Vous êtes sur la page 1sur 118

Computer Networks

Section1. Routing Algorithms and Network Layer Protocol


Presenter: Shu-Ping Lin

Outline
Routing Algorithms The Network Layer in The Internet

Outline
Routing Algorithms The Network Layer in The Internet

Store-and-Forward Packet Switching

Routing Algorithm
Routing algorithm
Datagram Virtual circuit

Differences between routing and forwarding Routing versus forwarding

Routing Algorithm (contd)


Requirement of routing algorithm
Correctness and Simplicity Robustness Stability Fairness Optimality

Datagram Routing

Virtual-Circuit Routing

Comparison

Routing Algorithm (contd)


Nonadaptive (static routing)
Do not base routing decisions on measurement of the current traffic and topology. The route used to get from node to node is computed in advance.

Adaptive (dynamic routing)


Change routing decisions to reflect changes in topology and traffic.

Optimality Principle
If router J is on the optimal path from router I to router K, the the optimal path from J to K also falls along the same route. Sink tree
The set of optimal routes from all sources to a given destination

The goal of all routing algorithms is to discover and use the sink trees for all routers.

Sink Tree

Shortest Path Routing


Link metric Dijkstra algorithm

Flooding
Every incoming packet is sent out on every outgoing link except the one it arrived on. Termination of flooding process
Hop counter Record of packet which has been flooded

Applications of flooding
Military application Comparison

Distance Vector Routing


Each router maintains a table giving the best known distance to each destination and which line to be used. Tables are updated by exchanging information with the neighbors. Items in the table
Preferred outgoing line Estimate of the distance to that distance

Distance Vector Routing (contd)

Distance Vector Routing (contd)


Converge to the correct answer, but do so slowly. It reacts rapidly to good news, but leisurely to bad news. The count-to-infinity problem

Count-to-Infinity Problem

Problem of DVR
Does not take line bandwidth into account. Take too long to converge.

Link State Routing


Learning about the neighbors
Sending a special HELLO packet on each point-to-point line

Measuring line cost to each of its neighbors


Round-trip time Traffic load Line bandwidth

Link State Routing (contd)


Building link state packet

Link State Routing (contd)


Distributing the link state packets
Flooding is used to distribute the link state packet. Each packet contains a sequence number that is incremented for each new packet sent.

Computing the new routes


Dijkstras algorithm can be run locally to construct the shortest path to all destinations.

Hierarchical Routing
As networks grow in size, the router routing tables grow proportionally. Each router has responsible for its region and knows nothing about the internal structure of other router.

Hierarchical Routing (contd)

Hierarchical Routing (contd)


Penalty of increased path length The optimal number of levels for an N router subnets is ln N.

Broadcasting Routing
The source simply send a distinct packet to each destination.
Waste bandwidth Have to know complete list of all destinations

Flooding
Generate too many packets Consume too much bandwidth

Broadcasting Routing (contd)


Multidestination routing
Each packet contains a list of destinations. The destination set is partitioned among the output lines.

Using spanning tree


Routers must know which of its lines belong to the spanning tree in advance. Copy an incoming packet onto all the spanning lines except the on it arrived on.

Broadcasting Routing (contd)


Reverse path forwarding
Router checks to see if the packet arrived on the line that is normally used for sending packets to the source of the broadcast.

Multicast Routing
Each router computes a spanning tree covering all other routers. When a process sends a multicast packet to a group
Examining the spanning tree of this group Removing all lines that do not lead to hosts that are members of group

Multicast Routing (contd)

Multicast Routing (contd)


Pruning spanning tree
Reverse path forwarding Router with no hosts interested in a particular group sends a PRUNE message. The subnet is recursively pruned.

Multicast Routing (contd)


Disadvantage of algorithm, suppose that network has n groups, each with an average of m members.
For each group, m pruned spanning trees must be stored Total of mn trees

Core-based tree
A host wanting to multicast sends packets to the core, which does the multicast along the spanning tree.

Routing for Mobile Hosts

Routing for Mobile Hosts (contd)


Mobile host
Migratory hosts Roaming hosts

All hosts are assumed to have a permanent home location that never changes. Hosts also have a permanent home address that can be used to determine their home location.

Routing for Mobile Hosts (contd)


Each area has a home agent which keeps track of hosts whose home is in the area, but who are currently visiting another area. Each area has foreign agent which keeps track of all mobile hosts visiting this area.

