, OSPF, VLSM I. INTRODUCTION Routing protocol is the language spoken by routers with their neighbouring routers to share information about the reachability and the status of the prefixes they know. These protocols determine the best path to a destination and note them in their routing tables. Choosing the correct protocol that satisfies the given requirements of a network is very crucial and generally a tedious task [2]. Many factors have to be considered, both about the network as well as the protocol that we would like to go with. The first question will be “Which protocol is better?” And the answer for it is “Well, it depends”. Now comes the question “Depends on what?” And answer to this question is the key part for designing any network. The decision parameters that help in deciding upon the right protocol to be used in the network are: Speed of convergence: How fast is the routing table convergence attained when updates or failures happen in the network? Scalability of network: Are there any chances of extending this network in the future? Will more devices be added to this network? Or is it going to be broken into many subnets? Utilization of resources: Which protocol adds less overhead to the routers? Also, the Protocol dependant resources like CPU, queue buffer, memory and bandwidth must be dealt with Maintenance and management: Which protocol makes the network more flexible and easily maintainable? Which protocol offers relatively easier troubleshooting and management? In this paper, we have measured the bandwidths used in sending updates and hello messages by various IPv4 dynamic routing protocols in real time using “Real-Time Bandwidth Monitor” from SolarWinds software and have compared them. We have also addressed the issue of selecting the optimal protocol for a network to the largest possible extent by discussing various possible network requirements. II. OVERVIEW OF DYNAMIC ROUTING PROTOCOLS A. Routing Information Protocol (RIP) The Routing Information Protocol (RIP) is a veteran distance-vector routing protocol that uses UDP port 520 for message encapsulation [7]. It consists of two message types [1]: 1. A request message is used to ask neighboring routers to send an update [7]. 2. A response message carries the update. When RIP is configured on a router, it sends broadcast packets containing the request message out the entire RIP enabled interfaces and then listens for response messages. Routers receiving the request message respond to it by sending their routing tables in the response message. This process continues until the network is converged. A RIP router sends out its full routing table in its update once in 30 seconds. If any new entry is found in an update, the RIP router enters it into the routing table along with the sending router’s address. It uses the hop count as a metric for determining best paths. The maximum hop count is 15; thereby preventing routing loops in the network. This also limits the size of the network supported by it. If the hop count of an incoming route is 16, it is considered to be inaccessible or undesirable and is at an infinite distance. 219 978-1-4673-0255-5/12/$31.00 c 2012 IEEE

Upload: muhammad-hafizan

Post on 17-Aug-2015

227 views

Category:

