Você está na página 1de 4

Mohammed Rizwan Adil

ELEC 6120

Routing Strategies and Algorithms


Background of the system: Routing Overview: The fundamental concern in switched data networks is that of routing. There are usually several possible routes through which packets may be transmitted in a packet switching network. In such a case, the network has to select the best possible route which can be done using a routing function. Such a routing function should be flexible enough to handle localized failures, overloads etc but should also incorporate stability. We are also concerned with the performance criterion which is employed to determine the best route. The simplest performance criterion is that of the least number of hops. In this technique, we take the path that has the least number of nodes or hops from source to destination. However, not all links are identical. Links may have different data rates, different queuing delays etc. We therefore assign a cost to each link and calculate the least cost route between the source and destination. Additionally, all nodes introduce transmission overhead and processing overhead which further impact the efficiency. Different approaches to routing: Routing can be done in several ways. In Centralized routing, a central node decides the path of the data packet. Even though such a technique is relatively simple, it is prone to error. In Source routing, the source determines the path across the whole grid. Such a method is best used for virtual circuit networks. Both these techniques suffer from shortcomings because of their localized control. Therefore, we use Dynamic Routing in most cases where in each node decides the link to which it will transmit the packet based on the varying cost of the link. In other words, the decisions are made on a packet by packet basis. Some strategies even use techniques like flooding and random transmissions and get the job done. Routing Algorithms: Routing usually follows one of the two Least-Cost Algorithms- Dijkstras algorithm and Bellman-Ford algorithm. In the Dijkstras algorithm, we find the shortest path from a given source node to all other nodes by developing the paths gradually. At each stage, we add a shortest path to a new node. This way, by the end, we have the least cost path to each node from a particular node. In the Bellman-Ford algorithm, the shortest path to a given source subject to the constraint that it has only one link is employed, followed by those with two links and so on. Technical problems to solve and challenges (Weaknesses in the current systems) The central problem behind all routing techniques is that there is no one perfect solution for all cases. Different approaches have different benefits and demand different compromises. a. Problems with dynamic routing: In dynamic routing the network can respond to sudden changes in the links but do not have centralized control. b. Problems with fixed routing: In the fixed routing technique, every node is assigned a routing table which directs the packet to a particular node. In the event of a broken link, the system fails. c. Problems with flooding: Duplicate copies of packets keep circulating around the system.

Mohammed Rizwan Adil

ELEC 6120

d. Problems with Design Algorithm: In the Dijkstras algorithm, at each stage we add a new node to the set of nodes and its path to the source is defined. In other words, it consumes lot of resources and does an unorganized search which consumes a lot of time. This makes it very difficult to employ Dijkstras algorithm to huge number if nodes [1]. e. In the Bellman-Ford algorithm, if a link or node gets broken, then the other nodes keep increasing their estimates until infinity. In other words, the process loops until all nodes find out that the weight of link to the last node is infinity. This problem is referred to as the Count to Infinity Problem. Yet another problem in the Bellman-Ford Algorithm is that of incorrect entries in the routing table. This may cause routing table loops which consume the time and resources but dont provide any connectivity [2]. f. In addition to both these problems, both algorithms take time to update which can cause problems in dynamic topologies. We also have the prospect of changing link costs over time depending upon the traffic. Network system design that meets the challenges (Proposed ideas for improvements) a. Proposed improvements for fixed routing: The routing table can be added with an alternate next node. The packet can be sent to this node if the actual node link is broken. b. Proposed improvements for flooding: This problem can be solved by setting a hop count field with each packet which decrements by one when it passes through a node. When the hop count becomes 0, the transmission stops. c. Improvements for the Dijkstras Algorithm: To use Dijkstras algorithm, we use its Adjacency Matrix and Distance Matrix, both of which contain several 0s and infinities. Both 0 and infinity take up lot of computing power and hence can cause problems in huge data sets [1]. Therefore, we use the Adjacent Node Algorithm which is an advanced version of Dijkstras algorithm. In this algorithm we have the Adjacent Matrix in which we record only nodes related to the network and the Judgment Matrix which is a condensed form of distance matrix. This way, we are able to simplify the Dijkstras algorithm. d. Improvements for the Bellman-Ford Algorithm: To avoid the Count to Infinity problem, we need to make the routers send information only to neighbors that are not exclusive links to the destination. The loop due to incorrect values in the routing table can be resolved by using the Labeled Distance Vector Routing Algorithm. In this technique, each node maintains vector labeled distances, labels are updated only when the new node distance is smaller than the existing distance [2]. e. To improve the routing after the link cost has been altered by dynamic topologies, fuzzy logic can be integrated into the routing process [3]. Further Recommendations: Floyd- Warshall Algorithm: The routing at the network layer can be categorized into two types: static and dynamic. In static algorithms, topology and packet flow cannot influence the routing decisions. Once the algorithm is downloaded and installed to the network, the routers blindly follow it. On the other hand, in dynamic routing, packet flow, topology etc impact the link cost which eventually influences the routing decision [4]. The central idea behind the Floyd Algorithm is that it doesnt give out any shortest path; it gives sum of all possible short paths. The Floyd Algorithm works in a recursive fashion. Given any two points A and B and a third

