2.5 Routing
The main features of a router include the following: -Routers work at the network layer. Routers are able to identify source and destination network addresses within packets.
-A router is able to keep track of multiple active paths between any given source and destination network. This makes it fault tolerant.
This is devided up into the following general hierarchy: -The globally unique unicast address range is indicated by the 2000::/3 address space (the first 3 bits). Ranges from this address space are allocated to RIRs in blocks from /3 to /32, as listed at gtsgo.to/48het.
-ISPs received allocations from their registry in the space from the /32 to /35. -End users receive allocations from their ISP in the /48 to /64 range. -End users can subnet their networks using the remainder of the network prefix left to them (if any).
The main features of a router include the following(Continued): -Routers provide excellent traffic management using sophisticated path selection; they select the best routes based on traffic loads, line speeds, number of hops or administrator pre-set costs. The parameters used for determining routes for packets are called metrics.
-Routers can share status and routing information with other routers and can listen to the network and identify which connections are busiest or not working. They then route network traffic avoiding slow or malfunctioning connections.
Support rapid convergence -convergence is the process whereby routers agree on routes through the network.
As the network topology changes constantly (what with router failures, addressing changes, and unforeseen events), routers must be capable of adapting to these changes and communicating them quickly to other routers to avoid loops. A network where all the routers share the same topology is described as steady state. The time taken to reach steady state is a measure of a routing protocol's convergence performance.
The relatively complex tasks performed by a router mean that they tend to be processing intensively.
A router may be a dedicated appliance with a port to each of the networks or it may be a NOS server with multiple interface cards (multi-homed). Routers very often also support the functions of a firewall.
Some dynamic routing algorithm types include:
Multipath Hierarchical Host-intelligent Distance vector Link-state
RIPng is an update to support IPv6.
RIPng uses UDP port 521.
Bandwidth
metrics based on bandwidth look at the aximum achievable bandwidth on a link and do not consider the available bandwidth. This is a less efficeient metric than delay-based metrics.
BGP works
over TCP on port 179.
Reliability
over a period of time, it might become obvious that some links between routers are more reliable than others. You can assign an arbitrary value for this reliability that routers can assess when determining an effective path.
Link-state
routers implementing a link-state algorithm propagate information about their own links to other routers on the internetwork. These smaller, frequent updates lead to more rapid convergence and more efficiently support larger networks. However, they are more processor intensive.
In either protocol, IPv4 or IPv6,once the IP layer has established that the destination host is on a different network,
routing must take place.
RIP sends regular updates about the routing table to neighboring routers plus ad-hoc updates whenever changes occur.
When a router receives an update from a neighbor, it updates the appropriate route in its own route table, increases the hop to the specified network. The router then propagates the update.
If the packet has been routed the Time to Live (TTL) is decreased by at least one. THis could be greater if the router is congested.
When the TTL is zero, the packet is discarded. This prevents badly addressed packets from permanently circulating the network.
Routers that can connect to muliple areas are known as Area Border Routers.
A backbone (always called Area 0) is created by the collection of border routers. This backbone is only visible to the border routers and invisible to the routers within a specific area.
Multiple Routing Protocols
A complex network may need to run more than one routing protocol.
High-level network routing prefixes (or CIDR blocks), which are 8 bits in length (//8s), are allocated by IANA to Regional internet Registries (RIR), such as ARIN (America) and RIPE (Europe).
A few are still held privately by companies such as IBM, Xerox, HP, or AT&T or by governemt agencies such as the DoD. You can view the assignments at gtsgo.to/yqu0j.
Open Shortest Path First (OSPF)
A hierarchical link-state protocol, Open Shortest Path First (OSPF) is better suited to large organizations with multiple redundant paths between various networks. It has high convergence performance compared to RIP and better scalability compared to EIGRP. It was designed from the outset to support classless addressing.
Path Selection All routers choose an appropriate path through the network using some sort of evaluation process of all possible paths available. Metrics are used in this process.
A metric is a unit of measurement; for example, how are the next hop router is,how long it will take to route a packet to the subsequent routers, what bandwidth is available on the selected path, how large a packet can be sent without fragmentation, and so on.
An Interior Gateway Protocol (IGP) is one that performs routing within a single private network, also referred to as an Autonomous System (AS).
