Inter national J our nal of Electrical and Computer Engineering (IJECE) V ol. 6, No. 1, February 2016, pp. 413 420 ISSN: 2088-8708 413       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     Study of BGP Con v er gence T ime Rohit Nilkanth De vikar * , D . V . P atil ** , and V . Chandraprakash * * Department of Computer Science and Engineering, K L Uni v ersity , Green Fields, V addesw aram, Guntur , (A.P .) 522502, India ** Department of Computer Engineering, Gokhale Education Society’ s R. H. Sapat Colle ge of Engineering, Management Studies and Research, Nashik - 422 005, (M.S.), India Article Inf o Article history: Recei v ed May 18, 2015 Re vised Sep 8, 2015 Accepted Oct 2, 2015 K eyw ord: BGP BGP policies Con v er gence T ime Instability Link F ailure ABSTRA CT Border Gate w ay Protocol (BGP), a path v ector routing protocol, is a widespread e xterior g ate w ay protocol (EGP) in the internet. Extensi v e deplo yment of the ne w technologies in internet, protocols need to ha v e continuous impro v ements in its beha vior and operations. Ne w routing technologies conserv e a top le v el of service a v ailability . Hence, due to topo- logical changes, BGP needs to achie v e a f ast netw ork con v er gence. No w a days size of the netw ork gro wing v ery rapidly . T o maintain the high scalability in the netw ork BGP needs to a v oid instability . The instability and f ailures may cause the netw ork into an unstable state, which significantly increases the netw ork con v er gence time. This paper summarizes the v arious approa ches lik e BGP policies, instability , and f ault detection etc. to impro v e the con v er gence time of BGP . Copyright c 2016 Institute of Advanced Engineering and Science . All rights r eserved. Corresponding A uthor: Rohit Nilkanth De vikar Research Scholar Departement of Computer Science and Engineering K L Uni v ersity , V addeshw aram, Guntur +919028339491 rohit.de vikar89@gmail.com 1. INTR ODUCTION Border Gate w ay Protocol (BGP) is interdomain routing protocol. Such protocol of fers routing functional- ity between autonomous system (AS). Earlier the objecti v e of BGP w as: 1) T o pro vide scalable and rob ust routing functionality , and 2) T ime required for the netw ork to reco v er from the f ailure. In this paper , we analyze the BGP problems and identify v arious algorithms that impro v es the con v er gence time signi ficantly . Labo vitz et al. [1], [2] noticed that sometimes BGP tak es a substantial amount of t imes and messages to con v er ge, and stabilize the f ailure of some node in the internet. The Y ehuda Afek et al. [3] has gi v en a minor modification to BGP , that eliminates the problem pointed out and substantially reduced the con v er gence time and communication comple xity . An important parameter in the BGP con v er gence time is minimum route adv ertisement interv al. Basically it is a amount of time BGP enforce between the sending of consecuti v e announcement from routers to its neighbors. Grif fin and Brian [4] sho ws that for each specific netw ork topology there is an optimal v alue of minimum route adv ertisement interv al (MRAI) that minimizes the con v er gence time. The MRAI v alue proposed in this approach changing from netw ork to netw ork that cant be ef ficient to impro v e the BGP performance. A ne w solution to reduce the con v er gence time comple xity w as introduced in [5]. The y use the information pro vided in ASP ath to define route consistenc y assertion and use this assertion to identify infeasible routes. Ho we v er t his technique requires e xtra computational resources for checking router consistenc y and to send e xtra information in the BGP messages. It also introduced dif ficulti es in some cases, when AS partitions and some routers in the AS become disconnected from other routers in the same AS. The Y ehuda Afek et al. [3] proposed a ghost flushing solution to reduce the problem of con v er gence time. In the netw ork sometimes incorrect information are forw arding for a long duration of time. This information is nothing b ut the ghost information. Such information disturbs the con v er gence of routers in case of bo t h f ail do wn and f ail o v er mechanism [1]. The ghost information is outdated for netw ork con v er gence that will enter the netw orks into unstable state. T o impro v e this problem [3] modify BGP by introducing ghost flushing rate and ghost flushing rule. J ournal Homepage: http://iaesjournal.com/online/inde x.php/IJECE       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     DOI:  10.11591/ijece.v6i1.8106 Evaluation Warning : The document was created with Spire.PDF for Python.
