Routing protocol

1       Principle for selecting a routing protocol

With the rapid development of Internet, TCP/IP has been in mainstream protocol. Many different types of routers running in the network carry the most important information. Without routing protocols, these routers can’t work together in phase. So when we begin to design a large-size network, choosing a suitable routing protocol is very important. Commonly used unicast routing protocols include RIP, IS-IS, BGP as well as Cisco private protocols of IGRP/EIGRP. According to the scale of network, you can select different protocol.

Connectivity is most fundamental requirements of IP network; this is the base of protocol choice. Other factors such as network typology, network management also should be considered:      

1) Compatibility of protocols: Compatibility of protocols makes sure connectivity and expansibility of network, because different manufactures can support these protocols. And customers also have more choice.  

2) Typology of network: Typology of network has direct influence on protocol choice. For example RIP is not suitable for complicated network, because its coverage is limited to a certain degree. So we need more powerful protocols, such as OSPF, EIGRP.   

3) Strong and Stable: Being the signaling protocol to ensure network connectivity, the routing protocols must be strong and stable. Various abnormalities will appear in the network such as hardware error or heavy load. Because routers locate at the decision-making points of the network, if routers have error, they may cause unpredictable network behavior. Routing protocols must be able to bear various abnormalities for a long time.

4) Best path selecting: Routing protocols aim at finding the best path in the network to ensure its connectivity. Each routing protocol has its own standard to judge a route quality (the judging parameters include next hop number, bandwidth and delay etc. Generally these parameters are quantified with “metric” for route data). To ensure the best network path, we should select proper measurement for different network environments.

5) Management and security: Dividing the autonomous system into different areas will decrease the possibility of route cycle or route unreachable; It also makes management of network more easy. And we also should consider security and strategy of routing information transmitting. According to these principles, open and standard, as well as robust of protocols are highly considered.

2       OSPF vs. EIGRP

OSPF and EIGRP are dynamical routing protocols which have been put into application in recent years. OSPF is IGP protocols recommended by IETF, and EIGRP is released by Cisco, they both have been widely used in the network. Now customer are concerned about which one is more suitable for future network, we’ll discuss their advantages and disadvantages from technical view.       

2.1      OSPF

2.1.1          OSPF Brief introduction

OSPF is the abbreviation of Open Shortest Path First. It is an internal routing protocol of the autonomous system based on link state developed by IETF. In IP networks, it dynamically finds and propagates routes by collecting and forwarding autonomous system link state.

Each router that runs OSPF protocol always describes the local network connection state (such as valid interface information and reachable neighbor information) with LSA (link state advertisement) and advertises it to the whole autonomous system. Thus, each router receives the LSA generated by all routers within the autonomous system. The LSA collection then forms LSDB (link state database). Because each LSA is the description of the surrounding network topology of a router, the whole LSDB is then the actual reflection of the autonomous system network topology.

Based on LSDB, the routers run the SPF (Shortest Path First) algorithm. Build a shortest path tree that takes itself as the root, and the tree gives out the route to nodes in the autonomous system. In graph theory, “tree” is a connection figure without loops. Therefore, routes calculated by OSPF are born to be without loops.

To reduce self-overhead, OSPF protocol brings out following concepts:

1)     DR:

In the various multi-address access networks, if there are two or more routers, the network should elect a DR (designated router). DR responds to the LSDB synchronization of all routers in the network segment. Thus two non-DR routers need not make LSDB synchronization, which can greatly reduce bandwidth overhead in the same network segment.

2)      AREA:

OSPF can divide the autonomous system into different areas according to the topology. Thus when the area border router (ABR) transmits routing information to other areas, it generates the brief LSA with the unit of segment. It will decrease the LSA number in the autonomous system and complexity of route calculation.

OSPF adopts four classes of routes that are arranged as follows with priority:

l  Internal routing

l  Inter-area routing

l  Type one external routing

l  Type two external routing

Internal area route and inter-area area route describes the internal network structure of the autonomous system, while external route describes how to choose routes to destinations outside the autonomous system. Generally, type one external routes correspond to the information introduced by OSPF from other internal routing protocols. Costs of these routes and costs of OSFP route itself are comparable. Type two external routes correspond to the information introduced by OSPF from external routing protocols. Costs of these routes are much larger than costs of OSFP route itself, so only external costs are considered for calculation.

2.1.2          OSPF Features

OSPF is a well developed routing protocol. It is suitable for most of networks, especially enterprise network. It has features such as:

1)     OSPF is a real loop-free routing protocol ：It benefits from the algorithm itself (Link state and shortest path first algorithm).

2)     Fast convergence：Transmitting routing changing information through whole AS and recalculating routes in very short time.

3)     Support equal cost load balancing.