An Exterior Gateway Protocol (EGP) is one that can perform routing between autonomous systems.
The registries then allocate blocks to national and local registries. Actual ISPs are generrally allocated blocks with prefixes of 20 bits or less.
Any routing over that boundary (that is, a /21 network or higher) takes place solely within the ISP's network rather than over the general internet. The ISP's network is referred to as an Autonomous Sytem.
Bgp work with CIDR IP network prefixes (Network Layer Reachability Information [NLRI]) Route selection is based on multiple metrics including hop count, weight, local preference, origin, and community.
BGP is no a pure distance-vector algorithm but uses a hybrid approach.
Performance All algorithms are designed to determine paths and switch packets across those paths. To achieve these goals, algorithms should:
Be efficient in selecting the best route Be efficient in routing packets Be reliable and flexible Support rapid convergence
Autonomous systems are designed to hide the complexity of private networks from the public internet. If all internet locations had to be propagated to internet routers, the routing tables would become too large to process.
Border (or edge) routers for each AS exchange only as much network-reachability information as is required to access other autonomous systems (the AS path), rather than networks and hosts within each AS. Autonomous System Numbers (ASN) are allocated to ISPs by IANA via the various regional rigistries.
To help prevent looping, the maximum hop count allowed is 15.
Consequently, this limits the maximum size of a RIP network as networks which have a hop count of 16 or higher are unreachable.
Like RIP, EIGRP is usually classed as a distance vector-based routing protocol. Unlike RIP, which is based on a simple hop count metric. EIGRP uses a metric composed of several administrator weighted elements, including reliability, bandwidth, delay, and load.
EIGRP also supports multiple paths to the destination network, again, unlike RIP. EIGRP may also therefore be described as an advanced distance vector protocol or as a hybrid routing protocol.
A router designed to connect a private network to the internet is called an edge router or border router. These routers can perform framing to repackage data from the private LAN frame format to the WAN internet access frame format.
Edge routers designed to work with DSL or cable modems are called SOHO Routers (Small Office or Home Office).
Routers are responsible for two functions in an internetwork.
Firstly, they must choose a route, and secondly, they must deliver the network packets to the destination using the selected route.
Routing Protocols
For larger organizations in is simply not practical to configure routing tables manually. Aside from anything else, routing information is seldom static as routers are reconfigured, taken temporarily offline, and even decommissioned.
Because it is widely adopted, well understood, and simple, RIP is ideally suited to small networks with fairly limited failover routes.
For more complex networks, with redundant paths, other network routing protocols should be considered.
Routing Information Protocol (RIP) is a long established distance vector-based routing protocol. It uses a hop count metric to determine the distance to the destination network.
Generally speaking, each router is assigned a hop count value of 1. RIP only considers a single route to a given destination network - that with the lowest hop count.
End Systems and Intermediate Systems All IP hosts are, in essence,capable of functioning as routers.
However, most workstation and server computers are configured with a single network adapter connected to only one network; although potentially capable of routing, they are not equipped with the necessary interfaces and knowledge of the location of other networks.
Different routing algorithms, as implemented in the various routing protocols, may use different metrics and make comparisons of available paths in different ways.
However, the desired end result is always the same - to choose the optimal path for a specific packet at a given moment.
Enhanced Interior Gateway Routing Protocol (EIGRP) Interior Gateway Routing Protocol (IGRP) was developed by Cisco to provide a routing protocol for routing within a domain, or autonomous system (that is, within a single organization).
IGRP was seen as a possible alternative to organizations limited by the inherent restrictions imposed by RIP, such as the hop count limit of 15.
Routing Basic The process of routing takes place when a host needs to communicate with a host on another network or in another subnet.
IP is able to determine that the target host is on a different network by deriving the network address from the full IP source and destination addresses and comparing them. The network address is obtained by masking the full IP address against a network prefix or subnet mask.
If the former, then the router consults its routing table, and determines the next hop router.
If no route exists, the packet is either forwarded to the default gateway of the router (also referred to as the default route or gateway of last resort) or dropped (and the source host is notified that it was undeliverable).
If so, it updates the frame with the MAC address of the destination host and merges the new frame onto the appropriate interface.