414 ISSN: 2088-8708 1.1. Moti v ation A long con v er gence time has been a critical problem for an y service pro vider or netw ork pro vider , because it af fects the data transmission process for a considerably longer period and till that interv al netw ork remains una v ailable. The aim of this study is to impro v e the a v ailability of netw ork and to reduce the con v er gence time in inter -domain routing. 1.2. Route Cause Notification The path v ector protocol does not ha v e periodic updat es/adv ertisements. The update message can be triggered only when changes happened in the connecti vity . These changes in the link connecti vity will be detected by routers which is adjacent to that link. F or an y gi v en destination, one of the node from tw o adjacent node may change its route, we call this node as root cause node. The root cause node attaches its ID to the update message, which will then propag ate in to the netw ork. Unlik e to flooding (in link state routing), the Simple P ath V ector Protocol-RCN (SPVP-RCN) [6] piggyback the root cause notificati on in the updates, due to this only direct neighbors and af fected nodes are notified [7], [8]. 1.3. Routing P erf ormance The ghost flushing [3], RCN [9], [6], and FECN [10] analysis uses U delay model. The limitation of U model is that for all nodes there is same netw ork-wide fix upper bound h(G,[v   u]). But in the netw ork topology each node may ha v e dif ferent upper bound h(G,[v   u]). The U model gi v es rough estimation of con v er gence ti me, also it f ails to sho w the relationship between netw ork topology and con v er gence time. Dan Pei et al. [11] proposed Q model which combines a queuing delay estimated into h(G,[v   u]) and better reflect BGP implementation. F or calculating queuing delay the y consider sum of transmission and propag ation delay on an y link (ld), maximum message processing time PMax, and summation of ld, queuing delay , and message processing time, when message is propag ated from u to v denoted as h(G,[v   u]). Inter -domain routing consists of problems lik e performance and route aggre g ation. Man y people use geographic information for routing and addressing mechanism. Researchers ha v e focused on reducing the geographical length of selected path by routing mechanism t o impro v e the routing performance. The T ao yu Lia et al. [12] suggested that abo v e technique does not used to impro v e the actual end to end transmission performance; rather the y de v eloped a performance model based on transmission delay . The transmission delay consists of both propag ation delay and queuing delay . By e xperimental result the y ha v e sho wn the impro v ement in the performance of routing mechanism up to 50 % by actual pack et deli v ery mechanism. The f ast pack et deli v ery can be achie v ed by reducing the con v er gence time required to update the routing tables. 2. BGP POLICIES The BGP allo ws an AS to apply dif ferent local policies for selecting route and propag ating reachability information to another domain. But autonomous systems ha v e conflicting policies that le ads to instability in routing. Sometimes routing oscillation reduces the performance of netw ork in terms of quality of service (QoS). Up till no w man y modification ha v e been made on BGP protocol that dynamically notice and solv e polic y-induced oscillation [13], [14]. In internet each AS ha v e there routing policies for pack et transmissi o n [15]. As a result, if an y polic y destruction occurs at intermediate AS, causes pack et dropping before reaching to the destination. BGP solv e the problem of pack et dropping. BGP w orks on the principle of hop by hop transmission, resulting in some routes are unreachable e v en though there is a ph ysical path a v ailable to reach destination. T o o v ercome this problem Jyh-ha w Y eh et al. [16] ha v e proposed a source polic y route disco v ery protocol, which will resolv e the f alse ne g ati v e unreachable destination in BGP . The B. Quoitin et al. [17] allo ws the internet service pro viders to control the incoming traf fic flo w by proposing the utilization of redistrib ution communities. This has been done by cont rolling the distrib ution of routes adv ertisement with the peers. L. Xiao et al. [18] systematically studied the lifetime of BGP session under certain netw ork congestion using statistical and simulation methods, which can be caused by w orm attacks or by traf fic engineering f ailure. Among independent ASs, when an y changes happened in inter -domain routing, there is a need of on-demand routing adjustment. T o resolv e this problem Osamu Akashi et al. [19] ha v e proposed a virtual router (VR) technique, which controls the con v entional BGP routers from e xterior w orld without an y protocol e xtension. The Huaming Guo et al. [20] fill the g aps and analyze the im p a ct of routing policies on con v er gence condition and con v er gence time including MED attrib utes. T o represent routing policies in BGP including MED [20] first introduce a timeless model, later on the y e xtended it to real time model by adding edge delay . The y also deri v e a suf ficient condition on the routing policies for rob ust con v er gence and an upper bound on con v er gence time. Martin O. Nicholes IJECE V ol. 6, No. 1, February 2016: 413 420 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 415 et al. [21] e v aluated the ef fecti v eness of Inter Domain Collaborati v e Routing (IDCR) using de gree algorithm, greedy algorithm, and f actor algorithm, which allo ws a friend routes to determine routes collaborati v ely . 