Congestion Control using Cross layer and Stochastic Approach in Distributed Networks

Transcription

1 Congestion Control using Cross layer an Stochastic Approach in Distribute Networks Selvarani R Department of Computer science an Engineering Alliance College of Engineering an Design Bangalore, Inia Vinoha K Department of Information science an Engineering The Oxfor College of Engineering Bangalore, Inia Abstract In recent past, the current Internet architecture has many challenges in supporting the magnificent network traffic. Among the various that affect the quality of communication in the massive architecture the challenge in maintaining congestion free flow of traffic is one of the major concerns. In this paper, we propose a novel technique to aress this issue using cross layer paraigm base on stochastic approach with extene markovian moel. The cross layer approach will brige the physical layer, link layer, network layer an transport layer to control congestion. The resource provisioning operation will be carrie out over link layer an the mechanism of exploring the congestion using stochastic approach will be implemente over the network layer. The Markov moeling is aopte to ientify the best routes amist of highly congeste paths an it is carrie out at the transport layer. An analytical research methoology will be aopte to prove that it is feasible to evelop a technique that can ientify the origination point of congestion an share the same with the entire network. It is foun that this approach for congestion control is effective with respect to en to en elay, packet elivery ratio an processing time. Keywors Distribute Network System; cross layer; congestion; Traffic Flow; Rate Control Metric I. INTRODUCTION Internet plays a vital role in issemination of knowlege an servicing seamless an ubiquitous communication in the present era. With the avancements in technologies like clou computing an optical network, offering high spee ata elivery, ata storage an retrieval is not an impossible task [1] yet, there is still a problem with the existing internet architecture. A closer look into the existing internet architecture revels that it is packe with various complexities Viz. incompatible in allowing connectivity with heterogeneous networks an its respective protocols, operating with various istribute networks [2][3]. The existing internet esign principles can only permit networking with less complexity in its routing an communication process. These principles are not scalable for the requirement of the future internet architecture. The reason behin this is untruste communication, more customer-oriente user environment, availability of many commercial network operators, atacentric utilities, an the worst part is intermittent connectivity [4]. Another challenging problem is its inclination towars Internet Protocol (IP) paraigm that makes it suitable for static internet users but not for mobile internet users. Therefore, whenever an application meets heterogeneity, it introuces a great eal of challenges for the network-base architecture an at the same time, it also leas to significant problems of resource allocation that can potentially affect the quality of performance. The connection technique of this architecture is characterize by one-to- many an many-to-many connections an also supports smart virtualization process. The significance of user-base participation is quite high with compatibility of multi-hop transmission scheme [5]. Unfortunately, none of the above mentione schemes are present even to a lesser extent in the existing internet architecture. The present paper eals with the problems relate to congestion control in future internet architecture. Although unerstaning the user-behavior over traffic an preicting it is an NP(noneterministic polynomial time) har problem, there are stuies existing in past that has alreay focuse on congestion control mechanism but less evience of stuies have focuse towars congestion control in future internet architecture. Essentially, this is built over three components viz. service, architecture an infrastructure. The next problem is interoperability. Given a scenario of multiple an heterogeneous network, it is a challenging task to process the control messages. This phenomenon is efinitely a big impeiment towars congestion. The next issue is for a given congestion over the ynamic network, it is quite challenging to maintain a balance between ientifying the point of congestion an processing heterogeneous control messages. Hence, it can be sai that it is quite a ifficult task to ientify an mitigate the level of congestion in this architecture. This paper presents a joint algorithm that incorporates cross layere mechanism with stochastic approach an Markov moeling to mitigate the potential issues of congestion in massive istribute system (future internet architecture). Section II reviews the existing literature for congestion control. The motivation an problem ientification is iscusse in section III. Section IV eal with the propose stuy an its significant contribution. The algorithms that are implemente to attain the goals are presente in section V. The results of the propose stuy are analyze in Section VI. The concluing remarks are iscusse in section VII. II. RELATED WORK The existing research in this area is revele here. Gholipour et al. [6] have carrie out an investigation on congestion problems in sensor network. The working principal of the sensor network operate with istribute algorithm. Here the authors iscusse a technique base on cost metric. The 192 P a g e