If not, then the router determines the appropriate path, and selects a router on that path, and inserts the next router's MAC address into a new frame (containing the original IP packet) and merges it onto the wire attached to the appropriate interface for the next hop router.
IPv4 and IPv6 Internet Routing Internet routers must be able to locate any host on the internet. As there are millions of networks and hosts, it is impossible to do this by storing routes to each of them.
Instead, internet backbone routers store routes only to large networks, as identified by their network routing prefix. This is referred to as route aggregation or summarization.
In IPv4, routers can be made responsible for calculating the Maximum Transmission Unit (MTU - or datagram size) for a given interface and fragmenting and reassembling datagrams that are too big.
In IPv6, the host is responsible for determining the MTU and routers cannot perform fragmentation.
There are, in fact, three implementations of RIP. RIPv2 provides for a level of authentication between RIP routers and uses more efficient multicast transmissions rather than broadcasting updates.
In addition, RIPv2 packets carry a subnet mask field and therefore support classless addressing.
Each router stores information in three tables:
Neighbor Table Topology Table Routing Table
The version in current use is OSPFv2;
OSPFv3 provides support for IPv6.
Routing Metrics Different routing algorithms use different metrics to help determine the appropriate path to use. These metrics might include:
Path length Reliability Delay Bandwidth Load (link utilization) MTU (Maximum Transmission Unit) Price
Most routing takes place using dynamic algorithms, encoded in a routing protocol.
Router use these protocols to exchange information about connected networks periodically and select the best available route to a particular destination.
Hosts are incapable of forwarding packets to other subnets, and therefore referred to as End Systems (ES). Routers that interconnect subnets and can perform this packet forwarding process are known as Intermediate Systems (IS).
Routers are simply IP hosts configured with multiple network interfaces and knowledge of the location of other networks. Information about the location of other networks is usually stored in a routing table of some sort. Early routers were manually configured with this routing information.
Router Appliances
Routers designed to service medium to large networks are complex and expensive appliances. They feature specialized processors to handle the routing and forwarding processes and memory to buffer data. Most routers of this class will also support plug-in cards for WAN interfaces.
Networks and their connected hosts and routers withing an autonomous system are grouped into OSPF areas.
Routers within a given area share the same topological database of the networks they serve.
Some of the most popular routing protocols are listed in the table below.
Routing Information Protocol (RIP) Enhanced Interior Gateway Routing Protocol (EIGRP) Open Shortest Path First (OSPF) Border Gateway Protocol (BGP)
Remember that both subnetting and supernetting require the use of a classless routing protocol (one that does not determine the network mask on the basis of the first octet in the IP address)
Routing protocols that support classless addressing include RIPv2, EIGRP, OSPF, and BGPv4.
Hierarchical In hierarchical routing systems, certain routers form a routing backbone. Other routers are grouped into logical collections, sometimes called areas or domains.
Some routers can communicate with routers in other domains, whilst the remainder are limited to communication with routers in the current domain. Non-hierarchical systems are referred to as flat; a situation in which all routers can inter-communicate with one another.
Default AD values are coded into the router but can be adjusted by the administrator if necessary:
Source - AD Local interface / Directly connected - 0 Static route - 1 BGP - 20 Internal EIGRP - 90 OSPF - 110 RIP - 120 External EIGRP - 170 Unknown - 255
Types of Routing Algorithms The algorithms used for path selection can be categorized according to the topology and metrics that they use to build and update routing tables.
Static routing is defined manually. You create the routing entries in the router's memory and they only change if you edit them. Routers configured this way are not flexible and do not support rapid convergence, but they might suit a small internetwork.
Whilst many organizations can manage to maintain routing tables for these internal changes, when connected to the internet, it becomes almost impossible. Consequently, router vendors provide support for routing protocols.
These routing protocols use various algorithms and metrics to build and maintain routing tables to provide reasonably current routing information about the networks to which they are connected.
IPv6 follows the same hierarchical structure, with the advantage of planning an efficient addressing topology from the start and having a larger address space to work with.
The full network prefix of an IPv6 address is 64 bits long.
Host-intelligent
in host-intelligent algorithms, it is assumed that end systems (hosts) can perform path determination and routers act as store-and-forward dev
In IPv4, the network prefix can be increased (to make more network numbers available) at the expense of the number of host IDs left available for each network.