2.1. Routing T ools Lixin Gao et al. [22] proposed detail BGP model and a set of guidelines for AS to follo w in setting its routing policies without considering global co-ordination among autonomous systems to impro v e the stability in the internet. Nic k Feamster et al. [23] de v eloped a tool called router c o nfi gu r ation check er (RCC), which identifies f ault in BGP configuration. RCC enables netw ork administrator to test a n d deb ug configurations before deplo ying them in the netw ork. The authors ha v e analyzed the configuration on 17 dif ferent ASs to detect v ariety of f aults which in turn used to impro v e internet routing infrastructure. F or designing a stable BGP protocol Grifn et al. [24] ha v e proposed a formal tool, b ut it f ails to pro vide guaranteed service continuity when deplo ying an y changes to BGP . T o o v ercome this problem Luca Cittadini et al. [25] proposed a Greedy+ algorithm (impro v e traditional Greedy algorithm) which pro vides correct reports for stability of a netw ork, also used to spotting the f ault pints in the oscillated path, and checks the con v er gence of BGP in an abstract model. BGP configuration f aults causes pack ets loss and forw arding loops that corresponds to f ailure in the netw ork infrastructure. BGP configuration f aults causes pack ets loss and forw arding loops that corresponds to f ailure in the netw ork infrastructure. 2.2. F ault Detection in BGP Distance v ector routing protocol ha v e slo w con v er gence problem. In distance v ector , each router maintains its routing table that contains information about reachability to destination. Due to changes in the topology distance v ector tak es longer time to con v er ge information, which were introduced count to infinity problem in the netw ork [26]. The solution to abo v e problem is the introduction of BGP , which pro vides a path v ector approach contains entire path to reach destination. Another attempts to o v ercome count to infinity problem includes split horizon, dif fusion update, and trigger update algorithm. The Craig Labo vitz et al. [1] analyze the impact and the rate at which inter - domain routing repairs and f ailures, adv ertise this information through the internet. The y proposed a no v el approach to impro v e the con v er gence, b ut changes increases the router o v erhead and comple xity . Pei et al. [9] impro v es the BGP con v er gence time by identifying a f ault location and indicating all the routers which are in the path of f ault zone to a v oid the incoming updates, due to this other routers not using that path for future transmission. 2.3. Hot P otato Routing Renata T eix eira et al. [27] proposed a Hot Potato routing technique w orks on the basis of link weights and link f ailure. The figure 1 sho ws router A will choose the e gress router C to tra v el the traf fic to dif ferent ASs. Suppose distance between A ! C changes from 9 to 11 intentionally or link between A ! C f ailed due to some interruption. Although the distance between A ! C changes, still there is a path between A ! C is a v ailable b ut has a lar ge distance. A chooses the path A ! B to forw ard the traf fic. This routing which changes the path dynamically called hot potato routing. The routing in the netw ork is fle xible and visible to all neighbors in the netw ork, which is ef ficient to impro v e the netw ork con v er gence. But hot potato technique has the chances of pack ets loss due to slo w con v er gence of BGP . The Alejandro Ruiz-Ri v era et al. [28] ha v e proposed a green netw orking technique in addition to hot potato called HO TPLEC that shutdo wn the least utilized links or routers during of f peak hours. Due to shutdo wn of unutilized links or routers reduces the ener gy consumption of netw ork without ne g ati v e impact on BGP . Figure 1. Hot potato routing changes from C to B Study of BGP Con ver g ence T ime (R. N. De vikar) Evaluation Warning : The document was created with Spire.PDF for Python.
416 ISSN: 2088-8708 2.4. D-BGP During route con v er gence, the transient routing f ailure, losses the end to end reachability of the internet path. Also, this f ailure causes the pack et losses in the netw ork which will create problems on v oice o v er IP pack et trans- mission. T o reduce this Feng W ang et al. [29] ha v e studied transient routing f ai lure during changes happened in the routing (such as f ailure and reco v ery in BGP system) by applying routing policies. Due to this netw ork administrator can impro v e the performance and stability of the netw ork. Also the y de v eloped T w o path di v ersity a w are routing protocols [30] D-BGP and B-BGP to impro v es the resilience of inter -domain routing. These protocols established multiple paths with lo w routing b urdens by e xploiting e xistence of path di v ersity in the netw ork infrastructure. Y i W ang et al. [31] on the basis of neighbor routers apply