Documents


1 download

DESCRIPTION

kkkk

TRANSCRIPT

Optimal Selection of Dynamic Routing Protocol with Real Time Case Studies G. P. Sai Kalyan, D.Venkata Vara PrasadDepartment of Computer Science and Engineering SSN College of Engineering, OMR Road, Chennai, India E-mail: [email protected], [email protected] AbstractSelectionofaDynamicRoutingProtocolhasgreat impactontheutilizationofvariousnetworkresources.The factorslikeusageofbandwidthandbehaviourtowardsgrowing entriesintheroutingtable,determinethesuitabilityofa protocol.Inthispaper,wehavemeasuredandcomparedthe bandwidths utilised by the three IPv4 dynamic routing protocols, RIPv2(RoutingInformationProtocol),EIGRP(Enhanced Interior Gateway Routing Protocol) and OSPFv2 (Open Shortest PathFirst)withrealtimecasestudies.Wehavealsoaddressed theissueofchoosingtheoptimalprotocolasperthenetwork specifications.Keywords- Bandwidth, Dynamic Routing Protocols, RIP, EIGRP, OSPF, VLSM I. INTRODUCTIONRoutingprotocolisthelanguagespokenbyrouterswith theirneighbouringrouterstoshareinformationaboutthe reachabilityandthestatusoftheprefixestheyknow.These protocolsdeterminethebestpathtoadestinationandnote them in their routing tables. Choosing the correct protocol that satisfiesthegivenrequirementsofanetworkisverycrucial and generally a tedious task [2].Many factors have to be considered, both about the network as well as the protocol that we would like to go with.The first questionwillbeWhichprotocolisbetter?Andtheanswer for it is Well, it depends. Now comes the question Dependsonwhat?Andanswertothisquestionisthekeypartfor designing any network.The decision parameters that help in deciding upon the right protocol to be used in the network are: Speed of convergence: Howfastistheroutingtableconvergenceattained when updates or failures happen in the network? Scalability of network: Arethereanychancesofextendingthisnetworkin thefuture?Willmoredevicesbeaddedtothis network?Orisitgoingtobebrokenintomany subnets? Utilization of resources: Whichprotocoladdslessoverheadtotherouters? Also,theProtocoldependantresourceslikeCPU, queuebuffer,memoryandbandwidthmustbedealt with Maintenance and management: Which protocol makes the network more flexible and easilymaintainable? Which protocoloffers relatively easier troubleshooting and management? In this paper, we have measured the bandwidths used insendingupdatesandhellomessagesbyvariousIPv4 dynamicroutingprotocolsinrealtimeusingReal-TimeBandwidthMonitorfromSolarWindssoftwareandhave compared them.Wehavealsoaddressedtheissueofselectingthe optimal protocol for a network to the largest possible extent by discussing various possible network requirements. II. OVERVIEW OF DYNAMIC ROUTING PROTOCOLS A. Routing Information Protocol (RIP) TheRoutingInformationProtocol(RIP)isaveteran distance-vectorroutingprotocolthatusesUDPport520for messageencapsulation[7].Itconsistsoftwomessagetypes [1]: 1.Arequestmessageisusedtoaskneighboringroutersto send an update [7].2. A response message carries the update. WhenRIPisconfiguredonarouter,itsends broadcastpacketscontainingtherequestmessageoutthe entireRIPenabledinterfacesandthenlistensforresponse messages. Routers receiving the request message respond to it bysendingtheirroutingtablesintheresponsemessage.Thisprocess continues until the network is converged.ARIProutersendsoutitsfullroutingtableinits updateoncein30seconds.Ifanynewentryisfoundinan update, theRIProuterentersitintotheroutingtablealong withthesendingroutersaddress.Itusesthehopcountasa metric for determining best paths. The maximum hop count is 15; thereby preventing routing loops in the network. This also limits the size of the network supported by it. If the hop count of an incoming routeis16, itis considered to be inaccessible or undesirable and is at an infinite distance.219 978-1-4673-0255-5/12/$31.00c 2012 IEEERIPpreventsinappropriateinformationfrom