The network prefix can be expressed in slash notation as a number of bits (172.16.0.0/20 for instance) or as a dotted decimal subnet mask (255.255.240.0 for instance).
Packet Delivery is the physical process of transmitting the packet to the destination host. As we have seen, when a host determines that it must route a packet, it determines the MAC address of the nearest router (its default gateway) and merges the data link frame onto the wire.
The router interface with the corresponding MAC address picks up the frame, and tries to determine if the destination host is directly attached to any of its interfaces.
Routing Algorithms and Metrics
The routing protocol you implement depends on a variety of factors. As different protocols support different routing algorithms, it is worth spending some time considering the different algorithms used.
Classless IPv4 routing uses a mix of flat and hierarchical structures to make more efficient use of the limited address space (compared to the old method of classful address allocation).
The system is based on the Classless Interdomain Routing (CIDR) specification.
This means, for example, that a static route will always be preferred to anything other than locally connected networks and that a route discovered by EIGRP would be preferred to one reported by RIP.
The value of 255 for unknown routes means that they will not be used.
Like EIGRP, messages are sent using OSPF's own datagram format. This is tagged as protocol number "89" in the IP datagram's protocol field.
There are various packet types and a number of mechanisms to ensure sequencing and reliable delivery and to check for errors. OSPF also supports plain text or cryptographic authentication.
Limitations in IGRP, such as lack of support for classless addressing, led to the development of Enhanced IGRP (EIGRP).
There are versions for IPv4 and IPv6. IGRP itself is now obsolete.
Router Placement As mentioned above, routers serve both to join physically remote networks and subdivide a single network into multiple subnets. Routers that join different types of network are called border or edge routers.
These are typified by distinguishing external (internet facing) and internal interfaces. These devices are placed at the network perimeter .
Routers use a Link State Advertisement (LSA) to update their routing tables. In a given area, routers exchange OSPF Hello messages, both as a form of keep-alive packet and in order to acquire neighbors with which to exchange routing information.
These exchanges or routing information enable the routers to each build a topological routing tree (a shortest-path tree) and keep it up-to-date. The use of areas to subdivide the network minimizes the amount of routing traffic that has to be passed around the network as a whole, improving convergence performance.
Once a router has received a packet, it goes through the same process that the source host did to calculate whether the packet needs to be routed to another router or whether the packet can be delivered locally to another interface (i.e. an interface other than the originating interface).
This is referred to as a directly connected route. If the latter, the router uses ARP (IPv4) or Neighbor Discovery (ND - IPv6) to determine the interface address of the destination host and merges the packet onto the appropriate interface.
The main features of a router include the following(Continued2): -Routers do not forward any information that does not have a correct network address. For this reason they do not forward bad data. They also filter broadcast traffic by not routing broadcast packets.
This means network broadcasts do not propagate throughout the internetwork and that broadcast storms are confined to a single subnet.
Be reliable and flexible
intermediate systems sit between networks and their failure may lead to significant problems. It is important that the routers are reliable and possibly fault tolerant. In addition, they must be capable of dealing with other failures on the network and still function within defined operational parameters.
The graphic below shows a simplified example of a typical network configuration. Basic switches provide ports and Virtual LANs (logical groupings of clients) for wired and (via an access point) wireless devices.
Traffic between logical networks is controlled by a router (or layer 3 switch). A WAN router provides access to the internet.
Protocol - Border Gateway Protocol (BGP)
Type- Distance-vector Class- Exterior Gateway Protocol (IGP) Transport - TCP (port 179)
Protocol - Routing Information Protocol (RIP)
Type- Distance-vector Class- Interior Gateway Protocol (IGP) Transport - UDP (port 520 or 521)
Protocol - Enhanced Interior Gateway Routing Protocol (EIGRP)
Type- Distance-vector (Hybrid) Class- Interior Gateway Protocol (IGP) Transport - Native IP (88)
Protocol - Open Shortest Path First (OSPF)
Type- Link-state Class- Interior Gateway Protocol (IGP) Transport - Native IP (89)
Delay
it takes time for a packet to traverse an internetwork. Delay-based metrics measure transit time (latency).
