An Adaptive Routing Algorithm for Communication Networks using Back Pressure Technique

Transcription

1 International OPEN ACCESS Journal Of Moern Engineering Research (IJMER) An Aaptive Routing Algorithm for Communication Networks using Back Pressure Technique Khasimpeera Mohamme 1, K. Kalpana 2 1 M. Tech (CSE), MVR College of Engineering an Technology, A.P., Inia. 2 Asst. Professor, Dept. of Computer Science & Engineering, MVR College of Engineering an Technology, A.P., Inia. Abstract: The basic iea of backpressure techniques is to prioritize transmissions over links that have the highest queue ifferentials. Backpressure metho effectively makes packets flow through the network as though pulle by gravity towars the estination en, which has the smallest queue size of zero. Uner high traffic conitions, this metho works very well, an backpressure is able to fully utilize the available network resources in a highly ynamic fashion. Uner low traffic conitions, however, because many other hosts may also have a small or zero queue size, there is inefficiency in terms of an increase in elay, as packets may loop or take a long time to make their way to the estination en. In this paper we use the concept of shaow queues. Each noe has to maintain some counters, calle as shaow queues, per estination. This is very similar to the iea of maintaining a routing table (for routing purpose) per estination. Using the concept of shaow queues, we partially ecouple routing an the scheuling. A shaow network is maintaine to upate a probabilistic routing table that packets use upon arrival at a noe. The same shaow network, with back-pressure technique, is use to activate transmissions between noes. The routing algorithm is esigne to minimize the average number of hops use by the packets in the network. This iea, along with the scheuling an routing ecoupling, leas to elay reuction compare with the traitional back-pressure algorithm. Keywors: Back pressure, Packets, Routing, Shaow queue. I. INTRODUCTION The back pressure metho first propose in [1] has recently shown a great potential for solving a number of issues in the wireless multi-hop networks. The central iea of back-pressure scheuling moel is that contention among the links shoul be resolve by scheuling the link which has the largest prouct of the queue ifferential backlog between its enpoints an the transmission rate at which the link can be serve. In a perfectly time-slotte meium access mechanism such as TDMA (Time Division Multiple Access), this will result into optimal throughput of flows while guaranteeing the queue stability (ingress traffic to a queue never excees its egress traffic). The utility maximization framework initially propose in [2] shows that iection rates of the flows shoul be chosen such that aggregate utility of the flows is to be maximize. Here the utility of flow represents a esirable effect on the entire network achieve by a particular rate of the flow. It was shown in [3] [4] that the backpressure scheuling an utility base rate control together can solve the global problem of the network utility maximization. The funamental challenge with back pressure metho is that solution of the unerlying scheuling strategy is NP-har [5]. Also, since it was propose for a centralize, time slotte, an synchronize system, a istribute implementation which can achieve even a closer approximation is very ifficult to evelop. Recently, [6] have attempte to incorporate backpressure technique base scheuling in ranom meium access protocols such as CSMA/CA. These protocols try to approximate the performance of the ieal back pressure scheuler by prioritizing the frame transmissions accoring to ifferential backlogs of the queues. Here, every noe in the network maintains a per estination queue (PDQ) an the packets estine to a particular estination host are store in the PDQ of that estination host until further forwaring ecisions are mae. Now, noes share their PDQ information with their neighboring noes, an this information is utilize by every noe to calculate the ifferential backlogs of its PDQs. The ifferential backlog of a PDQ at a noe is equal to the size of the PDQ minus the size of the PDQ of its upstream neighbor towars the estination en. To emulate the back pressure scheuling, the packets of the PDQ which has the highest ifferential backlog (highest back-pressure) in the neighborhoo are given the higher chances for the transmission. This way, the likelihoo that the packets are transmitte from a particular PDQ at a noe is proportional to its ifferential IJMER ISSN: Vol. 4 Iss.9 Sept