the filtering policies to the BGP routers, to impro v e the f ast transmission. Due to this routers select routes dynamically as per neighbor routers a v ailability . The main problem with DBGP [30] is it increases path di v ersity by adv ertising multiple paths. If route f ailure occ urs, D-BGP selects alternate path without considering its quality . T o o v ercome this problem [31] proposed a technique, which established shortest path using D-BGP , b ut it selects alternate path based on link a v ailability and bandwidth. This technique increases f ault tolerance and reduce message o v erheads and updates. Chaitan ya et al. [32] ha v e proposed a t echnique that will pro vides BGP short est path and OSPF lo west cost metric for mobile ad-hoc netw orks. It increases the routing table en- tries, b ut transmits the traf fic to destination with lo west cost. Mohammad Y anuar Hariya w an [33] compared dif ferent techniques lik e F ast Reroute one to one backs up, local rerouting, Haskin, 1+1 path protection reco v ery mechanism and PSL oriented path protection mechanism t echnique for f ast rerouting after f ailure. The performance sho ws that 1+1 path protection reco v ery mechanism has minimum pack et loss, b ut ha ving more cost. The Chengchen Hu et al. [34] pro vides a technique for reco v ery after f ailure. By using AS-le v el complete map and IXP database, the y measure the path di v ersity , reco v ery ratio and shifted the path in dif ferent f ailure scenarios. The Stef ano V issicchio et al. [35] sho ws that current system do not pro vides the guarantees for BGP reconfiguration with respecti v e traf fic disr up t ions and also for guaranteed pack et loss. The [35] proposed a BGP frame w ork that runs tw o separate BGP control plane in parallel, to enable the los sless reconfiguration. The first control plane stores initial configuration of routing table and second control plane store final configuration of routing table. The traf fic forw arding has been taking place on the basis of final RIB, which reduces the pack er loss. The a v ailability of wireless mesh netw ork is impro v e by introducing a medQoS routing protocol [36]. The main objecti v e of this protocol is to minimize the routing changes and reduce the o v erall o v erhead introduced by traditional routing protocols. The Miao Xue et al. [37] ha v e proposed a technique called source directed path di v ersity using which, sources can gi v es alternate paths to forw ard the traf fic. In pack et header , sources specify the tag called as source directed tag (SDT) that informs BGP routers for path selection. BGP routers on t he basis of Source indication, forw ard the traf fic independently on the indicated path. In order to address the link f ailure between autonomous systems LI Chun-xiu et al. [38] has proposed a f ast reroute scheme by incorporating Softw are Defined Netw orking (SDN) with BGP called softw are defined autonomous system le v el f ast rerouting (SD-FRR). By considering routing policies SD-FRR aims to pro vide polic y compliant path to protect forw arding of data locally , which a v oids pack et losses and ef ficiently impro v es the netw ork a v ailability . The abo v e section described the detection of f ault in the netw ork. Detection of f aults in early stage reduces the pack et loss, which will impro v e the performance of netw ork. 3. INST ABILITY IN INTER-DOMAIN R OUTING The oscillation in the internet causes w astage of bandwidth, due to e xtra and unnecessary route hops. T o reduce the oscillation in the BGP , V i vian Elliott et al. [15] used the e xplicit withdra w als technique. This technique will reduce the o v erall transaction traf fic and path length. Route oscillation and path e xploration reduces the performance of pack et forw arding that will increase BGP instability . The e xisting solution cannot ef ficiently solv e the BGP instability problem. The accidental acti vities such as f ailure, misconfiguration, route flapping, induced se v eral BGP instabilities in the netw ork lead to delays, loss of data and connecti vity . Route flap damping (RFD) is considered to be a good approach that stabilizes the internet routing. But has a problem to wrongly suppress relati v ely stable routes for a longer duration of time [39], [40]. This technique introduced a comple x interaction between BGP path e xploration and ho w RFD algorithm finds route flaps. Sahoo et al. [41] consider the group processing update technique called as batch processing updates. When an y update comes, router e xtracts the destination address from it and queue the update correctly , results in reducing the number of updates in the netw ork. In this technique authors maintain a separate logical queue for each and e v ery destination. The processing of all t he updates tak es place on the basis of destination. This technique is ef ficient for reducing updates. The Zhenhai Duan et al. [42] identify dif ferent BGP path e xploration characteristics that follo w e v ents such as links f ailure or rout es f ailure. The approach gi v en