Where a router is configured to run multiple routing protocols,
it uses a metric called Administrative Distance (AD) to determine which protocol to "trust" when presented with alternatives routes to a network.
Installing and Configuring Routers A router is the device that connects multiple networks and routes packets from one network to another (internetwork).
You should note that networks can be distinguished by different physical locations or by separate logical topologies (such as subnets). In the former case, routers join networks together; in the latter they can subdivide a single physical network to conform to a logical topology.
Load (link utilization)
a metric that bases routing decisions on how busy a particular route is.
These days, almost all routers use some sort of routing protocol to learn about remote networks
and the most efficient route to those networks.
RIP uses the User Datagram Protocol (UDP)
as the network transport protocol (port 520).
Border Gateway Protocol (BGP) is designed to be used
between routing domains, or Autonomous Systems (AS), and as such is used as the routing protocol on the internet, primarily between ISPs.
Neighbor Table
contains address and link state information about other routers to which there is a direct interface.
Topology Table
contains information about the wider network build from information obtained from neighbors, including hop count and link state information. The best path to a given network (called a successor) is built by DUAL (Diffusing Update ALgortihm). An alternative, backup path is referred to as a feasible successor.
Distance vector
distance vector algorithms require that routers propagate their entire routing table periodically to their immediate neighbors. Distance vector algorithms provide for slower convergence than with link-state algorithms..
All the protocols we have discussed so far have been classed as Interior Gateway Protocols (IGP) used for communications between routers within an autonomous system. BGP is a type of Exterior Gateway Protocol (in fact, it replaced a protocol named EGP)
for commuications between routers in diverse autonomous systems. When BGP is used withing an autonomous system, it is referred to as Interior BGP (IBGP) and when implemented between autonomous systems, it is referred to as Exterior BGP (EBGP).
A "flapping" interface is one that frequently changes
from online to offline and offline to online.
At each level, a router serving a particular set of networks needs only to be advertised by a routing prefix of a given length,
greatly reducing the number of routes that need to be stored in memory.
MTU (Maximum Transmission Unit)
how large a packet can be sent without the need for fragmentation.
Unlike RIP, changes to the topology are transmitted as updates after startup,
so much less unnecessary traffic is generated and convergence performance is better.
Multipath
some algorithms provide for only a single path to the destination, whilst others support multiple paths. The advantage of the multipath approach is that the different paths can be used simultaneously for load balancing.
Routing Table
stores the routes selected by DUAL as the best ones (successors).
The Reliable Transport Protocol does not have any connection to s.
the Real-time Transport Protocol, which is used in VoIP communications.
Path length
the end-to-end cost of using a route (hop count). You might assign an arbitrary value to a particular path between router A and B and between B and C. The end-to-end path length is the sum of A-to_B and B-to-C.
Be efficient in selecting the best route
the metrics used must be representative of the network and meaningful. For example, if the fastest route is more important than the cheapest route, then ensure the algorithm used is optimized for the least delay.
It is possible for two routers in sequence to be configured with a default gateway of each other;
this is one way to connect three subnets together, though not the most efficient.
The ISPs subdivide their allocations into different sized blocks for different customer requirements, ranging from dynamically allocated addresses for home users,
through fixed single IP addresses for small businesses, to smaller and larger ranges for medium sized and large enterprises.
Some protocols use a hybrid of different methods
to perform path selection more efficiently.
Be efficient in routing packets
too much overhead and complexity in the selection and routing process will slow the packet throughput, leading to unacceptable delays.
Unlike RIP, EIGRP is a native IP protocol,.
which means that it is encapsulated directly in IP datagrams rather than using TCP or UDP (it is tagged with the protocol number "88" in the protocol field of the IP header)
EIGRP builds on the strengths of RIP
whilst providing for more efficient route selection, better administrative control, and better fault tolerance.
In IPv6, the host ID is always 64 bits and the network prefix is only given in slash notation. Prefixes are allocated in a hierarchical manner, with the smaller prefix blocks (which contain more networks) going to internet registries,
who then allocate mid-size blocks to ISPs, who then allocate the blocks with the largest network prefixes to end users. End users themselves may then be left with a block to use to subnet their network.
Price
you can assign a monetary cost to various links and the router using a cost metric will try to select the cheapest link available. Useful for organizations routing on a budget.