Routing for Mobile Hosts (contd)


Registration procedure
Foreign agent searching Registration Foreign agent contacts the mobile hosts home agent. Verification of mobile hosts home agent Acknowledgement from the mobile hosts home agent

Routing for Mobile Hosts (contd)


Tunneling Encapsulate the original packet in the payload field of an outer packet and sends the latter to the foreign agent.

Routing for Mobile Hosts (contd)

Routing in Ad Hoc Networks


AODV (Ad hoc On-demand Distance Vector)
Distant relative of the Bellman-Ford distance vector algorithm Considering the limited bandwidth and low battery life On-demand algorithm

Routing in Ad Hoc Networks (contd)


AODV algorithm maintains a table at each node, keyed by destination and which neighbor to send packets in order to reach destination. Route request packet

Routing in Ad Hoc Networks (contd)


When a route request packet arrives at a node
The (Source address, request ID) pair is looked up in a local history table too see if this request has already been seen and processed. Receiver looks up the destination in its route table. If receiver does not know a fresh route to the destination, it increments the Hop count field and rebroadcast the REQUEST packet.

Routing in Ad Hoc Networks (contd)

Routing in Ad Hoc Networks (contd)


Destination constructs a ROUTE REPLY packet which follows the reverse path to source. Each intermediate node enters this packet into local routing table when
No route to destination is known. Sequence number for destination in REPLY is greater than the value in the routing table. Sequence number is equal but the new route is shorter.

Routing in Ad Hoc Networks (contd)


Problem of mobility Route maintenance
Periodically, each node broadcasts a Hello message. If no response is returned, router prune this link.

Active neighbor

Routing in Ad Hoc Networks (contd)

Routing in Ad Hoc Networks (contd)


Critical difference between AODV and DVR
Nodes do not send out periodic broadcasts containing their entire routing table.

Node Lookup in Peer-to-Peer Networks


One of p2p routing algorithmChord Each user node has an IP address that can be hashed to an m-bit number call node identifier. Successor (k) is the node identifier of the first actual node following k around the circle clockwise.

Node Lookup in Peer-to-Peer Networks (contd)


Finger table
Start field = k +2i (modulo 2m) If key falls between k and successort (k), then the node holding information about key is successor (k). Otherwise, the entry whose start field is the closest predecessor of key is tried.

Node Lookup in Peer-to-Peer Networks (contd)

Node Lookup in Peer-to-Peer Networks (contd)


Node r joining
New node r asks successor (r) for its predecessor. Insert r in between successor and predecessor. Successor should hand over those keys in the range predecessor (r)-r, which now belong to r.

Every node runs a background process that periodically recomputes each finger by calling successor.

Outline
Routing Algorithm The Network Layer in The Internet

The Network Layer in The Internet


Principles for network design
Make sure it works. Keep it simple. Make clear choices. Exploit modularity. Expect heterogeneity. Avoid static options and parameters. Look for a good design; it need not be perfect.

The Network Layer in The Internet (contd)


Think about scalability. Consider performance and cost.

Network layer provides best-efforts way to transport datagrams from source to destination.

The IP Protocol
The IPv4 header

IP Address
IP address formats

IP Address (contd)
Special IP address

Subnet
Split a network into several parts for internal use.

Subnet (contd)
Some bits are taken away from the host number to create a subnet number. For example, a university can use a 6-bit subnet number and a 10-bit host number, allowing fro up to 64 subnets. Outside the network, the subnet is not visible, so allocating a new subnet does not require contacting ICANN.

Subnet (contd)
Subnet mask indicates the split between network + subnet number and host.

Network Address Translation


Shortage of IP address Within the company, every computer gets a unique IP address, which is used for routing intramural traffic. An address translation takes place when a packet exits the company.

Network Address Translation (contd)

Network Address Translation (contd)


Use TCP port to distinguish the traffic. TCP source port field is replaced by an index into the NAT boxs translation table. Each entry contains original IP address and original source port.

Network Address Translation (contd)


Violation of the architectural model of IP
Every IP address uniquely identifies a machine.

NAT changes the Internet from a connectionless network to a kind of connection-oriented network. Violation of protocol layering
layer k may not make any assumptions about what layer k+1 has put into the payload field.

