Internet Routing - CS

smashlizardsNetworking and Communications

Oct 29, 2013 (3 years and 9 months ago)

91 views

Internet Routing

Chapter 27

Static Versus Dynamic Routing


Static


Does not change (except in the event of an error)


Dynamic


Assumes a basic static routing table


Route Propagation Software


Goal: optimal routes

Static Routing in Hosts and a Default
Route


Typical Situation


A host with one connection to a network with one
router


Use static routing! Do you need to know how to get to
Australia?


One Host, One Router

Dynamic Routing and Routers


Usually, a router should
not use static routing


In this figure the
administrator of R1 does
not want to have to
manually change routes
on Net 3



Routing in the Global Internet


A two
-
level routing hierarchy


Routers and networks are broken up into groups


One or more routers in each group summarize
information and pass it on to other groups


A group is called an Autonomous System


Each such system chooses its own routing protocol


If Calvin adds a subnet for CS, we do NOT tell Hope



The Two Types of Routing Protocols


Interior Gateway Protocols


IGPs


Several are available


Each autonomous system makes its own choice


Exterior Gateway Protocols


EGPs


Summarize the data


Implement policy constraints

When EGPs and IGPs Are Used

Optimal Routes, Routing Metrics, and
IGPs


Finding one optimal path is not realistic


What does “optimal” mean?


Interactive login


Least delay


Downloading a larger graphics file


Maximum throughput


Audio webcast


Least jitter


Typical internet routing combines hop count and
administrative cost


Example


An administrative fence


Two routes from A to B


The administration does not want internal traffic
to go over the first route


Add administrative cost


Since each autonomous system chooses its own
metrics


EGPs cannot compare apples and oranges so they
only care about the existence of a path

Routes and Data Traffic

The Border Gateway Protocol


BGP


Version 4


BGP
-
4


Characteristics


Routing among autonomous systems


no use of routing metrics


Provision for Policies


Classifies each AS as transit or stub


Reliable transport using TCP


The Routing Information Protocol


RIP


Routing within an AS


Hop Count Metric


Unreliable transport using UDP


Broadcast (v1) or multicast (v2)


Support fro default route propagation


Distance
-
vector algorithm


Passive version for hosts

RIP Packet Format

The Open Shortest Path First Protocol


OSPF


Much more scalable than RIP


Routes within an AS


Full CIDR and subnet support


Authenticated message exchange


Authentication refers to a guarantee of the source of a
message


Imported routes


From another protocol, say BGP


Link
-
State algorithm


Support for metrics


Designates a single router to broadcast on the
network


An Example OSPF Graph

OSPF Areas


A manager divides an AS into areas


Routers only communicate within their own
area


Except for designated routers which exchange
information with other areas


This allows OSPF to scale well

Multicast Routing


Up to now we have only considered


unicast


broadcast


Multicast is a major complication


Multicast uses designated IP addresses


Characterized by


dynamic membership


anonymous senders

Internet Group Membership Protocol


IGMP


Only used between a host and a router


Group members are hosts, not applications


Multiple copies are made for each application


Last one to leave turns out the lights, ie, host
computer informs the router that it is no longer a
member of the group

Forwarding and Discovery Techniques


Routers, not hosts, have responsibility for the
propagation of multicast routing information


A large group may span the globe


Routing protocols must adapt quickly to
changes in membership


Three different approaches have emerged

Flood
-
And
-
Prune


Ideal for small groups in which members are attached to
contiguous LANs


In the flooding stage routers forward each datagram to all
networks to which they are attached


Reverse Path Broadcasting (RPB) is used to avoid loops


A router receives a datagram from A on interface 3, say


It checks its routing tables to see if it would send to A on
interface 3


If not, it drops the datagram


Eventually, a router learns that no hosts on a given
network are members of the group and prunes that
network

Configuration
-
And
-
Tunneling


Ideal for groups spread out over long distances


A router at each site is configured to learn about
other sites


When a multicast datagram arrives, the router at
a site transmits on all attached LANs and
consults its configuration table to see which
other sites should receive a copy


Uses IP
-
in
-
IP tunneling (why?) to send a copy
to other sites

Core
-
Based Discovery


Allows multicast to scale gracefully from a
small group in one area to a large group with
members at arbitrary locations


Each multicast group has a core unicast address


When a router needs to reach a group, it sends a
datagram to the core address


A router that participates in the group


forwards a multicast datagram to the group


responds to a request to join the group by adding the
information to its routes and forwards a copy of each
subsequent multicast datagram to the new member

Multicast Protocols (Proposed)


Distance Vector Multicast Routing Protocol


DVMRP


local multicast


IP
-
in
-
IP to other sites


Core Based Trees


CBT


protocol software builds a delivery tree from a central point


Protocol Independent Multicast


Sparse Mode (PIM
-
SM)


tree approach


no particular unicast protocol


Dense Mode (PIM
-
DM)


within an organization


prunes


Multicast extensions (MOSPF)