ospf

A REPORT ON “ANALYSIS OF OSPF ROUTING PROTOCOL” Using OPNET 14.5 Modeler

NORTH CAROLINA STATE UNIVERSITY

SUBMITTED BY: SHOBHANK SHARMA [email protected]

Page 1

ANALYSIS OF OSPF ROUTING PROTOCOL A. Introduction OSPF (Open Shortest Path First) is an interior gateway routing protocol deployed typically in upper tier ISPs for intra-AS routing. It is a link state protocol employing Dijkstra’s algorithm to calculate least cost path. The following report dwells into the OSPF routing protocol investigating the following: 1) What are the ways of cost assignment supported by OSPF and what are the advantages and disadvantages of one over the other? 2) How the load balancing feature affects throughput? 3) What is the affect of dividing network into various areas? 4) Comparison of OSPF with other routing protocols like EIGRP and RIP. Therefore the analysis covers new features introduced by OSPF in the field of routing protocols. OPNET simulation is used for the stated analysis.

B. Cost assignment in OSPF The cost or metric as assigned to an interface in OSPF is an indication of the overhead necessary to allow packet transfer across the interface [1]. OSPF offers two types of cost assignments- the first, implicit way is by assigning bandwidth to the interface and the cost relates to bandwidth inversely with the reference bandwidth of 1000000000. A more flexible way would be to allow the privileged network administrator to assign cost explicitly in some units as desired. The later way can take into account the infrastructure cost, delay, administrative domain, bandwidth or any defined parameter. Experimental setup: The experimental setup consists of four ethernet4_slip8_gtwy node models having point to point (PPP link models) interconnections and exposed to three experimental scenarios: • With DS3 links • With all OC3 links • With one OC3 link and manual costs In the experiment we measure the utilization of various links in the three experimental settings. Fig. 1, Fig. 2 and Fig. 3 show the results. Results: Fig. 1, having all DS3 links, shows that all the links have utilization nearing to 100% and hence though economically viable due to cheaper DS3 links, it is not very efficient causing high load on links. Fig. 2 is an upgraded version of the previous scenario; in this all the links are OC3 and hence it is a costlier solution but gives lesser load on links. Fig. 3 is an intermediate solution in which there is only one OC3 link and costs are assigned manually. The cost can be assigned in such a way so that more traffic is on the link we want it to pass through. For example here we have tried to change path of traffic from A to D to pass through A to C and C to D as shown in Fig. 4. Hence the last scenario makes advantage of the flexibility offered by user defined cost assignment feature of OSPF. C. Load Balancing feature of OSPF In today’s internet the shortest path routing protocols are not enough to maximize guaranteed node traffic loads, scalable and fast bandwidth reservations, hence load balancing comes to rescue. Load balancing can affect throughput positively for arbitrary traffic pattern [2]. Router can learn multiple routes to a destination using the same routing process and chooses the path with the lowest cost to destination. Load balancing feature allows the routing process to install multiple routes in route table, to the same destination in order to balance the traffic flow. Page 2

Fig. 1 All nodes connected with PPP DS3 Link -Model.

Fig. 3 One link with PPP OC3 Link-Model and other links with PPP DS3 link model, with manual costs assigned.

Fig. 2 All nodes connected with PPP OC3 Link-Model.

Fig. 4 Route when cost is manually assigned

Experimental setup: The experimental setup for analyzing the effect of load balancing consists of similar topology as before i.e. four ethernet4_slip8_gtwy node models having PPP DS3 link model interconnections with subnets in the form of 100BaseT LAN node model. Application profile is configured to contain FTP application and Profile configuration is configured to download FTP traffic. One LAN model with 5 server workstations is connected to router A and another LAN model with 50

Page 3

host workstations is connected to router C. The simulation is performed to find the routes and throughput characteristic of the link between router C and the connected subnet, for the two cases.

Fig. 5 Route from A to C without load balancing.

Fig. 6 Route from A to C with load balancing.

Results: From the experiment conducted we obtain the snapshots as shown in Fig. 5-7. Comparing Fig. 5 and 6, it is clear that when load balancing is used the traffic does not pass through only one link but rather balances itself in order to lower the load on one particular link. Fig. 7 shows that the throughput of ‘with load balancing’ has improvement over ‘without load balancing’ topology. A further improvement in load balancing implementation is to assign some kind weights or load balancing coefficients. Balancing coefficients are optimized to maximize the network throughput while ensuring that nodes can generate and receive loads which are proportional to the allocated weights [2].