has useful for distinguishing BGP route updates from route flapping at the time of BGP path e xploration. The authors ha v e de v eloped a RFD+ algorithm that impro v es the stability of the inte rnet routing. The main objecti v e behind this IJECE V ol. 6, No. 1, February 2016: 413 420 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 417 approach is that, without af fecting occasionally f ail routes, i t can correctly suppress persistent route flaps. Shi v ani deshpande et al. [43] proposed the BGP instability detection mechanism that can be e x ecuted by indi vidual routers. The input data for detection of instability is BGP update messages recei v ed by routers from its neighbor . From this BGP update messages features (lik e AS path length, AS path edit distance) are e xtracted in e v ery v e minutes, this sho ws the change in topology . The GLR (Generalized Lik elihood Ratio test), Se gmentation boundary detection, Boundary position optimization algorithms are used to detect the changes. T odays internet routers are o v ercome by a number of BGP updates caused by e v ents such as f ailure, session reset, and polic y changes. Such e v ents can delay routing con v er gence, which de grades the performance of netw orks in terms of jitter and delay sensiti v e application. 3.1. Dynamic Routing Decision T o analyze, predict and troubleshoot the beha vior of netw ork the ISPs ha v e collects routing data. But this collected data is not complete and dif fic u l t to analyze manually . The Ashle y Fla v el et al. [44] combined the pieces of collected data to obtain a more complete vie w of netw ork state. Also the y ha v e presented a technique for real time scenario, which dynamically determines the routing decision for all routers in the autonomous system. The Qi Li et al. [45] proposed a “stableBGP” that e xperimentally solv e the BGP instability problem including path e xploration and route oscillation. The “stableBGP” quickly stabilize the route selection problem by addressing the causes of route changes. The Zhang Jun et al. [46] proposed a no v el approach for quickly checking border g ate w ay protocol route oscillation. The route update chain tag (R UCT) has been b uild to track the forw arding of update report and local routing library is used to record the change history of update report. The route oscill ation can be found out by analyzing R UCT and local routing library . F or checking oscillation time more ef ficiently , authors compares R UCT approach with relati v e preference (RP) approach and T ok en based approach. The e xperimental results sho w that R UCT needs lesser time to check route oscillation than RP and T ok en based approach. The D.P apadimitriou et al. [47] pro vides a stability metrics for stability of indi vidual routes, stability compu- tation for set of routing entries, most stable routes, and for best selected routes that described the stability properties of path v ector protocol. Also the y e xamine the ef fect of routing policies and instability on local routers. The duplicate announcements are the major BGP churn contrib utor analyzed by Ahmed Elmokshi et al. [48] for BGP up da tes. Jian Jiang et al. [49] proposed a v erification of routing path mechanism to detect path inconsistenc y . In this technique sender and recei v er autonomous systems generate routing e vidence and communicate with each other to v erify path., which is used to detect inconsiste n c y in announced path. Instability in the netw ork results in loss of pack ets, which in turn increases the latenc y and con v er gence time. Abo v e section described the research w ork done by dif ferent researchers to reduce the con v er gence time by stabilizing the netw ork. 4. REDUCING NUMBER OF UPD A TES W ei Sun et al. [50] propose the no v el approach of dif ferentiated processing in terms of BGP updates, which impro v e the routing con v er gence and reduces the routers load. Based on this approach the BGP updates are classi- fied into tw o classes. Higher priority updates are processed sooner , while the lo wer priority updates are delayed to reduce router load and processing. The con v er gence delay can be increase by increasing the f ailure in the internet. So for multiple f ailures, results in more con v er gence delay in the netw ork. The Amit Sahoo et al. [51] presented a dynamic scheme that selects optimal MRAI v alue based on the size of the b uf fer messages at router , which reduces con v er gence delay for lar ge netw ork f ailure and k eeping lo w v alue of delay for small f ailure. The y also e xamined the batch processing scheme, which reduces the generation of in v alid adv ertisement. These tw o techniques are designed to impro v e con v er gence delay in the netw ork. Geof f Huston et al. [8] proposed a P ath Exploration Damping (PED) technique which reduces the v olume of BGP update messages and decreases the a v erage time required to restore reach- ability . The y compare PED impact on con v er gence time with Mean route adv ertisement interv al (MRAI), Route Flap Damping (RFD), and W ithdra w al Rate Limiting (WRA TE). From e xperimental results it w as found that the total BGP announcement can decrease by up to 32%, and path e xploration reduced by 77% compared with traditional MRAI approach. Rajvir Gill et al. [52] proposed the FLD-MRAI (Fle xible Load Dispersing MRAI) algorithm that disperses the load in the netw ork, which results in reducing the routers o v erhead. The authors focused on routing policies and their ef fects on number of updates, con v er gence time. The FLD-MRAI algorithm w orks in case of both high and normal loads. When de gree of preference (DoP) chooses the shortest path, then FLD-MRAI belie v e this situation as normal load, and when DoP chooses the longest path then FLD-MRAI belie v e this situation as high load. The Mahesh K umar and Shishir K umar [53] proposed a technique that can k eep the v alue of the mean route adv ertisement interv al (MRAI) t imer v ariable instead of k eeping it const ant [26]. The MRAI v alue depends upon the present netw ork condi- tion, due to this con v er gence time become relati v ely lo w and updating of netw ork significantly f aster . Andre y Sape gin et al. [54] ha v e analyzed a BGP updates from multiple observ ation points. The y de v eloped a method that classifies Study of BGP Con ver g ence T ime (R. N. De vikar) Evaluation Warning : The document was created with Spire.PDF for Python.
418 ISSN: 2088-8708 BGP updates into correlated or non-correlated updates. F or forw arding and filtering of pack ets router requires lookup functionality . No w ada y s there are serious challenges for update performance, memory ef ficienc y and throughput. The Y anbiao Li et al. [55] presented a ne w parallel lookup model called split routing lookup model rather than looking for optimization techniques for traditional lookup model. In this model all the prefix es are split to produce redundancies, after that the y are remo v ed during information inte gration. The splitting of prefix es reduces routing updates also this model use for parallel processing for lookup address. The abo v e section discussed about reducing the routing updates, due to this congestion in the netw ork will reduced, which in turns impro v e the con v er gence time. T able 1. describes the comparison of v ari ous scenarios in terms of BGP policies, f ault detection, reducing the instability , and reducing the number of updates in the netw ork. T able 1. A Comparison of V arious Scenarios Scenario BGP Policies F ault Detection Reduce Instability Reduce Number of Updates Labo vitz et al. [1], [2] Y es No Y es No T ao yu Lia et al. [12] Y es No No Y es Dan Pei et al. [11] yes No No Y es Jyh-ha w Y eh et al. [16] Y es No No Y es Huaming Guo et al. [20] Y es No No No Lixin Gao et al. [22] Y es No Y es No Nick Feamster et al. [23] No Y es No Y es Renata T eix eira et al. [27] No Y es No Y es Feng W ang et al. [29] Y es Y es Y es No Y i W ang et al. [31] No Y es No Y es Miao Xue et al. [36] Y es Y es No No Chun-xiu et al. [37] Y es Y es No No V i vian Elliott et al. [15] Y es No Y es No Sahoo et al. [41] No No No Y es Zhenhai Duan et al. [42] No Y es Y es No Shi v ani deshpande et al. [43] Y es No Y es No Ashle y Fla v el et al. [44] No No Y es No Qi Li et al. [45] Y es Y es Y es Y es Zhang Jun et al. [46] Y es No No Y es D.P apadimitriou et al. [47] Y es No Y es No Jian Jiang et al. [49] Y es No Y es No W ei Sun et al. [50] No No No Y es Amit Sahoo et al. [51] No Y es Y es No Geof f Huston et al. [8] No No No Y es Rajvir Gill et al. [52] Y es No No Y es Mahesh K umar[53] No Y es No Y es Andre y Sape gin et al. [54] No No Y es Y es Y anbiao Li et al. [55] No No No Y es 5. CONCLUSION AND FUTURE DIRECTION This study sheds light on the ef fect of continuously increasing con v er gence time. From the surv e y we found that, BGP con v er gence time increases rapidly with the de gree of f ailure before le v eling of f and going do wn. This means that multiple f ailures can lead to considerable longer periods of instability as compared to single f ailures. In this paper , we surv e yed current ef forts to enhance the con v er gence speed of the BGP protocol and eliminate the duplicate adv ertisements in its operation to impro v e its stability . The needs for f ast con v er gence and stability in path v ector routing protocols continue to challenge for the researchers as the routing domains gro w lar ger and more comple x. In future we will focus on the scalability and security related issues and its impact on BGP con v er gence time. REFERENCES [1] C. Labo vitz, R. W attenhofer , S. V enkatachary , and A. Ahuja, The impact of Internet polic y and topology on delayed routing con v er gence, In Proc. INFOCOM ,Apr . 2001: 537-546. IJECE V ol. 6, No. 1, February 2016: 413 420 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 419 [2] C. Labo vitz, A. Ahuja, A. Bose, and F . Jahanianitz, Delayed Internet routing con v er gence, In Proc. SIGCOMM , Sept. 2000: 175-187. [3] Y ehuda Afek, Anat Bremler -Barr , and Shemer Schw arz , Impro v ed BGP Con v er gence via Ghost Flushing, IEEE Journal on Selected Areas in Communications , December 2004; 22(10): 1933-1948. [4] T imoth y G. Grif fin and Brian J. Premore, An e xperimental analysis of BGP con v er gence time, In Proc. ICNP , No v . 