propagating throughout the network,by theuseof its features like split horizon, route poisoning and hold down timers, thus providingstabilitytothenetwork.RIPcanperformload balancing for up to six equal-cost links [1]. Versions: RIPv1: RIPv1 supports classful routing [6]; therefore variablelength subnet masks (VLSM) cannot be used. There is also no authentication mechanism. RIPv2: RIPv2 supports Classless Inter-Domain Routing (CIDR).It uses MD5 mechanism for authentication. B. EIGRP (Enhanced Interior Gateway Routing Protocol) EIGRP,abrainchildofCiscoSystems,implements thebestfeaturesofdistancevectorprotocolsaswellaslink-stateprotocolsandhenceisconsideredasahybridprotocol. EIGRPfeaturesauniquediffusingupdatealgorithm(DUAL) forcalculatingbestpathstoadestination.Itisextremely powerfulinreducingtheconvergencedelaythatwetypically have in a modern network.EIGRP also has a great quality of being very easy on CPUutilizationfordevices.Itisscalable;itdoes accommodateverylargenetworks.EIGRPfeaturesavery simpleconfiguration.Automaticsummarization is enabledby default; so EIGRP acts in a classful manner and automatically summarizesprefixes.ItalsosupportsroutingformultiplenetworkprotocolslikeIP,IPX,andAppleTalkthroughthe conceptofProtocolDependentModules(PDM),bywhich EIGRPprocessusesadifferentroutetableforeachnetwork layer protocol. Metric:Five possible components used by EIGRP in metric calculation: Bandwidth: Weakest link bandwidth in the total path Delay: Sum of the delays for the entire path Reliability Load MTU Ifwepickupreliabilityandloadformetric calculation, themetricswould be changingway too often and thiswillcauseinstabilitiesandproblemsinCPUutilization. Soitwasdecidedtouseonlybandwidthanddelay.MTUis notinvolvedinmetricformulaatallanditisjustapotential tie-breaker[9].ThesecomponentshavecorrespondingK values.Andthemetriccanbemanipulatedaccordingly, changingtheseKvalues.FortwoEIGRProuterstobecomeneighbours, these values must match. DUAL Terminologies: Feasible Distance (FD) is simply the cost between the local router and the destination prefix. Advertised Distance (AD) is the cost from next hop to the destination prefix. It is also called Reported Distance (RD). Successor is the best (lowest cost) route to the destination. Feasible Successor is the next best route to the destination. TheadvantagewithDUAListhatwhenasuccessor fails,itisimmediatelyreplacedbyafeasiblesuccessorinto theroutingtable[5].Whilechoosingafeasiblesuccessor,it hastosatisfythefeasibilityconditionNexthopmusthave AD less than current FD of the current successor [9].EIGRP Status Codes: Passive network available Active network unavailable Update network is being updated Query outstanding query waiting for ACK Reply generating a reply to a query StuckInActive(SIA)routerisqueryingabouta networkthatisunavailableandnotgettingany responsesbackandthisleadstoconvergence problem.WecanpreventthisSIAstatebyusing summarization or Stub router concept. C. OSPF (Open Shortest Path First) TheOSPFisanopenstandardprotocolthatismost popularlyused in modern networks. It is a link stateprotocol.It features theconceptofareastoprovidescalability[8]. The key factor in designing an OSPF network is the assignment of router and its links to an area(s), which is whether it has to be inputinArea0(Backbone)oranyothernon-backbonearea. We take many factors into account while making this design: How many routers can be there in an area? How many neighbours will a router have? How many areas does this router support? Whatshouldbethedesignatedrouter(DR)and backup designated router (BDR)? For choosing an area, the most significant factors that aretobeconsideredarestablenessandredundancy.Thesize ofanareamustbeoptimalsothatthisenhancesthestability [6].Because,forsomechangeinstateofalinkforaroute, each router in that area needs to re-calculate its