CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

24
CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing
  • date post

    19-Dec-2015
  • Category

    Documents

  • view

    223
  • download

    2

Transcript of CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Page 1: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

CSCI 4550/8556Computer Networks

Comer, Chapter 25:Internet Routing

Page 2: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Introduction

While the transport protocols form the heart of the protocols used by applications, IP datagrams and their routing form the heart of the delivery system.

Routing tables are used by routers to make decisions on how to forward or deliver datagrams.

Routing tablesmust be built initially when a router is “powered up,” and

must be updated as the network changes.

Page 3: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Static Vs. Dynamic Routing

Static routes do not change.Dynamic routes

begin with an initial set of routesare updated with information from other computers to learn about optimal routes

Route propagation software, or just “routing software,” is started when a router begins execution.

Page 4: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Static Routing

Static routing isinexpensive and easy to use (no updates sent on the network)inflexible (no updates sent on the network!)used by many hosts with one network interface and one router connecting to the Internet.

The routing table is usually small, containing only two entries:

specific route to hosts on the network to which the host attachesdefault route to the single router

Page 5: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Static Routing Example

Page 6: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Dynamic Routing and RoutersStatic routing can be used in routers that deal only with a limited set of routes to a set of networks that never change.But even small network topologies need dynamic routing if networks administered by two or more organizations are connected.

manual routing table updates is slow and doesn’t scale.static routing can’t accommodate exceptional conditions (network interface or router failures).

To handle routing changes, each router runs software that

learns about destinations other routers can reach (using incoming information), andinforms other routers about destinations it can reach.

Page 7: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Routing in the Global InternetThe global internet is too large for each router to maintain complete information:

There are too many individual networks involved.The routing information exchange would overwhelm the networks.

Instead, a two-level routing hierarchy is used.Routers and networks are divided into autonomous groups.

Routers within a group exchange information.At least one router in each group exchanges summary information with other groups.

The size of the groups, and the protocols routers use to exchange information, may vary between organizations.

Page 8: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Autonomous System ConceptAutonomous System: contiguous set of networks and routers under control of one administrative authority. [For example, UNOmaha]Size and administrative authority definitions are flexible enough to accommodate disparate groups.If an organization grows, it may be convenient to divide its networks into two autonomous systems, each with its own internal routers, and each connected to an ISP, rather than using much larger routers and more network traffic related to routing.Routing traffic grows O (n 2), where n is the number of routers.

Page 9: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Interior and Exterior ProtocolsRouters within an autonomous system use an Interior Gateway Protocol (IGP) to exchange information.

There are several different IGPs available.Usually easy to install and operate.Each has its own limitations.

An Exterior Gateway Protocol (EGP) is used to exchange routing information with a router in another autonomous system.

More complex, but more flexible than IGPs.Summarizes routing information before passing it on.Can use policy constraints to restrict information release to the outside.

Page 10: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

IGP/ECP Illustration

IGP1 and IGP2 (different IGPs) used in the two autonomous systems for communication between routers.EGP used for exchange of routing information between the systems.

Page 11: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Optimal Routes, Routing Metrics, and IGPsOptimal routes are selected using routing metrics comprised of two components:

administrative cost – assigned manually to reflect organizational routing policieshop count – reflects actual number of routers/networks in a path

IGPs use routing metrics to find optimal routes within an autonomous system, since the routing metrics are (hopefully) consistent within the organization.EGPs do not use routing metrics, since they have no basis for comparing the metrics used by different autonomous systems.

Page 12: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Routes and Data TrafficData traffic for a given destination flows in exactly the opposite direction of routing traffic.This is because only as a result of a router advertising a route can data arrive using that route.

Page 13: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

The Border Gateway Protocol (BGP-4)

Border Gateway Protocol version 4 (BGP-4) is the most popular EGP in the Internet.

Provides routes only among autonomous systems, not individual routersHas provision for policies that can restrict route disclosureClassifies systems as transit systems (agree to pass traffic) or stub systems (won’t pass traffic)Uses TCP for reliable communicationUsed by all major ISPsUsed by the routing arbiter system to maintain a distributed database of all possible destination in the Internet, with copies running on separate route servers

Page 14: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

The Routing Information Protocol (RIP)RIP, one of the first IGP protocols used with IP, is implemented with the UNIX routed program.

Is an Interior Gateway ProtocolMeasures distance in hops (1 hop to an adjacent network)Uses UDP for all routing messagesIntended for use on LANs; version 1 uses broadcast, version 2 allows multicastAllows routers to advertise a default routeUses the distance-vector routing algorithm, with route advertisements giving (destination, distance) pairs.Allows hosts to retrieve routing information

RIP’s chief advantage is simplicity.