2 backlog compare to the ifferential backlogs of PDQs of all noes in the neighborhoo. This prioritization quickly moves the traffic from long back logge queues to the shorter queues achieving an improve throughput an a better overall stability of queues. II. RELATED WORK The first theoretical work on backpressure scheuling technique is the classic result in [1], proving that this queue-ifferential base scheuling metho is throughput optimal (i.e., it can stabilize any feasible rate vector in a network). Since then, researchers have combine the basic backpressure technique with utility optimization to provie a comprehensive approach to the stochastic network optimization [7], [8]. In [9], the authors present a metho whereby only one real queue is maintaine for each neighbor noe, along with virtual counters an shaow queues for all estinations, an show that this yiels elay improvements. In [10], a novel variant of the backpressure scheuling technique is propose which uses hea of line packet elay instea of the queue lengths as the basis of the backpressure weight calculation for each link or commoity, also yieling enhance elay performance. However, these works both assume the existence of the static fixe routes. In [11] an [12], the authors present two works on backpressure routing aime specifically for clusterbase intermittently connecte networks. In [11], the authors evelop a two phase routing metho, combining the backpressure routing with source routing for the cluster-base networks, separating intra cluster routing from inter-cluster routing. They show that this scheme results in large queues at only a subset of the noes, yieling smaller elays than conventional backpressure technique. In [12], the authors implement the abovementione algorithm in a real experimental network an show that the elay improvements empirically. The key ifference of these works from our propose work is that we o not make any assumption about the intermittently connecte network being organize in a cluster base hierarchy an we require no previous knowlege of noes mobility. III. PROPOSED WORK 1. Throughput Optimal Backpressure Algorithm This algorithm works on wireless networks. This back pressure algorithm shoul be epening on one main proceure. The main proceure is the estination queue proceure by using these proceure only we can reucing the poor elay of the time an the istance every noe at the time of routing file in between the two or more locations preestination queues maintaining the two phases: routing algorithm scheuling algorithm. In this system says istance is very high an time is very high. So may raw backs will be occurre at the time of working on the routing algorithm that s why we are implementing a new algorithm. Scheuling algorithm combine with the routing algorithm. Our scheuling algorithm says what to proviing the scheule for prepress of the over come to raw backs that s why we are using the roun robin algorithm an shortest path routing algorithm. In this algorithm proviing quality of services to the en users epening on the service only routing will be ecie it is best routing or the ba routing. QOS may be epens 4 phases: Banwith, frequency, time, istance. In any network epens on QOS. Banwith an frequency is very high an time istance is low QOS is very high. At each link the algorithm assigns a weight to each possible estination noe that is calle back pressure. Define the back pressure at link (n,j) for estination noe at slot t to be: where Qn(t) enotes the number of packets at noe n enote for noe at the beginning of time-slot t. Uner this notation, Qnn (t) =0, t. Assign a weight w to each link (n, j), where w is efine to be the maximum back pressure overall possible estinations. i.e. IJMER ISSN: Vol. 4 Iss.9 Sept

3 2. Min Resource Routing The back-pressure algorithm explores all paths in the entire network an, as a result, may choose paths that are unnecessarily long, which may even contain some loops, thus leaing to poor performance. We aress this problem by introucing a cost function that measures the total amount of the resources use by all flows in the network. Specially, we a up the traffic loas on all links in the network an use this as our cost function. The goal then is to minimize this cost subject to the network capacity constraints. Given a set of packet arrival rates that lie within the capacity area, our goal is to fin the routes for the flows so that we use as few resources as possible in the network. Thus, we formulate the following optimization problem: min s. t. f F x N, n N { }( n, j) L ( ) L f I{ b( f ) n, e( f ) } (ln) L ln ( ) L Figure 1 illustrates how the M back pressure algorithm works in a simple wireline network. All links in the network can be activate simultaneously without interfering with each other. Note that the backlog ifference of route 1 is 6 an the backlog ifference of the route 2 is 4. Because the backlog ifference of the route 2 is smaller than M, the route 2 is blocke at current traffic loa. The M -back-pressure algorithm will automatically choose the route 1, which is shorter. Therefore, a proper M can avoi the long routes in when the traffic is not close to capacity. Fig. 1: Link weights uner the M -back-pressure algorithm. 3. Shaow Queue Algorithm In aitional to the real queues, each noe n also maintains a counter, which is known as shaow queue, pn for each estination. Unlike the real queues, the counters are much easier to maintain even if the number of counters at each noe grows linearly with the size of the overall network. The shaow queues are upate base on the movement of the fictitious entities calle as shaow packets in the network. The movement of the fictitious packets can be thought of as an exchange of the control messages for the purposes of the routing an scheule. Just like the real packets, shaow packets arrive from outsie the network an eventually exit the network. The back-pressure for estination on link (n j)is taken to be as follows: w [ p [ p [ M n j The evolution of the shaow queue p n [ is p n [ t 1] p { ln[ } l:(ln) L n [ ln[ I{ b( f ) n, e( f ) } f F I{ [ } u[ j:( ) L I u a [ f IJMER ISSN: Vol. 4 Iss.9 Sept

