Border Gateway Protocol And Routing Scalability

Discuss about the Border Gateway Protocol and Routing Scalability.
 

RIPv2 is an internal routing protocol that is meant for smaller networks. RIPv2 is based on the original RIPv1 protocol which is now obsolete because of its limitations and deficiencies. RIPv2 is a distance vector routing protocol according to RFC 1723 definition. It is classless meaning that in its routing updates, the subnet mask that is within the network is included. Unlike RIPv2, RIPv1 is classful thus it supported only the networks without any subnets.

RIPv2 is based on RFC 1388, 2453 and 1723 (ortbitco, 2015) and it has the capability to support a network with more than 15 hops (number of routers that can be traversed in a route) and anything with more than 15 hops is considered to be infinity thus considering the route as invalid. A router using RIPv2 routing protocol sends its full routing table to all the other routers that are connected every 30 seconds and the connected routers have to send their complete routing table too. Because of this exchange in routing tables, calculation of routes does not happen and no real-neighbor relationship exists thus all the routers have firsthand information of available networks. If a router goes down before expiration of the 30 seconds a triggered update is done. Routing in RIP is rumor oriented thus is more prone to loops as compared to other routing protocols. Some of the advantages of using RIPv2 are;

• RIPv2 is compatible with a diverse range of network devices.
• It’s most suitable for smaller networks because of the minimal overhead in terms of the configuration, management time and the bandwidth used.
• RIPv2 has support for legacy host systems.

Implementation or configuration is simple and straightforward and is done through the following three steps;

1. The first step is to enable RIP using the global configuration router rip
2. The second step is to instruct the router to use RIPv2 by use of version 2
3. The third step is instructing the RIP on which networks it’s supposed to advertise by use of a series of network commands.

Step 1 and step 2 are easy to perform but step 3 requires a bit more time. The network commands used in step 3 are used to specify the interfaces that will participate in the routing process by taking a parameter of a classful network and enabling RIP on the corresponding interfaces (Davis, 2006).

Since the algorithm used by RIPv2 to calculate best paths to remote networks does not include any loop prevention mechanisms, RIPv2 has additional loop prevention mechanisms built into it. These mechanisms include;

• Counting to infinity- this mechanism happens as described above where by if the number of routers traversed exceeds 15 i.e. more than 15 hops is considered to be infinity thus that route is invalid.
• Split horizon- take routes that are learnt and installed into the RIPv2 routing table and then advertise to all the other RIPv2 participating interfaces excluding the interface that the route was learnt from.
• Split horizon with poisoned reverse-similar to split horizon except that router advertising to all the other parts is done in an infinite metric.
• Triggered update- RIPv2 sends an update immediately after routing information changes. It does not wait for the expiration of the update-timer.
• Invalid after timer- this is timer per route that is reset and will begin after an update of a route is received. Its default time is 180 seconds. If routes from the source experience failure and no updates are received the invalid after-timer is prompted and if no update is received by the end of the invalid after-timer the hold own timer is triggered (ComputerNetworkingNotes, 2018).
• Hold-down timer- This timer for every route that begins after the route is declared invalid. The router starts to advertise the route as being unreachable and ignores any update information and ignores routing entry modifications until the hold down timer expires.
• Flushed after timer- this is a timer for every route that is reset and is started after an update of a route is received. If no update is received and the timer times out the route is removed from the routing table.

Because RIPv2 sends automatic periodic updates every 30 seconds it tends to consume more CPU and bandwidth resources if it’s used for medium or large networks because of the large routing tables. But when used for small networks this is not an issue as it is best suited for smaller networks.

This introduces scalability issues because the bigger the network gets the bigger the routing table gets and the more CPU and bandwidth resources are used to send periodic updates. 

Implementation

RIPv2 is best used for smaller networks for example a small office block with a small network due to its scalability issue discussed in section 1.4 above.

OSPF is initials for Open Shortest Path First. It is a routing protocol where connected routers exchange information about the routes they have information on and their cost of getting there. For a large network, all the routers running OSPF protocol have to learn about all the routes. Each router shares the information with its adjacent neighbor and information is passed along to all OSPF routers. Instead of sending routing table updates OSPF uses the concept of sharing link-state-advertisements (LSA) thus the network converges in a timely manner (Banks, 2014).

The two main OSPF interfaces are;

• OSPF broadcast interface- this is interface is connected to shared networks e.g. Ethernet
• OSPF point-to-point – interface is connected to a link with a single OSPF router on each end e.g. WAN.