Mohammed Rizwan Adil

ELEC 6120

point C, it finds the smaller one between AB and ACB. This process continues until the smallest path is found [5]. Yet another innovative and simple idea proposed is that of robotic routers. These can be used in places where the usage of routers is very sparse but connectivity is needed, for instance on large farms or even in the military [6]. Suggestions to improve the routing process: 1. Several routers can be connected to a central buffer which can be used data on any of them increases their individual buffer limit. This way, congestion can be avoided to some extent. 2. Routers with low data at any instant should offer their services to hold data packets for other routers until they process their data. This will also help in prevention of data loss. 3. The values in the routing table which get used a lot can be separated and another temporary table be created for quicker transmission. Relation with course material: We have tried to address the fundamental issue of routing by using different techniques. Fundamental techniques and issues for routing packet data have been enunciated. Throughout this term paper, we have further explored Dijkstras algorithm and Bellman-Ford algorithms from the text book and tried to address their shortcomings. We have proposed some ways to overcome the disadvantages because of both these techniques. Conclusion: Routing is perhaps the most integral aspect in communication. It is all the more significant when it comes to the internet because the whole hardware set up of routers is going to be dynamic forever. In such situation, the networks need to be able to adapt to varying nodes and provide uninterrupted connectivity to the end user. Several algorithms and improvements upon these algorithms have been proposed but there is still a lot more that can be done.

Mohammed Rizwan Adil References:

ELEC 6120

1. Zhang Fuhao; Liu Jiping; , "An Algorithm of Shortest Path Based on Dijkstra for Huge Data," Fuzzy Systems and Knowledge Discovery, 2009. FSKD '09. Sixth International Conference on , vol.4, no., pp.244-247, 14-16 Aug. 2009 doi: 10.1109/FSKD.2009.848 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=5359145&isnumber=53 59075 2. Garcia-Luna-Aceves, J.J.; , "Distributed routing with labeled distances," INFOCOM '92. Eleventh Annual Joint Conference of the IEEE Computer and Communications Societies, IEEE , vol., no., pp.633-643 vol.2, 4-8 May 1992 doi: 10.1109/INFCOM.1992.263467 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=263467&isnumber=662 5 3. Aboelela, E.; Douligeris, C.; , "Routing in multimetric networks using a fuzzy link cost," Computers and Communications, 1997. Proceedings., Second IEEE Symposium on , vol., no., pp.397-401, 1-3 Jul 1997 doi: 10.1109/ISCC.1997.616031 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=616031&isnumber=134 14 4. Matt Bishop, Computer Security Art and Science. Electronic Industry Press, Beijing, 2004 5. Ming-Xin Yang; Bing-Tong Wang; Wen-Dong Guo; , "Research on the performence of dynamic routing algorithm," Machine Learning and Cybernetics, 2009 International Conference on , vol.5, no., pp.2647-2650, 12-15 July 2009 doi: 10.1109/ICMLC.2009.5212662 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=5212662&isnumber=52 12095 6. Tekdas, O.; Isler, V.; , "Robotic routers," Robotics and Automation, 2008. ICRA 2008. IEEE International Conference on , vol., no., pp.1513-1518, 19-23 May 2008 doi: 10.1109/ROBOT.2008.4543416 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4543416&isnumber=45 43169

Você também pode gostar