4 Where u [ is the number of shaow packets transmitte over the link (n j) in time t. [ is the estination noe that has the max weight on the link (n j). a f [ is the number of shaow packets generate in time t for flow f. 4. Aaptive Routing Algorithm When a packet arrive at a noe, then it use route as follows: Let [ is the number of shaow packets from noe n to the noe j for estination uring time t. Let enote an estimate of the expecte value of [ at time t. At each time slot t, the following sequence of operations occurs at each noe. A packet arriving at noe n for estination is inserte in the real queue P [ [ k:( nk) L nk[ q for next-hop neighbor j with probability: Packets waiting at link (n j) are transmitte over the link when that link is scheule as shown in figure 2. Fig. 2: Probabilistic splitting algorithm at noe IV. CONCLUSION ackpressure scheuling an routing, in which packets are preferentially transmitte over communication links with high queue ifferentials, offers the promise of throughput-optimal operation for a wie range of the communication networks. However, when the traffic loa is very low, ue to the corresponing low queue occupancy, then the backpressure scheuling/routing experiences long elays. This is particularly of concern in the intermittent encounter-base mobile networks which are alreay elay-limite ue to the sparse an highly ynamic network connectivity. While state of the art mechanisms for such networks have propose the use of the reunant transmissions to improve the elay, they o not work well when the traffic loa is high. We ecouple (to a certain egree) routing an scheuling in the network through the use of the probabilistic routing tables an the so-calle shaow queues. We use the same number of shaow queues as the back-pressure technique, but the number of real queues is very small (per neighbor). The new iea here is to perform the routing via probabilistic splitting, which allows the ramatic reuction in the number of the real queues. Finally, an important observation in this work is that the partial ecoupling of the shaow back-pressure an real packet transmission allows us to activate more links than a regular backpressure metho woul. This iea appears to be essential to reuce the elays in the routing case. IJMER ISSN: Vol. 4 Iss.9 Sept

5 REFERENCES [1] L. Tassiulas an A. Ephremies, Stability properties of constraine queueing systems an scheuling policies for maximum throughput in multihop raio networks, ec 1990, pp vol.4. [2] F. Kelly, A. Maulloo, an D. Tan, Rate control in communication networks: shaow prices, proportional fairness an stability, in Journal of the Operational Research Society, vol. 49, [3] L. Chen, S. H. Low, M. Chiang, an J. C. Doyle, Cross-layer congestion control, routing an scheuling esign in a hoc wireless networks, in INFOCOM th IEEE International Conference on Computer Communications. Proceeings, 2006, pp [4] A. Eryilmaz an R. Srikant, Joint congestion control, routing, an mac for stability an fairness in wireless networks, Selecte Areas in Communications, IEEE Journal on, vol. 24, no. 8, pp , [5] M. J. N. Georgiais Leonias an L. Tassiulas, Resource allocation an cross-layer control in wireless networks, in Founations an Trens in Networking, [6] U. Akyol, M. Anrews, P. Gupta, J. Hobby, I. Saniee, an A. Stolyar, Joint scheuling an congestion control in mobile a-hoc networks, april 2008, pp [7] X. Lin, N. Shroff, an R. Srikant, A tutorial on cross-layer optimization in wireless networks, IEEE Journal on Selecte Areas in Communications, vol. 24, no. 8, [8] M. J. Neely, Stochastic network optimization with application to communication an queueing systems, Synthesis Lectures on Communication Networks, vol. 3, no. 1, [9] L. Bui, R. Srikant, an A. Stolyar, Novel architectures an algorithms for elay reuction in back-pressure scheuling an routing, IEEE INFOCOM [10] B. Ji, C. Joo, an N. Shroff, Delay-base back-pressure scheuling in multi-hop wireless networks, IEEE INFOCOM, [11] J. Ryu, L. Ying, an S. Shakkottai, Back-pressure routing for intermittently connecte networks, IEEE INFOCOM [12] J. Ryu, V. Bhargava, N. Paine, an S. Shakkottai, Back-pressure routing an rate control for icns, ACM Mobicom, IJMER ISSN: Vol. 4 Iss.9 Sept