2001: 5361. [5] D. Pei, X. Zhao, L. W ang, D. Masse y , A. Mankin, S. F . W u, and L. Zhang, Impro ving BGP con v er gence through consistenc y assertions, In Proc. IEEE INFOCOM , 2002: 902-911. [6] Dan Pei, Matt Azuma, Dan Masse y , Lixia Zhang, BGP-RCN: Impro ving BGP con v er gence through root cause notication, Computer Netw orks , Y ear 2005; 48(2): 175-194. [7] Y ang Richard Y ang, Haiyong Xie, Hao W ang, A vi Silberschatz, Arvind Kri shnamurth y , Y anbin Liu, Li Erran Li, On Route Selection for Inter -domain T raf fic Engineering, IEEE Netw ork , No v ember/December 2005: 20-27. [8] Geof f Huston, Mattia Rossi, and Gren ville Armitage, A T echnique for Reducing BGP Update Announcements through P ath Exploration Damping”, IEEE Journal on Selected Areas in Communications , October 2010; 28(8): 1271-1286. [9] Pei, D., Zhang, B., et al., An analysis of path-v ector routing protocol con v er gence algorithms. Computer Netw orks , (2005), http://www .cs.arizona.edu/ bzhang/paper/con v er gence.pdf. [10] www .ieee802.or g/1/files/public/docs2007/au-jain-fecn-20070124.pdf [11] Dan Pei, Beichuan Zhang, Daniel Masse y , Lixia Zhang, An analysis of con v er gence delay in path v ector routing protocols, Computer Netw orks: The International Journal of Computer and T elecommunications Netw orking , 22 Febraury 2006; 50(3): 398 - 421. [12] T ao yu Lia, Y uncheng Zhu, K eXu, Maok e Chen, Performance model and e v aluation on geographic -based routing, Computer Communications ,2009; 32(2): 343-348. [13] T . G. Grifn and G. T . W ilfong, A safe path v ec tor protocol, INFOCOM 2000. Nineteenth Annual Joint Confer - ence of the IEEE Computer and Communications Societies , 2000; 2: 490-499. [14] C. T . Ee, V . Ramachandran, B.-G. Chun, K. Lakshminarayanan, and S. Shenk er , Resolving inter -domain polic y disputes, in Proc. SIG-COMM , 2007. [15] V i vian Elliott, K enneth J. Christensen, Charact erizing and reducing route oscillations in the Internet, Else vier Science, Computer Communications ,2003; 26(2): 143-153. [16] Jyh-ha w Y eh a, W ei Zhang a, W en-chen Hu b, a nd Chung-wei Lee, Design and simulation of a supplemental protocol for BGP , Science Direct, Else vier , Computer Netw orks ,March 2005; 49(2): 172-200. [17] B. Quoitina, S. T andela, S. Uhligb, O. Bona v enture, Interdomain trafc engineering with redistrib ution communi- ties, Computer Communications , 2004; 27(4): 355-363. [18] Li Xiao, Guanghui He, Klara Nahrstedt, BGP session lifetime modeling in congested netw orks, Science Direct, Else vier , Computer Netw ork , 2006; 50(17): 3315-3333. [19] Osamu Akashia, K ensuk e Fukud, T oshio Hirotsu, T oshiharu Sug a w ara, Polic y-based BGP-control architecture for inter -AS routing adjustment, Computer Communications , 2008; 31(13): 2996-3002. [20] Huaming Guo, W ei Su, Hongk e Zhang, Sy-Y en K uo, On the con v er gence condition and con v er gence time of BGP , Computer Communications , 2011; 34(2): 192-199. [21] Martin O. Nichole s, Chen-Nee Chuah, S. Felix W u, Bisw anath Mukherjee, Inter -domain collaborati v e routing (IDCR): Serv er selection for optimal client performance, Computer Communications , 2011; 34(15): 1798-1809. [22] Lixin Gao and Jennifer Re xford, Stable Internet Routing W ithout Global Co-ordination, IEEE/A CM T ransaction on Netw orking , December 2001; 9(6): 681-692. [23] Nick Feamster and Hari Balakrishnan, Detect BGP Configuration F aults with Static Analysis, Proceeding NSDI05 Proceedings of the 2nd conference on Symposium on Netw ork ed Systems Design and Implementation , 2005; 2: 43-56. [24] T . G. Grifn and J. L. Sobrinho, “Metarouting”, in Proc. SIGCOMM , 2005. [25] Luca Cittadini, Massimo Rimondini, Stef ano V issicchio, Matteo Corea, and Giuseppe Di Battista, From Theory to Practice: Efciently Checking BGP Congurations for Guaranteed Con v er gence, IEEE T ransactions on Netw ork and Service Management , December 2011; 8(4): 387 - 400. [26] R. Perlman, Interconnections, 2nd ed. Reading,MA: Addison-W esle y , 1999. [27] Renata T eix eira , Aman Shaikh , T im Grif fin , Jennifer Re xford, Damping BGP route flaps, Journal IEEE/A CM T ransactions on Netw orking , December 2008; 16(6): 1295-1307. [28] Alejandro Ruiz-Ri v era, Kw an-W u Chin, Sieteng Soh, A no v el frame w ork to mitig ate t h e ne g ati v e impacts of green techniques on BGP , Journal of Netw ork and Computer Applications , 2015; 48: 22-34. [29] W ang Feng, Qiu Jian, Gao Lixin, W ang Ji a, On understanding transient interdomain routing f ailures, IEEE/A CM Study of BGP Con ver g ence T ime (R. N. De vikar) Evaluation Warning : The document was created with Spire.PDF for Python.