4)     OSPF divides the autonomous system into different areas according to the topology. Thus when the area border router (ABR) transmits routing information to other areas, it generates the brief LSA with the unit of segment. It will decrease the LSA number in the autonomous system and complexity of route calculation. So the route information will not increase very rapidly with network expanding. 

5)     Overload is as small as possible：

l  Hello packet is very short and doesn’t contain routes information. Packets containing routes information will be sent only in the case of routes changed.

l  In broadcast network, OSPF adopt multicast address to send packet, it reduces interference to other equipments which don’t run OSPF.

l  In the various multi-address access networks, if there are two or more routers, the network should elect a DR (designated router). It greatly reduce bandwidth overhead in the same network segment by decreasing times of route changing from O(N*N) to O(N).

l  Routes out of AS will not be import into stub area.

l  ABR support routes aggregation, it reduces routing information transmission between areas.

l  In point-to-point interface types, OSPF will not send hello packet on time until network typology has really changed.

6)   OSPF adopts restrict four classes of routes to provide more reliable routes choice.

7)   OSPF support two types of packet authentication modes. One is the common clear text authentication mode; the other is the cipher text authentication mode with MD5 algorithm.

8)   OSPF is suitable for any size network ，and in can support thousands routers at most.

9)   OSPF can expand to support Traffic engineering because of link-state awareness.

2.2      EIGRP

2.2.1          EIGRP Brief Introduction

EIGRP and early IGRP are released by Cisco. They’re both distance vector protocols. EIGRP is enhanced edition of Interior Gateway Routing Protocol. Though it adopts distance vector algorithm, it has some features of link state protocol. EIGRP has improved a lot compared to IGRP; it relies on the Diffused Update Algorithm (DUAL) to calculate the shortest path a destination within a network. It is totally loop-free, and has very fast convergence speed among all the routing protocols. 

2.2.2          EIGRP Features

EIGRP has features as below:

1)  Accurately routing load calculating and heterogeneous network protocols supporting. EIGRP inherits advantages of IGRP. EIGRP calculates routes according to information such as network bandwidth, total delay, path reliability, path loading, so the routes table is more accurate. EIGRP also support IPX，CLNP。

2)  Low usage of network resource. During normal operation, usage of network resource is very low; only hello packets are transmitted on a stable network. When a change occurs, only routing table changes are propagated, not entire routing table; this reduces the load the routing protocol itself places on the network. EIGRP also can control the packets transmission and reduce the usage of interface bandwidth, so it can avoid influence to normal services data packets.    

3)  Loop-free and fast convergence. EIGRP uses DUAL , only routing table changes are propagated; and to one route ,only relative routers will recalculates。

4)  The cipher text authentication mode with MD5 algorithm is supported.

5)  Variable Length Subnet Mask routes aggregation. EIGRP support Variable Length Subnet Mask routes aggregation by configuration, is reduces transmission of routing information and save bandwidth.

6)  Support load-balance over equal cost or unequal cost .EIGRP can send traffic in proportion over several unequal cost paths, this promotes the utility rate of network resource; but is also increase workload of routers, so this way is not commended even by Cisco.

7)  Configuration is simple. There’s no complicated area setting and it need not adopt different configuration to different network interface. It only needs to start EIGRP process on routers, and uses network command to configure interface.

3       Compare OSPF to EIGRP

Both OSPF and EIGRP are fast convergence routing protocols, both are using algorithm which are loop free, secure, and take up small bandwidth. Nevertheless, from the analysis of the previous chapters, we can see that each has it own strong points and weak points.

3.1.1          Disadvantages of OSPF

1)  complexity of configuration：because of the complexity of network attribute and dividing areas when running OSPF, the network administrators need to possess solid knowledge of data communication and computer networks in order to make OSPF working well, with OSPF getting more and more popular, this is not considered a big problem.

2)  can not support unequal load balance：OSPF creates the metric of a link based on the bandwidth of the link by default, OSPF only picks path with the smallest metric towards the same destination (OSPF supports equal metric load balance). OSPF does not support unequal load balance. This is not like EIGRP, which supports unequal path load balance by configuration.

3.1.2          Disadvantages of EIGRP

1)  there is no area in EIGRP，so it is not good at dealing with big hierarchy network. When running OSPF on a big network, we can make the network hierarchy by dividing the network into some areas.  Obviously, EIGRP is not a good choice for a big network. This is also a restriction of distance-vector routing protocol (like RIP, RIPII). If EIGRP be a routing protocol for a big network, we can separate the network into different EIGRP domain, then import routing table to each other, but it is not a optimal network design, and very few network has been designed like this.