Fig. 7 Throughput comparison of with and and without load balancing.

Fig. 8 Topology consisting for ten routers

interconnected using PPP DS3 Link Model. Page 4

D. Areas Configuration in OSPF OSPF provides feature by which whole network can be split up into different areas which assist in restricting the link state updates over the network and often cuts down on the unneeded information of one area to be passed on to other. Experimental setup: The experimental setting comprise of ten routers interconnected using PPP DS3 node model as shown in Fig. 8 The simulation is setup to demonstrate and compare the OSPF traffic sent in bits/sec for withand without-area configuration. Area configuration (see Fig. 8) for links is as follows: r1<->r2, r1<->r3, r3<->r4, r2<->r4—area 0.0.0.1 (blue) r3<->r5, r5<->r6, r4<->r6, r5<->r7, r6<->r8—area 0.0.0.0 (green) r7<->r8, r8<->r10, r10<->r9, r9<->r7—area 0.0.0.2 (red)

Fig. 9 OSPF Protocol traffic sent for with and Without area configuration.

Fig. 10 Average traffic sent by routing Protocols.

Fig. 11 Network convergence duration Comparison of routing protocols.

Fig. 12 Network convergence activity Comparison of routing protocols. Page 5

Results: As can be seen from the Fig. 9 the overall OSPF protcol traffic flow is lesser in the case of with-area configured network as compared to without-area configured network. The reason for such a behavior is because of restriction on LSAs to be confined to its own area only hence reducing the routing overhead. Hence, Dividing an Open Shortest Path First (OSPF) Autonomous System (AS) into independent routing areas allows area topology abstraction, reducing route overhead, table size, and convergence time, while providing some isolation from bad routing data [3]. E. Comparison with EIGRP and RIP In this experiment we will compare OSPF with other interior gateway routing protocol. EIGRP is Enhanced Interior Gateway Routing Protocol. It is a distance vector protocol like RIP, with routing optimizations based on Diffusing Update Algorithm (DUAL) that gives loop-free operation and provides mechanism for faster convergence. Experimental Setup: The topology used is same as shown in Fig. 8 but without areas being configured. The simulation is setup to demonstrate and compare network convergence duration, network convergence activity and the average traffic sent in b/s for EIGRP, OSPF and RIP. Results: As can be seen from Fig. 10 the least traffic sent is for RIP since it is very simple protocol without much overhead. But it is demonstrated in Fig. 11 that EIGRP has smallest convergence duration as compared to OSPF and RIP. This is verified by Fig. 12 which shows network activity and here also EIGRP can be seen to converge faster than OSPF and RIP. F. Conclusion OSPF offers advantageous features like assigning manual costs, load balancing, dividing network into areas to restrict LSAs as well as for administrative reasons, encrypted protocol traffic etc. But there is other side of coin too. The OSPF area configuration reduces connectivity, increases complexity, routing path lengths and traffic concentration. EIGRP has better convergence performance and lesser overheads. In spite of these disadvantages OSPF still triumphs amongst routing protocol and continues to be used in upper tier ISPs.

G. References [1] www.cisco.com Document ID: 7039 [2] Antic M., Maksic N., Knezevic P. and Smiljanic A., "Two phase load balanced routing using OSPF" IEEE Journal on Selected Areas in Communications January 2010. [3] Manousakis K. and McAuley A.J.,"Using stochastic approximation to design OSPF routing areas that satisfy multiple and diverse end-to-end performance requirements" Proceedings of the 6th intl. symposium on modeling and optimization Wiopt 2008. [4] Kurose and Ross, Computer Networking: A top-Down Approach Featuring the Internet.

Page 6

H. Readme Page Note: For Examiner: 1) Path to opnet models: J:\.eos.ncsu.edu\lockers\workspace\csc\csc570\001\ssharma5\op_models. 2) There are three folders in the above directory path: linkcostanalysis – for section B Loadbalancing OSPF – for section C OSPF Comparison – for section D and E 3) The simulation in all the scenarios is set for 10 minutes. 4) Each folder has .project folder consisting of .prj file. 5) From the opnet modeler this .prj file can be opened and it automatically loads the simulation results. If needed the simulation can be rerun.

Page 7