2 results were compare with respect to energy an packet. Efthymiopoulos et al. [7] have presente a stuy on congestion minimization pertaining to real-time streaming. The authors have introuce a technique that can provie traffic management in ifferent omains of the network base on the banwith. The system is purely mae for the internet-base peer-to-peer traffic. Jose et al. [8] have presente a congestion minimization technique that evaluates the rate of communicating signals in highly istribute manner. The outcome of the stuy was evaluate with respect to transmission rate an foun that it offers better rate control mechanism for minimizing the congestion. Zaki et al. [9] have presente a solution towars mitigating congestion that is witnesse over highly unpreictable mobile networks. The authors teste their fining over the continuous ate occurre on 3G network. The outcome of the stuy was evaluate with respect to throughput an elay to fin that propose system offers better resiliency for internet-base congestion. Ichrak et al. [10] have also investigate the problems of congestion in TCP-IP(Transmission Control Protocol/Internet Protocol)base connection. Sonmez et al. [11] have presente a technique that focuses congestion ientification an reuction owing to multimeia transmission. The stuy focuses on the congestion control an its effect on the quality of the transmitte multimeia files using fuzzy logic mechanism. The outcome of the stuy was evaluate with respect to Peak Signal-to-Noise Ratio (PSNR). Rey an Krishna [12] has presente cross layer approach in orer to mitigate the congestion issues in mesh network using TCP New Reno protocol. They focuse on efficient channel capacity optimizing uring the massive multimeia transmission an the results were assesse using packet transmission rate an elay. A scheme for controlling the congestion over TCP-IP base network was presente by the authors Carofiglio et al. [13]. They has use the principle of active queue management to control the congestion an foun that the technique possesse an effective winow size, roun trip time, an queue size. Further stuies towars istribute system were carrie out by Antoniais et al. [14]. Although they worke on a small network, the principle applie was consiere as a guiing factor for large scale istribute network as it focuses on aressing an effective traffic management technique using game theory. Cai et al. [15] have presente a moel for controlling congestion in TCP-base communication system. The propose methoology controls congestion over the wireless network base on noe-to-noe interactions. Using the case stuy of ahoc-base network, they have prove that their metho offere better congestion control. Uner the constraint of the faing channel, Ye et al [16] use probability theory to show that the congestion control moel for vehicular network offere improvement in the energy efficiency an ata packet transmission over ahocbase networks. Similar kin of work was carrie out by Bouassia an Shawky [17].They presente ynamic scheuling algorithm base on the priority of the messages. The focuse on improving ata reliability of real-time vehicular network. Kas et al. [18] have presente a technique for performing scheuling over ynamic channels. The aim of this is to increase the throughput from application viewpoint. A specific level of weight is assigne to each noe that is arbitrarily finetune base on saturation level of the queue. The results were evaluate with respect to en-to-en elay an packet elivery ratio over Constant Bit Rate traffic. Li et al. [19] have investigate congestion control for elay-base network. The authors have compare their work with respect to voice an ata traffic an showe that it can control congestion base on the available elay information. Misra et al. [20] has presente a unique technique base on automata theory for managing the congestion over wire network. The author have also applie stochastic-learning base mechanism an cellular automata for managing an effective queue size. The outcomes were assesse using sequence number, queue factor, etc. Uthra et al. [21] have propose a rate control mechanism for governing the traffic so that efficient throughput can be manage to ensure transmission free from any sorts of collision. The outcome of the simulation-base stuy is recore an compare with the existing preictive-base mechanism to control congestion an foun that the presente system minimizes the traffic congestion an also enhances the traffic performance. The following section presents the problem that is ientifie after reviewing the work that was carrie out by researchers in the fiel of congestion control. III. MOTIVATION AND PROBLEM IDENTIFICATION The following are the areas to be consiere for efficient performance of future Internet. The prevailing research in this area fail to aress the following:- The network quality parameters like elay, latency an channel capacity are not consiere efficiently