routes and this woulddefinitelytakesupasignificantamountoftherouters CPU resources.Whenthereexistmultipleequalcostpathstothe samedestination,OSPFperformsloadsharingacrossallthe links [4]. OSPF supportsonly manual summarization and that too, only at the Area Border Routers (ABRs) and Autonomous SystemBoundaryRouters(ASBRs)[5].Hierarchicnetwork 220 2012 International Conference on Recent Advances in Computing and Software Systemsdesignandanorderedaddressassignmentschemedecidethe scalability of the network.Ifwehaveahighcapacitylinkandifthecountof prefixesissmall,thenthenewrouterscanbeadded.Each OSPFroutersendsLink-StateAdvertisements(LSA)overallitsadjacencies[1].Baseduponthewaytheroutinghasto happen, areas are classified into five types [8]. Backbone (area 0) AllowsRouterLSA,NetworkLSA,Network SummaryLSA,ASBRSummaryLSAandAS External LSA Non-backbone, non-stub AllowsRouterLSA,NetworkLSA,Network SummaryLSA,ASBRSummaryLSAandAS External LSA Stub AllowsRouterLSA,NetworkLSA,Network Summary LSA Totally Stub Allows Router LSA and Network LSA Not-so-stubby AllowsRouterLSA,NetworkLSA,Network SummaryLSA,ASBRSummaryLSAandNSSA External LSA. OSPFusesbandwidthformetriccalculation.Based uponthebandwidthofthelinkthatisbeingused,ametric valueisassigned.Thehigherthebandwidth,theloweristhe metric (cost) assigned.Forexample,foranEthernetlinkofbandwidth10 Mbps, the cost assigned would be 10. Sum of the costs for the entirepathgivesthemetricforaroute.Andthemetricfora summaryroutewouldbethebestmetricoftheindividual routes present in that summary.Based upon the information available in the topology table,eachOSPFrouterrunsSPF(ShortestpathFirst) algorithmandcalculatestheshortestpathtoeveryprefix withinthesamearea.Incaseofanychangeinthestateofa link, the OSPF router sends it in a partial update and is flooded throughout the entire network. III. REAL TIME CASE STUDIES A. Platforms used 2 Cisco 2801IntegratedServices Router, IOS 12.5T, RAM 256 MB, Flash 64 MB. 1 Cisco 2811IntegratedServices Router, IOS 12.5T, RAM 256 MB, Flash 128 MB. 1 Dell XPS Laptop. Fig: 1Experimental setup comprising of Routers Fig:2 Graphical Representation of the experimental setup shown in Fig:1 B. Software used SolarWinds Real-Time Bandwidth Monitor Edraw Max 6.0 Wehaveconfiguredthenetworksoneachrouteras showninthefigure:2,foreachroutingprotocolandhave determinedthebandwidthusageofvariousroutingprotocols throughgraphsforsendingupdates andhellopackets at fa0/1 interface(100Mbps)ofSSN2router.Wehavealsoestimated thebandwidthusagesoftheselectedprotocolsforEthernet (10Mbps),T1(1.544Mbps)andSerial(64Kbps)linksand compared them. CASE I Whenthereisanetworkoutageorwhenanew network is added to the existing one, routers inform each other bysendingupdatesandhencebandwidthisutilisedinthis process. Under such conditions, the following graphs of Time (X-axis)vs.Bandwidth(Y-axis)areplottedfortheselected protocols. RIP: 2012 International Conference on Recent Advances in Computing and Software Systems 221EIGRP: OSPF:TABLEI.COMPARISONOFBANDWIDTHSUSEDFORSENDING UPDATES ProtocolFast EthernetEthernetT1Serial RIP0.00140.0140.0932.179 EIGRP0.00160.0160.10562.475 OSPF0.0010.010.0661.547 From the table I, we can infer the following points: 1. IfthenetworkinvolvesmoreSerialorT1links,it bettertogoforOSPFprotocolasitusesrelatively lesser bandwidth in sending updates. 