Internet Control Protocols


Internet Control Message Protocol (ICMP)

Address Resolution Protocol


How do IP address get mapped onto data link layer addresses, such as Ethernet? Configuration file ARP
Broadcast a packet onto the Ethernet asking: Who owns IP address xxx.xxx.xxx.xxx. Correspondent will reply with its Ethernet address

Address Resolution Protocol (contd)

Address Resolution Protocol (contd)


Make ARP work more efficiently
Local cache ARP request with requestors Ethernet address

Transmit data to remote network.

RARP, BOOTP, and DHCP


When a computer is booted how does it get the IP address. RARP
Broadcast a packet to ask RARP server. Packet cannot be forwarded by routers, so RARP server is needed on each network.

RARP, BOOTP, and DHCP (contd)


BOOTP
Use UDP message, which are forwarded over routers. Require manual configuration of tables mapping IP address to Ethernet address.

DHCP
Both manual IP address assignment and automatic assignment

RARP, BOOTP, and DHCP (contd)


DHCP

OSPFThe Interior Gateway Routing Protocol


Routing algorithm within as autonomous system (AS) is called an interior gateway protocol. The original Internet interior gateway protocol was distance vector protocol (RIP). Link state protocol replaced RIP in 1979. OSPF began work in 1988.

OSPFThe Interior Gateway Routing Protocol (contd)


Internet is made up of a large number of autonomous systems. Each AS is operated by a different organization and can use its own routing algorithm. Every AS has a backbone area. All areas are connected to the backbone and can communicate each other via backbone.

OSPFThe Interior Gateway Routing Protocol (contd)

OSPFThe Interior Gateway Routing Protocol (contd)


Using flooding, each router informs all the other router in its area of its neighbors and costs. This information allows each router to construct the graph for its area and compute the shortest path. Backbone routers accept information from the area border routers in order to compute the best route from each backbone router to every other router.

BGPThe Exterior Gateway Routing Protocol


All interior gateway protocol has to do is to move packets as efficiently as possible without worrying about politics. Exterior gateway protocol routers have to worry about politics.
No transit traffic through certain ASes. Traffic starting or ending at IBM should not transit Microsoft.

Policies are typically manually configured into each BGP router and are not part of protocol.

BGPThe Exterior Gateway Routing Protocol (contd)


Stub networks
Has only one connection to BGP graph and cannot be used for transit traffic.

Multiconnected networks
Could be used for transit traffic, except that they refuse

Transit networks
Willing to handle third-party packet

BGPThe Exterior Gateway Routing Protocol (contd)


Border Gateway Protocol (BGP)
Fundamentally a distance vector protocol, but each BGP router keeps track of the path used.

Internet Multicasting
IP uses class D address to support multicast. Two kinds of group address
Permanent address Temporary address

Internet Multicasting (contd)


224.0.0.1 All systems on a LAN 224.0.0.2 All routers on a LAN 224.0.0.5 All OSPF routers on a LAN 224.0.0.6 All designated OSPF routers on a LAN Temporary groups must be created before they can be used.

Internet Multicasting (contd)


Multicasting routing is done using spanning tree. Each multicast router exchanges information with its neighbors, using a modified distance vector protocol.

Mobile IP
Increment of portable computers. Major goals
Mobile host must be able to use its home IP address anywhere. Changes to software and router are not permitted. No overhead should be incurred when a mobile host is at home.

Mobile IP (contd)
See section 5.2.9 (Routing for Mobile Hosts) When foreign shows up at a foreign site, it contacts the foreign host there and registers. The foreign host contacts the users home agent and gives it a care-of-address.

Mobile IP (contd)
Each foreign agent periodically broadcast its address and type of service, which called advertisement. Registration for impolite mobile hosts that leave without saying goodbye.

IPv6
While NAT may buy a few more years time, IP in its current form (IPv4) is numbered. IETF issued a call for proposal and discussion in RFC 1550. IPv6 is not compatible with IPv4, but it is compatible with other auxiliary Internet protocol.

IPv6 (contd)
Main features
Longer addresses Simplification of the header Better support for options Security Quality of service

IPv6 (contd)
The IPv6 fixed header

IPv6 (contd)
New notation of 16-byte address
8000:0000:0000:0000:0123:4567:89AB:CDEF 8000::123:4567:89AB:CDEF

IPv4 can be written as ::192.168.20.46 Vanishment of IPv4 headers


IHL Protocol Checksum

IPv6 (contd)
Extension header is encoded as (Type, Length, Value) tuple.
Type is a 1-byte field telling which option this is. Length is a 1-byte field telling how long the value is. Value is any information required.