for congestion control mechanism. The current cross layer esign allows manipulation of various layer parameters which leas to complication of congestion control an error management. In aition to that it is observe that the complexity of ientifying the source of congestion is ifficult because of the inefficient hanling of ranomness of traffic in heterogeneous network. IV. PROPOSED SYSTEM The aim of the propose system is to evelop a novel algorithm that can ientify the origin of congestion in istribute network system. Here, the emphasis on network resource allocation for ynamic ata flow control is given. As future internet architecture will possess all the possible complexities of existing internet as well as other networking stanars (owing to reconfigurable nature), it is essential to aress the issues through empirical an analytical moeling. In aition to regular quality parameters our research aresses the issues relate to air meium e.g interference an ifferent levels of noise over wireless channel for moeling the traffic. This paper is a continuation of our work where we have offere a packet level congestion by introucing a parameter i.e., Rate Control Metric (RCM)[24]. This metric is esigne to offer an efficient control over the highly istribute network. The system is esigne with the principle of cross layer paraigm. The ranomness in the heterogeneous network is stuie 193 P a g e

3 through stochastic base probability moel. This system is viewe as a massive network through graph theory moeling for better analysis of traffic congestion. In orer to stuy an mitigate the traffic congestion in heterogeneous network (Future internet architecture) the performance metric were analyze through RCM. In this research we have consiere cross layer approach for effective communication between networks. The resource provisioning technique is enhance through stochastic base approach where, the moel base on markovian moeling provies an optimize search for favorable noe for routing. This moel supports in ientifying the best possible noe amist of congeste noes for speey transfer of ata which enhances the throughput of the network. A. Cross Layer network moel Ientifying the origination point of congestion an etermining the control messages for processing the routing process to mitigate congestion requires a robust mechanism. Cross-layere approach is use to overcome this problem by controlling physical layer, ata link layer, network layer, an transport layer in the protocol stack of future internet architecture. The cross layer network moel plays an important role in arbitrary provisioning of network resources for congestion control. The presente scheme uses a significant routing factor that supports ynamic communication through multiple hops. It also uses a scheme that controls an manages the rate of traffic flow to achieve the fairness in sharing of network resources. A cost efficient provisioning algorithm is esigne that moels a novel queuing technique for maintaining queue stability. One of the significant focuses of the presente technique is to inclue the scenarios of noise an interference. This approach helps in processing control messages of multiple layers to ajust the rate of traffic flow uring peak hours an also select favorable noes for communication between two ifferent networks. Hence, the cross-layer scheme offers flexibility to process the control request with less elay an also ensures that it is applicable for istribute network with heterogeneity. B. Stochastic Approach The stochastic approach of the propose system mainly involves an integrate implementation of resource provisioning, communication an controlling the traffic irections. This approach initializes the iscrete networking states followe by selection of highly stabilize links, an apply provisioning. The system uses graph theory to esign an algorithm that works over the istribute machines. The significant contribution of this approach is to evelop a network moel that uses noise an signal power to categorize the quality of the links. It also consier the constraints of first two layers (link, physical) where there is no assure link rate for assigne time instances in istribute networking system. There is also a possibility that the capacity of the route may vary over a perio of time that will lea to a significant stochastic problem. In orer to solve this issue, we have implemente Rate Control Metric (RCM) [24] that can extract the exact information about the traffic rate thereby giving more information about the capacity of the routes. In orer to solve the problems relate to computational complexity, we also initialize a hypothetical matrix that stores an extracts the best provisione values, which acts as alternative for the congestion states of traffic. Hence, there is no significant control overhea ue to this. Moreover, the provisional matrix is regularly upate which makes the propose system inepenent from any egree of congestion foun in a specific transmission area. C. Markov Moelling The main aim of Markov Moeling is to optimize the stochastic approach use for congestion control. The goal of this moule will be to minimize the en-to-en elay in istribute networking system. We apply probability theory along with Markov moel to