420 ISSN: 2088-8708 T rans Netw ork , 2009; 17(3): 740-751. [30] W ang Feng, Gao Lixin, P ath di v ersity a w are interdomain routing, In: IEEE INFOCOM , 2009: 307315. [31] W ang Y i, Schapira Michael, Re xford Jennifer , Neighbor -specic BGP: more e xible routing policies while impro v- ing global stability , In: METRICS measurement and modeling of computer systems , 2009: 217-28. [32] Chaitan ya K Krishna, Kishore Ch Ra vi, An approach to shortest path technique for BGP using OSPF , Interna- tional Journal Computer Science and Information T echnology , 2010; 1(5): 389-391. [33] Mohammad Y anuar Hariya w an, Comparison An a lysis of Reco v ery Mechanism at MPLS Netw ork, International Journal of Electrical and Computer Engineering(IJECE) , December 2011; 1(2): 151-160. [34] Chengchen Hu, Kai Chen, Y an Chen, Bin Liu, Athanasios V . V asilak os, A Measurement Study onPotential Inter -Domain Routing Di v ersity , IEEE TRANSA CTIONS ON NETW ORK AND SER VICE MAN A GEMENT , SEPTEMBER 2012; 9(3): 268-278. [35] Stef ano V issicchio, Laurent V anbe v er , Cristel Pelsser , Luca Cittadini, Pierre Francois, and Oli vier Bona v en- ture, Impro ving Netw ork Agility W ith Seamless BGP Recongurations, IEEE/A CM TRANSA CTIONS ON NET - W ORKING , JUNE 2013; 21(3): 990-1002. [36] Muhammad Haikal Satria, Jasmy bin Y unus, Ek o Supriyanto, 802.11s QoS Routing for T el emedicine Service, International Journal of Electrical and Computer Engineering (IJECE) , April 2014; 4(2): 265-277. [37] Miao Xue, Gang Fu, Ruitao Ma, and Longshe Huo, Source-Directed P ath Di v ersity in the Interdomain Routing, Journal of Netw orks , December 2014; 9(12):3251-3262. [38] LI Chun-xiu, LI Xin, LI K e, ZHANG Hong, SHI Y u-long, CHEN Shan-zhi, T o w ard softw are defined AS-le v el f ast rerouting, The Journal of China Uni v ersities of Posts and T elecommunications , December 2014; 21(6): 100- 108. [39] Zhuoqing Morle y Mao, Ramesh Go vindan, Geor ge V ar ghese, Randy H. Katz, Route Flap Damping Exacerbates Internet Routing Con v er gence, Proceedings of the 2002 conference on Applications, technologies, architectures, and protocols for computer communications ,2002; 221-233. [40] http://en.wikipedia.or g/wiki/Route flapping. [41] Sahoo A., Kant K., and Mohapatra P ., Impro ving BGP con v er gence delay for lar ge-scale f ailures, In Proceedings of the international conference on dependable systems and netw orks , 2006; 323-332. [42] Zhenhai Duan, Chandrashekar , J., Krask y , J. ,K uai Xu, Damping BGP Route Flaps, Journal of IEEE Communi- cation and Netw orks , December 2007; 9(4): 490-498. [43] Shi v ani Deshpande, Marina Thottan, T i nKamHo, and Biplab Sikdar , An Onli ne Mechanism for BGP Instability Detection and Analysis, IEEE T ransactions on Computers , No v ember 2009; 58(11): 1470-1484. [44] Ashle y Fla v el, Jeremy McMahon, Aman Shaikh, Matthe w Roughan, Nigel Bean, BGP route prediction within ISPs, Computer Communications , 2010; 33(10): 1180-1190. [45] Qi Li, Mingwei Xua, JianpingW u, P atrick P .C. Lee, Dah Ming Chiu, T o w ard a practical approach for BGP stability with root cause check, Journal of P arallel Distrib uted Computing , 2011; 71(8): 1098-1110. [46] ZHANG Jun, HU Ziying, ZHANG T ao, Update Chain-based Approach for Checking Route Oscillation of BGP , Chinese Journal of Aeronautics , 2011; 24(2): 202-209. [47] D.P apadimitriou and A.Cabellos, Analysis of P ath-v ector Routing Stabilit y , Pro vided by T echnical Uni v ersity of Catalania , April 2012. [48] Ahmed Elmokash, Amund Kv albein, and Constantine Do vrolis, BGP Churn Ev olution: A Perspecti v e from the Core, IEEE/A CM TRANSA CTIONS ON NETW ORKING , APRIL 2012; 20(2): 571-583. [49] Jian Jiang, W ei Li, Junzhou Luo, Jing T an, A netw ork accountability based v erication mechanism for detecting inter -domain routing path inconsistenc y , Journal of Netw ork and Computer Applications , 2013; 36(6): 1671-1683. [50] W ei Sun, Zhuoqing Morle y Mao, Kang G. Shin, Dif ferentiated BGP Update Processing for Impro v ed Routing Con v er gence, IEEE Conference Publication, ICNP Netw ork Protocol , 12-15 No v ember 2006; 280-289. [51] Amit Sahoo , Krishna Kant, Prasant Mohapatra, BGP con v er gence delay after multiple simultaneous router f ailures: Characterization and solutions, Computer Communications , 2009; 32(7-10): 1207-1218. [52] Rajvir Gill, Ra vinder P aul, and Ljiljana T rajk o vic, Ef fect of MRAI T imers and Routing Policies on BGP Con- v er gence T imes, In proc. IPCCC, IEEE 31st International Conference , 1-3 December 2012; 314 - 323. [53] Mahesh K umar , Shishir K umar , Analyzing and Impro ving Netw ork A v ailability in Inter -domain Routing, Springer Sci ence+Business Media Ne w Y ork 2013, W ireless Pers Communication , 27 March 2013; 72(4): 2069- 2080. [54] Andre y Sape gin, Ste v e Uhlig, On the e xtent of correlation in BGP updates in the Internet and what it tells us about locality of BGP routing e v ents, Computer Communications , 2013; 36: 1592-1605. [55] Y anbiao Li, Daf ang Zhang, K un Huang, Dacheng He, W eiping Long, A memory-efcient parallel routing lookup model with f ast updates, Computer Communications , 2014; 38: 60-71. IJECE V ol. 6, No. 1, February 2016: 413 420 Evaluation Warning : The document was created with Spire.PDF for Python.