IPv6 (contd)
IPv6 extension header

IPv6 (contd)
Controversies
Hop limit field length Maximum packet size Checksum Mobile issue

Section2. The Internet Transport Protocols: TCP

Introduction to TCP
Function of providing reliable end-to-end byte stream over an unreliable internetwork. Lack of IP
No guarantee that packets will be delivered properly. Packets may arrive in the wrong order.

TCP Service Model


TCP service is obtained by both sender and receiver creating sockets. Each socket has IP address and a 16-bit number called port. Port number below 1024 are called wellknown ports and reserved for standard service.

TCP Service Model (contd)


Well-Known Ports

TCP Service Model (contd)


Characteristics of TCP connections
Full duplex Point-to-point Byte stream, not a message stream

TCP Service Model (contd)


A key feature of TCP is that every packet on a TCP connection has its own 32-bit sequence number. TCP segment consisting of fixed 20-byte header is used for exchanging data between sender and receiver.

TCP Service Model (contd)


Segment size issue
Fit in the 65515-byte IP paylo.ad Cant exceed maximum transfer unit (MTU). MTU is generally 1500 bytes.

The basic protocol used by TCP is the sliding window protocol.


Timer Acknowledgement number

TCP Segment Header

TCP Connection Establishment


Three-way handshake.

TCP Connection Release


To release a connection , either party can send a TCP segment with the FIN bit set. When the FIN is acknowledged, that direction is shut down for new data. Normally, four TCP segments are required to release a connection.

TCP Connection Management Modeling

TCP Transmission Policy

TCP Transmission Policy (contd)


When the window size is 0, the sender stop sending data to receiver except
Urgent data may be sent. Send a 1-byte segment to make the receiver reannounce the next byte expected and window size.

TCP Transmission Policy (contd)


Consider the worst case in performance issue. For receivers
Delay acknowledgements and window updates for 500 msec in the hope of acquiring some data.

For senders
Nagles algorithm

TCP Transmission Policy (contd)


Silly window syndrome
Data are passed to the sending TCP entity in large blocks, but an interactive application on the receiving side reads data 1 byte at a time.

TCP Transmission Policy (contd)


Clarks algorithm
It forces window update to wait until it has a decent amount of space available.

Combination of Nagles and Clarks algorithm.

TCP Congestion Control


When the load offered to any network is more than it can handle, congestion builds up. TCP achieves congestion control by dynamically manipulating the window size. First step of congestion control: detection. All the Internet TCP algorithms assume that timeouts are caused by congestion.

TCP Congestion Control (contd)

TCP Congestion Control (contd)


Two potential problemsnetwork capacity and receiver capacity. Each sender maintains two windows receiver window and congestion window. The number of bytes that may be sent is the minimum of the two windows.

TCP Congestion Control (contd)


Slow start

TCP Timer Management


Retransmission timer Difficulties of setting retransmission timer

TCP Timer Management (contd)


Estimate RTT
RTT TRR (1 )M

Where is a smothing factor that determines how much weight is given to the old value. TCP uses RTT to estimate retransmission timer.

TCP Timer Management (contd)


Jacobson proposed proposed to use mean deviation as a cheap estimator of the standard deviation. D D (1 ) | RTT M |
Timeout RTT 4 D

TCP Timer Management (contd)


Potential problem
When the ack of retransmission packet comes in, it is unclear whether the ack refers to the first transmission or a later one. Karns algorithm

Persistence timer is design to prevent the deadlock. Keepalive timer.

Wireless TCP and UDP


Problem due to congestion control. TCP assumes that timeouts are caused by congestion, not by lost packets. A packet is lost on a wired network, the sender should slow down. When one is lost on a wireless network, the sender should try harder.

Wireless TCP and UDP (contd)


Bakne and Badrinath propose indirect TCP which split the TCP connection into two separate connections.

Wireless TCP and UDP (contd)


The advantage of this scheme is that bothconnections are now homogeneous. Timeouts on the first connection can slow the sender down, whereas timeouts on the second one can speed it up.z The disadvantage of the scheme is that it violates the semantics of TCP.

Transactional TCP
Efficiency problem of TCP