Interface types are used to make sure that all the routers know about all the routes from the other routers. For point to point, the two OSPF routers exchange routes. For broadcast, there are many OSPF routers thus neighbor relationship is minimized by electing a designated router to neighbor with all the other router and to share everyone’s routes with all the routers.

Operations of OSPF consists of three main elements;

• Neighbor discovery
• Link-state information exchange
• Best-path calculation.

For best-patch calculation, OSPF uses the SPF (Shortest Path First) algorithm which is also known as Dijkstra’s algorithm. SPF calculation is done using link-state information inputs. Implementation of OSPF is done by following these steps (Teare, Graziani & Vachon, 2018);

• Coming up with a complete OSPF strategy and plan for the proposed network e.g. deciding on the number of areas.
• Enabling IPV6 on the interface.
• Configuring authentication. 

Convergence capabilities of OSPF provides redundancy and resiliency to circuit failures. OSPF routing convergence has two components (Ruhann, 2018);

• Detection of changes is topology- This happens in two ways. The first way is failure or status change on the physical interface e.g. loss of the carrier. The second way is a timing out of the OSPF hello timer. An OSPF neighbor is assumed to have failed if the waiting time for a hello packet exceeds the dead timer which is four times the value of the hello timer. Hello timer is 10 seconds for broadcast and 30 for non-broadcast.
• Recalculation of routes- Every router performs recalculation of routes after a failure is detected. A LSA is broadcasted to all routers in the OSPF area to signal a topology change. The resulting action is recalculation of routes using SPF and this is uses a lot of CPU resources especially for a large network.

Convergence can be increased by decreasing the value of the hello timer if a link goes down and is not detected by layer 2. However the timer should not be set too low to avoid unnecessary recalculations of topology.

If a change in topology is detected, LSA is automatically generated and sent to all the devices in the network. Routes recalculation does not occur until the SPF timer expires. SPF timer has a default value of 5 seconds. To delay consecutive SPF calculations SPF hold time is used and ha a default value of 10 seconds. This results to the minimum time for convergence of routes in case of a failure being more than 5 secs less only if SPF timers tuning is done using OSPF throttle timers.

Routing overhead in OSPF occurs mainly due to route recalculation which is done by each router after a failure is detected. For a large network that has unreliable links recalculation causes CPU overload. Thus the factors that affect the scalability of OSPF are;

• Count of the adjacent neighbors for every single router- this is because in case of a failure OSPF floods LSA to all routers in that area thus router with most neighbors have a lot of work to do.
• Count of adjacent routers in an area—the larger and more unstable the area is the more chances of performance problems associated with recalculation of the routing protocol.
• Number of areas a single router supports- Because of running LSA for each change in link-state one router should be restricted to not less than 3 areas.
• Designated router selection- A good practice is to select routers that are free in terms of CPU usage instead of selecting the ones that are already overloaded.

Convergence

The best application for OSPF protocol is enterprise networks that are running on multiple sites for example a university with many campuses.

EIGRP is a routing protocol based on the original IGRP (Interior Routing Protocol). EIGRP was designed to improve on the deficiencies and weaknesses of IGRP. It’s owned by Cisco. EIFRP improved on the easiness of configuration, speed and reliability.

EIGRP is based on a distance vector algorithm that is used to determine the path to a destination. It’s similar to RIP but uses a more complex metric as compared to RIP’s hop count. The metric used by EIGRP is based on the net delay and minimum bandwidth along each possible path thus it has an ability to support larger networks. Support for large networks while achieving high efficiency and reliability can be attributed to its concept of sending non periodic updates which contain information only routes that have changed instead of sending the whole routing table as in the case of RIPv2. Exchange of hello packets is what happens the rest of the time there is no updates for verification of availability of their routing peers. For this reason EIGRP uses very low bandwidth (orbitco, 2015).

All routers in an EIGRP network include a topology table that is a central feature of the DUAL algorithm. The topology table is updated every time a router receives a piece of information thus giving it an up-to-date and reliable image of every connection in the network that is in use. 

The steps involved in implementing EIGRP are;

• Configuration of EIGRP on the network.
• Redistribution of routes into EIGRP
• Creating of the default route in EIGRP.

EIGRP uses rapid convergence through the use of the DUAL (Diffusing Update Algorithm). This is achieved by making sur that a router that is using EIGRP keeps a storage of all available backup routes so that in case of failure, the router can quickly adapt to an alternative route. If there is no alternative back up route in the local routing table the router sends an enquiry to its neighbors to get an alternate route.