fin out alternate routes by exploring non-congeste paths for routing. Markov chain is use for mapping the network moel that uses queuing theory over the layere esign. (The stuy oesn t emphasize much on queuing mechanism explicitly as there is alreay robust mechanism specific to routing protocols in istribute network). The system maintains two types of traffics in a matrix i.e. local traffic an global traffic. Local traffic can be accesse at any instance of time an global traffic information can be accesse only when the noe has better resiual energy. The energy moel base on first orer raio moel or Raio Frequency (RF) circuitry principle [26] will be implemente in the propose system. The next section iscusses the implementation of cross layer algorithm, stochastic approach algorithm an markov moeling. V. ALGORITHM IMPLEMENTATION A. Algorithm for Cross-Layer Approach In this work, we aress the mitigation of congestion in the istribute network system through cross layer approach. The algorithm an implementation are as epicte below. Algorithm: Input: n (noes), ρ (queue) Output: upation of Link 1. init n, ρ, // n is the number of noes & ρ is the queue size 2. Estimate generate packets pkt on conition pkt J 3. Define link cost C s I 0, ( t) arg max [ ] 4. L[t]{Ls,[t]} 5. Define enhancement in cross-layer provisioning L Lc ( t) ( [ x], ( Lo Li )) x 0 6. Selectran[t]ϵL ran[t]=lc[t] 7. upate L[t] 194 P a g e

4 En The algorithm is formalize by consiering the number of noes n an initial queue size ρ. Fig.1 illustrates the complete process flow of the propose cross layer base provisioning. The queue stability of the istribute networking system is efine by the following equation which is use to filter out all the links that have their queue size tening to infinity. Reuce tcum Provisioning Value lim F t cum Init n, ρ If ρ is stable? T Set N/W cap Generate pkt Define Link cost Enhance Cross Layer Provisioning Store it in Matrix ( t) If ran[t] = Line [t]? T Sort it Ranom Extraction If Prob[Line-5] > maxval [Line-5] T Upate provisioning Matrix Stop Fig. 1. Process flow for Cross Layer Provisioning F F Abort it The link capacity of the network is expresse in terms of the pkt. Here, it is assume that at time t the noes in the network generate ata packets equivalent to queue size ρ with controlle variables I an J (I an J are positive integers) are as shown in line2 of the algorithm. As the future internet architecture supports higher range of heterogeneity in evice integration, there are possibilities of signal collision that leas to channel interference. In orer to istinguish the quality of links (or routes) a new measurement link cost C is consiere an is estimate base Δ ρ, where the variable Δ ρ represents ifference between the source queue ρ s an estination queue ρ at the time t. The link provier metric L[t] that is equivalent to L s, [t], where L represents a matrix of provier that consist of non-colliing links between any source(s) an estination() is consiere as the main parameter of the algorithm. It is assume that initially the buffer is share among each recipient noe. The link provisioning matrix is upate consiering the maximum value of the two arguments i.e. queue ρ[x] an ifference between outgoing capacity of link Lo an incoming capacity of link Li. The link provier will arbitrarily select an element from the matrix that satisfy the conition i.e. probability of selecte element is equivalent to enhance value (Line-6). It is upate as follows L Lc ( t) ( [ x],( Lo Li )) x 0 Here, the link metric L[t] is estimate in terms of its queue size an the link capacity helps in the route selection process. B. Algorithm for Stochastic Approach In istribute networking system the traffic may unergo uncertainties like ynamic topology, ranom mobility etc. in high egree of ranomness. The state of the network with uncertainty is analyze through stochastic process in which the future noe is ientifie with the theory of cross layer architecture. The noes are initialize an their etails are maintaine on a ata structure manage by graph theory. Owing to the istribute nature of the system, we assume that the control messages are free from errors or noise. After the implementation of cross-layer approach, we assume that there is no eviation or variance in the route capacity over the avancement of time. Algorithm Input: Es(energy for transmitting), δs, (gain factor of the power), β (capacity of the channel), ψ (noise ensity) Output: Provisioning state 1. Evaluate SNR SNR s, E s. s,. 2. Evaluate capacity of link Lcap = β log2(1+snrs,) 3. Define uplicate groups p { ps s, N, ps N s N } 4. Function for uplicate groups f ( s, p) { ps ( sps)^( ps p)^( ps 2)} 5. If s S Than 6. for all i D o 7. rcm(t)argmin(rcmmax, scaler_mult(t)); 8. Apply Algorithm-1 9. Transmit ata from s to 10. Upate scalar_multi(t)state of provisioning En The algorithm is implemente by efining a network moel, Signal-to-Noise Ratio (SNR) an Link Capacity (Line- 1 an 2 of the algorithm). The uplicate control messages for analysis purpose are generate using Line-3 of the algorithm. In the above algorithm the source noe s is ientifie as s i an matrix of uplicate control messages containing information 195 P a g e