2)  does not support DoD：EIGRP maintains the adjacency relationship by sending HELLO message to each other periodically, even though running on dial up link. However, the HELLO message may bring the dial link up, this is not what we want on a dial up link, and especially the dial up link is a backup link. When we run EIGRP on a dial up link, in order to prevent this from happening, we usual put a Dialer list and Dialer group on the interface so as not to let the HELLO message bring the dial link up. By doing this way, we sacrifice the router resource. OSPF takes advantage at this point by supporting DoD, Dial on Demand.

3)  The fast convergence and loop free characteristics are based on the EIGRP DUAL algorithm. Basically, the DUAL algorithm is working by sending query to its neighbor about the active routes(uncertainty routes), then convergence upon receiving the reply. If the routes are uncertainty routes for its neighbors too, the neighbors send out query to their neighbors, the process will be going on and on until get the reply or after a certain time, the routes will be considered not available and get purged from the routing table. Thus in some cases, the active routes will be put into “stack in active” status for quit a long time, this affect the fast convergence seriously. OSPF does not have this problem. Although EIGRP is also fast convergence routing protocol, when working on some special topology, EIGRP is fairly slow. for instance, in a long and narrow network, if something has changed, it would take EIGRP a long time to send the message from one side to the other side.

4)  In a broadcast network, EIGRP sets up a full mesh adjacency relationship with each other, the routers exchange information with other. This would waste a lot of bandwidth. OSPF does not work like this. OSPF elects DR and BDR instead. The DR other routers only need to set up adjacency with DR, and exchange link state advertisement with DR in the network. That will save a lot of bandwidth.

5)  EIGRP is a protocol come up with by Cisco, it is a private protocol, not a open standard, Cisco is the only company who has the right to use it and make change of it, Cisco has the right to make any change of the protocol as they want without having to inform any customers and other vendors, this would be a big security concern for customers. Besides, If customers choose to run EIGRP on their network, they are no way to choose other vendors products when they upgrade their network afterwards. This is unfair both for customers and other vendors. On the contrary, OSPF is a open standard routing protocol, come up with by IETF. Most the mainstream vendors in this industry support it, so the compatibility among different vendors is guaranteed. Under the support of many vendors, the protocol will be getting better and better.

3.1.3          Summary of comparison between OSPF and EIGRP

The following form is the summary of compare of OSPF and EIGRP：

Compare point	OSPF	EIGRP
standard	Open standard of IETF，supported by most vendors.	Cisco owned private routing protocol，not been supported by any other vendors；is not as mature as OSPF.
popularity	Most popular IGP in the world	Only a few networks designed by EIGRP，and is getting less and less popular.
algorithm	SPF algorithm fast convergence, loop free.	DUAL algorithm could be in SIA status, query could spread out the whole network.
topology	Can build a hierarchy and scaleable network.	Can not build a hierarchy network with this protocol.
Supportive of new technology	Support OSPF-TE	Does not support TE.

4       How migrating EIGRP to OSPF

It’s always been a big concern for customers to migrate an EIGRP network to an OSPF network smoothly without breaking the current network when they upgrade or rearrange their network. The easiest way is take off the EIGRP protocol and put on OSPF protocol. However, doing this way would break the current network. This is not the way it is supposed to be.

Actually, smooth migrate a routing protocol to another routing protocol is not that difficult. If we follow the right procedure, we can make a perfect migration without breaking the current network, all we need to do is the configuration change.

The main idea of migrating a routing protocol to another routing protocol is utilizing the different precedence of different routing protocol(it is called preference in Huawei VRP, and it is called distance in Cisco IOS), when we run multi routing protocols on a network, the working routing protocol is picked by their precedence, the less, the better. The following is the procedure of migration:

1)     Design the OSPF network, like how to divide the areas, what’s the routing policy.

2)     Configure OSPF routing protocol on the EIGRP network, make the precedence of OSPF lower than precedence of EIGRP, then put the interfaces into OSPF areas.

3)     Complete the configuration followed the second step, because the precedence of EIGRP is higher than OSPF, so only the EIGRP learned routes show up on routing table, OSPF learned routes do not show up on the routing table. So OSPF does not work at this time around.

4)     After making sure everything is fine, changes the precedence of EIGRP and OSPF, and makes the precedence of EIGRP lower than OSPF. Because the precedence of OSPF is higher than EIGRP, so the routing table change to OSPF learned routes, EIGRP learned route will disappear, and OSPF will be the working routing protocol.

5)     After having made sure the OSPF routing table is perfect, we can delete the EIGRP configuration, and change the precedence of OSPF back to default value. After have done that, the migration is completed.

By following the previous procedure, we can fulfill the migration without breaking the current network. Because we keep two routing protocols on the network, so if something is wrong with OSPF, something is wrong with OSPF learned routing table, we can change back the precedence of EIGRP lower than OSPF, by doing this way, we don’t take the risk of breaking the current network.