EIGRP achieves less overhead as it used unicast and multicast as opposed to using broadcast thus end stations are not affected by topology information requests and routing updates. This means that EIGRP is very scalable as it can be expanded to support large networks without compromising on performance and efficiency as it uses unicast and multicast. EIGRP also supports unequal metric load balancing. This in turn allows better distribution of traffic flow in the network.

Routing overhead and scalability

The best type of network to use with EIGRP is a large network that requires a lot of scalability for example a WAN.

BGP is an advanced and complex distance EGP (Exterior Gateway Protocol) (Orbitco, 2015). It uses BPSA (Best Path Selection Algorithm) to find and save the best routes in the routing table. Unlike the other protocols discusses in the sections above BGP does not run on a network of a company but is mostly used in exchange of information between Internet Service Providers or large clients and the ISPs. For large enterprise networks, it’s used to interconnect administrative or geographic regions.

There are different types of BGP;

• Internal BGP also called iBGP and operates inside autonomous system (AS).
• External BGP also called inter-domain routing protocol. Its operates outside an autonomous system and connects one autonomous system to another

These two types of BGP use the same protocol. The only difference is where they are applied for use. An autonomous system can be an ISP, an entire corporate network, or even a company. The key characteristics of BGP are;

• The distance-vector protocol used by BGP is more advanced.
• At the start of the session, BGP sends full routing updates then trigger updates are sent later on.
• To maintain connection BGP sends periodic keepalives.
• Connection between peers is maintained using TCP, port 179.
• When a notification, keepalive or update is not received BGP sends a triggered update.
• BGP has its own routing table. However it’s still capable of inquiring and sharing the interior IP routing table.
• The source of the strength for BGP is its complex metric which is referred to as attributes. The metric gives more flexibility during path selection.

Configuration of BGP uses a neighbor-based approach where all configurations for a single neighbor should be grouped in one place under the configuration of that neighbor. Peer groups are not supported for sharing of update messages or for configuration sharing between neighbors. BGP has set of configuration groups that are used as templates to replace the concept of peer groups.

Since BGP is a distance vector protocol, BGP routing accepts many routing updates and only shares the best routes to its peers. Because it does not use the concept of sending periodic updates, BGP does an explicit withdrawal in the triggered UPDATE message and this signals the neighbors about a failed route (Lapukhov, 2013). BGP also sends signals implicitly if it encounters the same prefix from a previously learned information then it is replaced.

The scalability of BGP is mainly affected by the need send updates to peers and to store and process updates received from peers. Bandwidth is not a problem but memory requirements poses challenges when storing and processing large BGP tables (Beijnum, 2015). Thus an increase in the number of updates requires more processing power and memory.  

References

Beijnum, I. (2015). Border Gateway Protocol and Routing Scalability. Retrieved from https://www.techopedia.com/2/31630/networking/border-gateway-protocol-and-routing-scalability

Banks, E. (2014). Open Shortest Path First OSPF Protocol Explained. Retrieved from https://www.auvik.com/media/blog/ospf-protocol-explained/

ComputerNetworkingNotes. (2018). RIP - Routing Information Protocol Explained. Retrieved from https://www.computernetworkingnotes.com/ccna-study-guide/rip-routing-information-protocol-explained.html

Davis, D. (2006). Cisco administration 101: Know the basics about RIPv2. Retrieved from https://www.techrepublic.com/article/cisco-administration-101-know-the-basics-about-ripv2/

Lapukhov, P. (2013). Understanding BGP Convergence. Retrieved from https://blog.ine.com/2010/11/22/understanding-bgp-convergence/

Orbitco. (2015). What is BGP? Explained with Examples - orbit-computer-solutions.com. Retrieved from https://www.orbit-computer-solutions.com/border-gateway-protocol-bgp/

Orbitco. (2015). What is EIGRP? Explained with Examples. Retrieved from https://www.orbit-computer-solutions.com/enhanced-interior-gateway-routing-protocol-eigrp/

Ortbitco. (2015). What is RIPv2? Explained with Examples. Retrieved from https://www.orbit-computer-solutions.com/ripv2/

Ruhann, V. (2018). OSPF Convergence. Retrieved from https://routing-bits.com/2009/08/06/ospf-convergence/

Teare, D., Graziani, R., & Vachon, B. (2018). OSPF Implementation > Establishing OSPF Neighbor Relationships. Retrieved from https://www.ciscopress.com/articles/article.asp?p=2294214