5 about s as s.p. The uplicate groups are formulate using the equation as shown in Line-4. The cross layer architecture of the future internet is esigne in such a way that each source noe s can access its routing table N s. For reliable routing uring peak traffic the algorithm allows noe s to construct multiple hops with other noes for proviing alternate routes. The one imensional matrix is generate by scalar multiplication of s an p an the same is store at every noe. However, for all the uplicate control messages ps, only the noe that has highest value of i is chosen an is use in the computation process. The algorithm looks for all the source noes s (S is total source noes) an attempts to control the flow of packets. It then checks all the respective estination noes an uses rate control metric (RCM) [24] to further enhance the provisioning for the ata transmission. Finally, with the help of cross layer provisioning algorithm the ate is transmitte towars the estination. The significance of the stochastic base approach algorithm is that it further enhances the resource provisioning offere by cross-layer base provisioning technique at the link layer an also supports better communication in the network layer by favoring multiple hops routing in istribute networking system. Finally, the ata transmission is improve by applying rate control metric [24] which assigns an appropriate rate at the transport layer for effective en to en communication. Hence, the algorithm completely supports the cross-layer paraigm for future internet architecture to ensure interoperability among heterogeneous networks an achieve efficient ata transmission. C. Algorithm for Markov Moeling The Markov moeling is use to further strengthen the algorithm iscusse in the above sections an to apply stochastic moeling to further enhance the congestion control algorithm an offer a better solution to control traffic congestion. In Markov moeling each noe is represente as Mc that is compose of the total number of layers corresponing to Lcap+1 (numerically). The amount of ata packets pkt processe on each layer shoul be equivalents to Lcap such that 0<pkt<Lcap. We consier two ifferent forms of layers Viz. passive layer PL an active layer AL. Passive layer represents the passive process when the noes oesn t have any packet to forwar (pkt=0) whereas in active layer, noes always have packets for forwaring (pkt>0). As the future internet architecture possess ifferent wireless noes it is assume that there are other feasible communication outages that will call for retransmission phenomenon. We enote φ as the amount of retransmission an Wn to be amount of unit trail of transmission. The algorithm for Markov moeling is given below. Algorithm Input: pkt (Packet), Lcap (Link capacity w.r.t queue), PL(Passive Layer), AL(Active Layer), φ (Maximum amount of retransmission) Output: Ientification of free/busy routes 1. init pkt, Lcap, PL, AL, φ 2. Determine TP, PP, γp, IP. 3. Define area of collision A { s, } A A As A 4. s,, s, 5. Obtain Fs, =As / As, 6. Evaluate size matrix As,, Fs,, an F,s 7. Estimate number of Noes size( As,, Fs,, an F,s ).network ensity 8. Estimate the probability of minimum transmission p As 1 9. Evaluate b1, b2, b3 & Mc=Algorithm-2{b1, b2, b3} 10. Fin busy routes an free routes. En The problem of congestion in future internet architecture leas to network jamming that isrupt the process of ientifying the best noes for forwaring the ata packets. This problem can be aresse by esigning an algorithm that applies Markov moeling for evaluating the free an busy routes at the peak traffic situation. The algorithm takes the require inputs an computes maximum probability of passive state transition TP, preliminary state component PP, passive state probability component per states γp, an inter-arrival probability IP. Fig.2 shows the process flow for Markov moeling. The Markov moel is esigne by consiering three probability matrices viz. b1, b2 an b3. The matrix b1 an b2 represents the probability of a noe ientifying the busy channel in the first an secon Markov process. The matrix b3 represents the feasibility that the packet forwaring process fails ue to ata packet collision or interference or noise. Line- 3 shows the collision area A for the source noe s. The area A is efine as a transmission zone where there is interference of the neighboring noes resulting in traffic congestion in that particular transmission area. Hence, area A represents the possible congestion area. As shown in Line-4, it can be interprete that both the sener noe s an estination noe will lie within As,. There can also be another possible transmission zone Fs, as per Line-5 which may be unetecte in the area As.. It means that there may be an area e.g. F, which goes unetecte an the state of congestion is not etermine owing to ynamic mobility of noes in mobile networks. In this case a source noe or any intermeiate noe in area F s, cannot forwar the message to estination noe. 196 P a g e