2. Ifthechanceoffailureismoreinthenetwork, EIGRP canbeopted for,providedthere aremultiple paths to the destination. CASE II The routers send hello messages to their neighbors to showtheirpresence.Usuallythisisnotamajorissueina networkwithlink-stateprotocolsbutaseriousissuewith distancevectorprotocolwhichsendsitscompleterouting table at every specific interval.Undersuchconditions,thebandwidthusageof variousroutingprotocolsatfa0/1interface(100Mbps)of SSN2 router are determined the following graphs of Time (X-axis)vs.Bandwidth(Y-axis).Wehavealsoestimatedthe bandwidthusagesoftheselectedprotocolsforEthernet,T1and Serial links and compared them. RIP: EIGRP:OSPF:TABLEII.COMPARISONOFBANDWIDTHSUSEDFORSENDING HELLO PACKETS ProtocolFast EthernetEthernetT1Serial RIP0.0010.010.0661.547 EIGRP0.00040.0040.02640.619 OSPF0.000370.00370.02440.572 Here,intableII,foreachhellomessagesent,RIP sends out its complete routing table creating heavy traffic flow 222 2012 International Conference on Recent Advances in Computing and Software Systemsevery30seconds[3].Thereforeinanetworkwherethe bandwidth is very limited, RIP must be avoided.IV. OPTIMAL SELECTION Thesameprotocolmaynotbethebestforallkinds ofnetworks.Italldependsonthenetworkenvironmentand business needs [2]. While choosing the optimal protocol for a network with given requirements,convergencetimeisaveryimportantfactorin decidingthecorrectprotocolforanetwork.Becausedata transmissioncannottakeplacewhiletheroutersare converging, it is averygood pluspoint for a routingprotocol to have very less convergence time. EIGRP shows sub second convergenceandnextcomesOSPFfollowedbyRIP.Incase ofEIGRPandOSPF,thebandwidthisusedinitiallyduring formationofneighborsandnetworkupdates,unlikeRIP which sends out its complete routing table as periodic updates. The scaling of network is also an important factor for limiting theentriesintheroutingtable.EIGRPusesAutonomous System(AS)andOSPFusesareasforthispurpose.UsuallyEIGRPscaleswellcomparedtoOSPF.RIPdoesnotscale well. Taking CPU resources into account, RIP and EIGRP consume lesser thanOSPF (complex SPF algorithm).And talking aboutnetworks with multiple vendor devices, RIP and OSPF will be theonlypossibleoptionsasEIGRPisCiscoproprietary.If thereisanypossibilityofrunningIPXorAppleTalkonour network, only EIGRP has the capability to route these packets [9].AllroutingsoftwareanddevicessupportRIP.Noother routingprotocolhasthisadvantage. Forexample,ifthereare some old Unix routers running in the network, then we do not have a choice but to use RIP. Whilecalculatingcostofaroute, OSPFv4andRIPtake bandwidthand hopcountasthemetricrespectively,whereas thisisnotthecaseinEIGRPwhichtakesintoaccount bandwidth, delay, reliability and load as metrics [3]. The order ofcomplexityofconfiguringRIPislesswhencomparedto EIGRPandOSPF.Verylargenetworksaresupportedby EIGRP,usingtheflexibilityofconfiguringmultipleASona single router and by OSPF using the concept of areas. V. CONCLUSION Manyfactorsplayakeyroleindecidinguponthe routing protocol to be chosen. Not all requirements can be met by a single protocol at all times. Therefore, we have suggested anoptimalselection,throughwhichchoosingthecorrect protocolbecomeseasier.Also,wehavecomparedthe bandwidthsutilisedbyvariousroutingprotocolsfortheir managementpurposes;therebymaking it easier tochoosethe right one when bandwidth is a bottle-neck factor. REFERENCES [1] Jeff Doyle, Jennifer Carroll, CCIE Professional Development Routing TCP/IP Volume 1, Second Edition,Cisco Press. [2] Poprzen, N.; Gospic, N,Scaling and convergence speed of EIGRPv4 and OSPFv2 dynamic routing protocols in huband spoke network,IEEE International Conference on Telecommunication in Modern Satellite, Cable, and Broadcasting Services, 2009.[3] W Richard Stevens, G Gabrani, TCP/IP Illustrated:TheProtocols (Volume - I) [4] The Internetworking Technologies Handbook - Conflicts,Empire And National Identity Fourth Edition, Cisco Press. [5] Todd Lammle,CCNA: Cisco Certified Network AssociateStudy Guide, Seventh Edition, Sybex Press. [6] Behrouz Forouzan , TCP/IP Protocol Suite. [7] RFC 1058, Routing Information Protocol, C. Hendrik, TheInternet Society (June 1988) [8] RFC 2328, OSPF Version 2, Moy. J, The Internet Society (April 1998). [9] Cisco Documentation on EIGRP. 2012 International Conference on Recent Advances in Computing and Software Systems 223