Page 15: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

RIP Packet FormatRIP packets contain

An address and subnet mask for each networkNext hop address and distance for each netAddress family (2 for IP)A command (1=request, 2=response, others obsolete)Route tags to allow entries to be grouped

Page 16: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Open Shortest Path First Protocol (OSPF)RIP messages contain complete lists of destinations and distances, making them large. It works well with a small number of routers but doesn’t scale well.A newer protocol, OSPF, that can scale to large organizations was devised by the IETF.

OSPF is an Interior Gateway Protocol.Subnet masks are included with network addresses.Message exchanges can be authenticated.Routes can be imported from other sources (e.g. BGP).Uses the link-state routing (SPF) algorithm (based on Dijkstra’s shortest path algorithm)Designates a single router on broadcast networks to handle all broadcasts (rather than having each router broadcast)

Page 17: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

OSPF Example

(a) An internet consisting of seven networks interconnected by routers(b) a corresponding OSPF graph. In the simplest case, each router corresponds to a node in the graph.

Page 18: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

OSPF Areas

OSPF includes the ability to perform hierarchical routing.

Instead of dealing with just a single autonomous network, OSPF can deal with subsets called areas.Routers (using OSPF) within an area exchange link status messages periodically.One router in each area also communicates summary information with one or more routers in other areas.Each broadcast is limited to a specific area.

Because of hierarchical routing abilities, OSPF can scale to handle a larger number of routers than other IGPs.

Page 19: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Multicast RoutingA multicast allows a single address to be used for a dynamically changing group of recipients not necessarily on the same network.Unicast routing design goals include stability: routes don’t change frequently.Multicast routing differs significantly:

Hosts can choose to join a multicast group at any time and begin receiving a copy of all packets sent to the group.Hosts can choose to leave a multicast group at any time.Multicast group membership is private; no receiver or sender knows the identity or number of group members.Any application can send a message to any multicast group at any time.

Page 20: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

The IGMP Protocol

The Internet Group Multicast Protocol (IGMP) is used by a host to inform a nearby router that it wishes to change membership status in a multicast group.

Only hosts are members of groups, not individual applications; multiple applications on the same host that wish to receive multicast messages will receive copies from the host network software.IGMP is used only on the network between the host and the router.

Page 21: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Forwarding and Discovery TechniquesWhen a host joins a multicast group, the router it informs must

establish a path to the groupforward datagrams for the group to the host

Dynamic groups and anonymous senders make general-purpose multicasting extremely difficult.The size of groups can vary significantly, from small (3-5 hosts) to worldwide groups.Multicast routing protocols must be able to change routing quickly and continually.

Page 22: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Multicast Data Forwarding TechniquesFlood-and-prune – good for small groups with hosts connected to contiguous LANs.

Multicast messages sent to all networks, where routers use hardware broadcast on LANs.When a router reports back that it has no group members, flooding to that network stops (it’s pruned).

Configuration-and-tunneling – good for geographically dispersed groups.

One router at each site knows about other sites.Messages broadcast on directly attached LANs.Other sites receive copies using IP-in-IP tunneling.

Core-based discovery – good for graceful scaling of group sizes

A core unicast address is associated with the group.The routers participating in the group form a tree rooted at the core address, and communicate using IP-in-IP tunneling.

Page 23: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

Multicast ProtocolsDistance Vector Multicast Routing Protocol – performs local multicast and uses IP-in-IP to send to other sites.Core Based Trees – builds a tree from a central point, with membership messages sent toward the core.Protocol Independent Multicast – Sparse Mode – like CBT, but doesn’t depend on any particular unicast routing protocol.Protocol Independent Multicast – Dense Mode – uses flood and prune; good for use within an organization.Multicast extensions to OSPF – for use within an organization; uses many of the OSPF concepts and facilities.

Page 24: CSCI 4550/8556 Computer Networks Comer, Chapter 25: Internet Routing.

SummaryHosts and routers use IP routing tables, mostly static in hosts and dynamic in routers.Internet is divided into autonomous systems.

EGPs are used to pass routing information between systems; BGP is the primary EGP in the Internet.IGPs are used to exchange routing information within a system; RIP and OSPF are popular IGPs.

Multicast route propagation is difficult because membership is dynamic and arbitrary hosts can send messages to the group. No Internet-wide multicast technology exists.