6 P L pkt, Lcap, PL, AL, φ pkt = 0 Initialize moeling Parameter Construct M c Define Layers Determine b1, b2, b3 Determine Area of Collision Determine F s, As A A L pkt > 0 1 p As Fig. 2. Process Flow for Markov Moeling Estimate the probability of minimum transmission Extract communicating noes Formulate routes Stop The next phase of the algorithm is to compute the size of the transmission zones as per line-6. The algorithm computes the number of noes in transmission zones by scalar multiplication of network size an network ensity as per line- 7. The probability of minimum number of noes require for forwaring ata packets is compute as per Line-8. We use a simple variable ϑ that is equivalent to summation of the probability of all noes carrying out ata packet forwaring ivie by total probability of the noes forwaring ata packets from the congeste area. This phenomenon will mean that propose Markov moeling attempts to fin the existence of atleast one noe which is in fair position to perform ata transmission. The Markov moeling procees further to fin similar kin of noes an upates the matrix of ata communication path that was previously manage by the algorithm of stochastic approach. The upate matrices helps to fin the links between favorable noes as the best possible alternate routes for packet forwaring uring the peak traffic conition. VI. RESULTS AND DISCUSSION This section iscusses about the results generate from the network simulation through NS2 simulator. The simulation parameters are as shown in Table 1. TABLE. I. SIMULATION PARAMETERS Parameter Value Parameter Value Network area( 1000 x Control packet Simulation) area 1200 m2 size 32 bits Simulation Time Data packet size secons bytes Routing Protocol NetFlow Antenna Moel Omniirectional Pathloss exponent 0.5 Maximum Spee of noe 50 m/s MAC Type Minimum Spee noe 1m/s Traffic Moel CBR/VBR Transmission range 10m Mobility Moel Ranom Transmission Energy 0.5 J consumption Channel Moel Urban Receiving Energy consumption 0.25J Channel capacity 300 Mbps Ieal moe Energy consumption J Sleep moe Channel sensing 0.2 sec Power time consumption 0.02J Initial battery Energy of each 10J noe The propose work focuses in fining an effective solution for congestion control in istribute networking system. The performance parameters like packet elivery ratio, en-to-en elay an processing time are consiere to analyze the effectiveness of the propose system. It is benchmarke with similar stuies of Otoshi et al. [25] an Sahuquillo et al. [27]. Otoshi et al. [25] who have presente a stochastic moeling with preictive analysis for ientifying iscrete states of traffic in istribute networking system. This technique has use a preictive control scheme to minimize the possibilities of preictive error consiering network constraints e.g number of hops, length of the hops etc. The mean length of the hops was consiere as cost function, which was subjecte to optimization using CPLEX solver. The outcome of the work was quite convincing as it has offere better scalability for future internet architecture. Similarly, we consier the work carrie out by Sahuquillo et al. [27] as it offers solution to the congestion control for a practical case stuy of istribute networking system eg. High Performance Computing. The authors have use a mechanism that integrates injection throttle an segregation of congeste traffic. We perform a minor moification to techniques introuce in [25] [27] in orer to make a suitable testbe for carrying out the comparative analysis. The parameters consiere for analysis are en-to-en elay, packet elivery ratio an processing time. 197 P a g e

7 A. Comparative Analysis of En-to-En Delay The en to en analysis is carrie out by transmitting the test ata of 2000 bytes. The result is as shown in Fig.3. The graph shows that propose system is able to minimize the en-to-en elay to a larger extent as compare to existing stuies of Otoshi et al. [25] an Sahuquillo et al. [27]. The reason behin this is the technique that is aopte for processing search request an control messages by the propose system. Fig. 3. Comparative Analysis of Delay (sec) In the propose system, owing to Markov moeling, it becomes essential for a noe to obtain the significant aress information of another communication noe which coul possible resie in transmission zone of F s, or A s,. As both F s, an A s, are ifferent transmission zones, extraction of the noe aress will be a quite ifficult. We simplify this problem by eveloping a cross layer paraigm that can carry out the task of processing control messages in transport layer thereby minimizing the complexity. Here, the task of one layer is to aggregate the respective aresses of the noes an keep on exchanging it with other layers. This operation of interoperability is manage by the network layer. It is the responsibility of the network layer for carrying out the processing of control message as it maintains the communication stanars of each transmission zone. This process helps in ientifying the point of congestion an makes it aware to the entire network. This process has two avantages viz. i) all noes can quickly ecie about alternate routes an ecrease the impact of congestion uring peak traffic an ii) egree of congestion at the origination point is reuce by implementing active queue management that irects the packets from highly congeste area to less congeste point. Hence, en-to-en elay of the propose work is reuce in the presence of mobility of the noes which varies at every simulation track points. The problem explore in Otoshi et al. [25] is a preictive scheme. Here, the stochastic processing is aapte to preict an ientify the possible preiction error. Hence, the elay factor using this technique cannot be implemente for istribute system of ynamic nature like that of future internet architecture. Similarly, the work one by Sahuquillo et al. [27] have focuse on ientifying congestion by using control messages which is quite time consuming in its nature. Using Markov moeling, propose system offers optimize solution for ientifying the point of congeste an also offers best quality routes for packet forwaring thereby reucing the elay. B. Comparative Analysis of Packet Delivery Ratio Packet elivery ratio is compute by analyzing the amount of ata packets receive by the estination noe to total amount of ata transmitte by the source noe. The result shown in Fig.4 exhibits that the propose system offers better packet elivery ratio compare to Otoshi et al. [25] an Sahuquillo et al. [27]. This is because the propose system provies a better processing of ata generate by multiple networking omains in future internet architecture through cross layer paraigm. We start by analyzing the work one by Sahuquillo et al. [27]. The authors have implemente a technique where the incoming packets are organize at the input ports of the switches. The system emphasizes more on organization an less on queuing. This operation when implemente in our scenario reuces the packet elivery ratio. Moreover, the process of ientification of the congestion an notify it to other noes for upates are not iscusse in that paper [27]. It is also not sure whether the upates were one over the highly congeste area. This issue creates a negative impact on other neighboring noes by consuming more time to take ecision for routing. Hence, packet elivery ratio will be affecte when this technique is use in future internet architecture. Delratio-Propose Delratio-Otoshi Fig. 4. Analysis of Packet Delivery Ratio Delratio-Sahuquillo The technique propose by Otoshi et al. [25] has use the concept of traffic engineering. This technique was implemente through stochastic moeling which is more preictive in nature. The preictive analytic moel is assesse for its accuracy of traffic moeling using ranomness by aopting traffic engineering with cost as a function on the stochastic moel. This is much better than the technique iscusse by Sahuquillo et al. [27] as it can accomplish better packet elivery ratio. The main rawback of this technique is that it uses control server to optimize the cost function which leas to less efficient istribute routing. Although, the authors have use relaxation mechanism to sort out this problem, but the probability factors assume is less when compare to realtime traffic constraints. Hence, its packet elivery ratio is not better than the propose system. The propose system overcomes this problem by the algorithm-2 (stochastic) an algorithm-3 (Markov Moeling). These algorithms assist in ientifying the best possible routes from non-congeste area as well as congeste area. The upating mechanism is quite instantaneous with a pause time of secons in 198 P a g e

8 simulation stuy that leas to better packet elivery ratio for a longer perio of time. C. Analysis of Processing Time It is known that an effective congestion control mechanism must have a reuce processing time as far as possible. Lower the processing time means the network can ensure better instantaneous ata elivery process. We analyze the processing time with increasing traffic loa (packets per secons). A closer look into the Fig 5 shows that processing time gets reuce linearly with increasing traffic loa, which is one of the unique patterns of the propose stuy. Usually with increase network traffic, the processing time shoul be increasing but ue to cross layer approach the time complexity is reuce. The cross layer approach briges physical layer, link layer, network layer an transport layer. The provisioning operation is carrie out over link layer, the mechanism of exploring the congestion using stochastic is implemente over network layer an Markov moeling for further optimizing the best routes (even from highly congeste area) is carrie out at the transport layer. Fig. 5. Analysis of Processing Time Hence, the system maintains ifferent functionalities over ifferent layers of protocol stack resulting in reuce processing time in the propose system. For a given simulation environment, Otoshi et al. [25] an Sahuquillo et al. [27] work oesn t meets the emans of the istribute traffic scenario with ense congestion leaing to higher processing time when compare to propose system. VII. CONCLUSION Owing to the complexity in the esign principle of istribute networking systems e.g. future internet architecture, the existing algorithms an techniques o not provie solution for mitigating congestion. The propose system, therefore, presents a technique that uses conglomeration of cross layere approach, stochastic approach, an Markov moeling for aressing the problems of congestion in highly istribute networking system. We have aopte an analytical research methoology to prove that it is feasible to evelop a technique that can ientify the origination point of congestion an share the same with the entire network. The interesting point of implementation is that propose technique attempts to use the existing network resources for harnessing the channel capacity in accorance with the state ientifie by the propose system. The outcome of the stuy were compare with existing system respect to en-to-en elay, packet elivery ratio, an processing time an foun that propose system offers better solution for congestion control. REFERENCES [1] C. White, Data Communications an Computer Networks: A Business User s Approach, Cengage Learning, Computers, 2015 [2] P. Verissimo, L. Rorigues, Distribute Systems for System Architects, Springer Science & Business Meia, Computers, 2012 [3] N. Rajan, The Digitize Imagination: Encounters with the Virtual Worl, Taylor & Francis, Social Science, 2012 [4] J. Holler, V. Tsiatsis, C. Mulligan, S. Avesan, S. Karnouskos, D. Boyle, From Machine-to-Machine to the Internet of Things: Introuction to a New Age of Intelligence, Acaemic Press, Technology & Engineering, 2014 [5] H-Y Wei, J. Rykowski, S. Dixit, WiFi, WiMAX an LTE Multi-hop Mesh Networks: Basic Communication Protocols an Application Areas, John Wiley & Sons, Technology & Engineering, 2013 [6] M. Gholipour, A. T. Haghighat, M. R. Meyboi, Hop-by-hop trafficaware routing to congestion control in wireless sensor networks, Springer- Eurasip Journal on Wireless Communications an Networking, vol.15, 2015 [7] N. Efthymiopoulos, A. Christakiis, M. Efthymiopoulou, Congestion Control For P2P Live Streaming, International Journal of Peer to Peer Networksm, Vol.6, No.2, August 2015 [8] L. Jose, L. Yan, M. Alizaehy, G. Varghese, High Spee Networks Nee Proactive Congestion Control, ACM-Proceeings of 14 th ACM Workshop on Hot Topics in Network Article, No.4, 2015 [9] Y. Zaki, T. Potsch, J. Chen, Aaptive Congestion Control for Unpreictable Cellular Networks, ACM-SIGCOMM, 2015 [10] T. Ichrak, S. Nawal an A. Mustapha, Systematic Mapping Stuy on the Congestion Control Problem in TCP/IP, Contemporary Engineering Sciences, Vol. 7, no. 27, pp , 2014 [11] C. Sonmez, O. D. Incel, S. Isik, M. Y. Donmez, Fuzzy-base congestion control for wireless multimeia sensor networks, Springer- Eurasip Journal on Wireless Communications an Networkin, vol. 63, 2014 [12] C. P. Rey, P. V. Krishna, Cross Layer Base Congestion Control in Wireless Mesh Networks, Cybernetics An Information Technologies, Vol.14, No 2, 2014 [13] G. Carofiglio, M. Gallox, L. Muscariellox an M. Papaliniz, Multipath Congestion Control in Content-Centric Networks, IEEE-Conference on Computer Communication Workshop, pp , 2013 [14] P. Antoniais, S. Fia, C. Griffin, Y. Jin, Distribute meium access control with conitionally altruistic users, Springer- Eurasip Journal onwireless Communications an Networking, vol.202, 2013 [15] Y. Cai, S. Jiang, Q. Guan, an F R. Yu, Decoupling congestion control from TCP (semi-tcp) for multi-hop wireless networks, Springer- Eurasip Journal onwireless Communications an Networking, vol. 149, 2013 [16] F. Ye, R. Yim, J. Zhang, S. Roy, Congestion Control to Achieve Optimal Broacast Efficiency in VANETs, IEEE-International Conference on Communication, pp.1-5, 2010 [17] M. S. Bouassia an M. Shawky, A Cooperative Congestion Control Approach within VANETs: Formal Verification an Performance Evaluation, Hinawi Publishing Corporation, Eurasip Journal on Wireless Communications an Networking, Article ID , 2010 [18] M. Kas, I. Korpeoglu, an E. Karasan, Utilization-Base Dynamic Scheuling Algorithm for WirelessMesh Networks, Hinawi Publishing Corporation, Eurasip Journal on Wireless Communications an Networking, Article ID , 2010 [19] Y. Li, A. Papachristooulou, M. Chiang, A. R. Calerbank, Congestion control an its stability in networks with elay sensitive traffic, Elsevier-Computer Networks, vol.55, pp.20-32, 2011 [20] S. Misra, B. J. Oommen, S. Yanamanra, Ranom Early Detection for Congestion Avoiance in Wire Networks: A Discretize Pursuit 199 P a g e

9 Learning-Automata-Like Solution, IEEE Transactions On Systems, Man, An Cybernetics, vol. 40, no. 1, February 2010 [21] R. A. Uthra, S. V. K. Raja, A. Jeyasekar, A. J. Lattanze, A probabilistic approach for preictive congestion control in wireless sensor networks, Journal of Zhejiang University-Science, vol.15, Iss.3, pp.: , 2014 [22] K.Vinoha, R. Selvarani, Congestion Control in Distribute Networksa comparative stuy, Springer Avances in Computer Science an Information Technology, pp , 2012 [23] K.Vinoha, R. Selvarani, Congestion Control in Distribute Networking System-A Review, International Journal of Computer Applications, Vol.83, No 6, December 2013 [24] S. Rangaswamy an V. Krishnarey, An efficient traffic regulation mechanism for istribute networks, Springer- Eurasip Journal on Wireless Communications an Networking, vol.154, 2015 [25] T. Otoshiy, Y. Ohsitay, M. Muratay, Traffic Engineering Base on Stochastic Moel Preictive Control for Uncertain Traffic Change IEEE International Symposium on Integrate Network Management, pp , 2015 [26] A. Grebennikov, N. Kumar, B. S. Yarman, Broaban RF an Microwave Amplifiers, Taylor & Francis, Computers, 2015 [27] J. E.Sahuquillo, E. G. Gran, P. J. Garcia, Efficient an Cost-Effective Hybri Congestion Control for HPC Interconnection Networks, IEEE Transactions On Parallel An Distribute Systems, vol. 26, no. 1, January P a g e