Lectures on 4.10.2000
1st lecture 12:15-13:00
• The Internet Protocol (IPv4)
• IP Addresses:
• Structure of
• Problems with
• Registration of
• Address Resolution Protocol (ARP) 2nd lecture 13:15-14:00
• Routing IP Datagrams
• Direct and Indirect Delivery
• The Need for Routing Protocols
• Dividing the Internet into Autonomous Systems
• Exterior Routing (EGP, BGP)
• Interior Routing (RIP2, EIGRP, OSPF)
• On Choosing Routing Protocols
The Internet Protocol (IP)
• The Internet Protocol was designed to provide a connectionless datagram service in internetworks
• From the user’s point of view, internet is a virtual network interconnecting all the hosts connected to it - the internal structure of the internet is unessential
• Conceptually, a TCP/IP-based internet provides three levels of services (the protocols named in parentheses implement
these services):
• Application level services (e.g. SMTP, FTP, HTTP)
• Reliable transport service (TCP)
• Connectionless packet service (IP)
• The basic architecture of the Internet was defined in 1974 by Vinton Cerf and Robert Kahn and has remained unchanged
• The success of TCP/IP is largely based on the good func- tioning and adaptability of the architecture (layer structure)
The Internet Protocol (IP)
• The IP-based packet-forwarding service provided by an internet can be characterized as follows:
• The service is unreliable, because IP does not use
acknowledgements and the delivery of a datagram is not guaranteed
• The service is connectionless, because every packet is routed separately and independently of other packets
• The service is based on best-effort delivery, because the internet software is doing its best to deliver every packet
• IP defines three significant things:
• The structure of the Internet datagram
• The routing of datagrams
• A number of rules implementing the principles of connectionless packet service
• IP is the key technology on which the Internet is based
IPv4 Datagram Header
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|Version| IHL |Type of Service| Total Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Identification |Flags| Fragment Offset | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Time to Live | Protocol | Header Checksum | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Source Address | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Destination Address | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Options | Padding | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
• The figure depicting the header of an IPv4 datagram is taken directly from RFC-791
Type of Service Field (RFC791)
Bits 0-2: Precedence.
Bit 3: 0 = Normal Delay, 1 = Low Delay.
Bits 4: 0 = Normal Throughput, 1 = High Throughput.
Bits 5: 0 = Normal Reliability, 1 = High Reliability.
Bit 6-7: Reserved for Future Use.
0 1 2 3 4 5 6 7 +---+---+---+---+---+---+---+---+
| | | | | | | | PRECEDENCE | D | T | R | 0 | 0 | | | | | | | | +---+---+---+---+---+---+---+---+
Precedence
111 - Network Control
110 - Internetwork Control 101 - CRITIC/ECP
100 - Flash Override 011 - Flash
010 - Immediate 001 - Priority 000 - Routine
IP Addresses
• Internet is a logical network consisting of a number of physical networks interconnected via routers
• Universal communication service = any-to-any communication
• Internet is based on compact, standardized, binary addresses - namely IP addresses
• Every network interface connected to the Internet has a
unique 32-bit IP address that is used in communications with this interface
• DNS maps the domain names onto IP addresses
• Routers route a packet based on its destination IP address
• Routing protocols help routers find a path to the destination
• The Address Resolution Protocol (ARP) maps IP addresses onto physical (MAC-layer) addresses
IP Addresses
• IP address is written as four decimal numbers, each between 0 and 255, separated with points
• For example the binary address “11000001 11010010
00001001 00111111” is written “193.210.9.63” and it stands for node “63” of the class C network “193.210.9”
• “0” as the number of a node or a network stands for “this”
and ”-1” (number with all bits 1) stands for ”all” - for this reason, they are not used as node or network numbers
• There can be at most 126 class A networks and each of them can have at most 16,777,214 nodes
• There can be at most 16,382 class B networks and each of them can have at most 65,534 nodes
• There can be at most 2,097,150 class C networks and each of them can have at most 254 nodes
Structure of IP Addresses
• IP addresses are 32-bit integers and they are divided into three main classes: A, B and C
• Class D was added for multicasting and class E reserved for future use
1 Network # (14 bit) Node # (16 bit)
Class B 0
Class C 1 1 0 Network # (21 bit) Node # (8 bit)
Class D 1 1 1 0 Network # (28 bit)
Class E 1 1 1 1 Reserved for future use
0 Network # (7 b) Node # (24 bit)
Class A
0 1 2 3 4 5 6 7 8 16 24 31
IP Addresses
• An IP address conceptually consists of a pair: (net-id, host-id)
• It is easy to separate the net and host parts from an address
• An IP address does not determine a host but a network interface of a host
• For example a router has an IP address for each of the networks it is attached to
• An IP address can also indicate an entire network
• An address with all host-id bits "0" indicates a network
• An address with all host-id bits "1" is a broadcast address
• Directed broadcast = all hosts in a given network
• Limited broadcast (local network broadcast) = all hosts in this network
• We normally try to limit broadcasts to as small a scope as possible
• Multicasting is a non-trivial function to implement
Some IP Addresses
• Special addresses
• Limited broadcast: "255.255.255.255"
• Directed broadcast:
"A.255.255.255", "B.B.255.255", "C.C.C.255"
• Loopback: "127.X.Y.Z"
• All systems on this subnet: "224.0.0.1"
• All routers on this subnet: "224.0.0.2"
• Some examples of IP addresses:
• "0.0.0.0" = some host (source only)
• "255.255.255.255" = any host (destination only)
• "129.34.0.3" = host #3 in n/wk 129.34
• "129.34.0.0" = some host in n/wk 129.34 (source only)
• "129.34.255.255" = any host in n/wk 129.34 (destination)
• "255.255.0.3" = host #3 in this n/wk
• "127.0.0.1" = this host (local loopback)
Problems with IP Addresses
• IP address maps into a network interface => when a host moves into another network its IP address changes
(this is a problem e.g. in portable PCs)
• When a network grows, changing the network number and IP addresses becomes very difficult
• If a host has several IP addresses, the path to and reachability of the host depend on the address used
• The biggest single problem with IP addresses is its inadequate 32-bit address space
(the ”ROADS” problem, Running Out of Address Space)
• Especially the number of class B networks is too small
• Classless Inter-Domain Routing (CIDR) and Network Address Translation (NAT) provide an interim solution to the problem
• IPv6 (IP Next Generation, IPng) solves the problem by using 128-bit addresses but moving into IPv6 is very difficult
Registration of IP Addresses
• In an intranet it is (at least in principle) possible to use unregistered IP addresses
• Network connected into the Internet need to be registered
• Network numbers are governed by the Internet Assigned Number Authority (IANA)
• In practice numbers are assigned by the Internet Network Information Center (INTERNIC)
• Today most companies and private people get their IP addresses from an Internet Service Provider (ISP)
• Who ”owns” the addresses?
- normally the ISP, which poses limitations to competition
• Private Internet Address Space (RFC-1597, including the old ARPANET class A network 10), some registered addresses, and NAT provide a working solution for most companies
Mapping IP Addresses to Physical Addr.
• Two hosts connected to the same physical network can
communicate if they know each other’s physical addresses
• In IP terminology " physical address” usually means, depending on the network, either the network address (e.g. in X.25) or MAC address (e.g. in LANs)
• In the physical layer of OSI no addresses are used
• Assume that A and B are connected to the same physical network and they have IP addresses IA and IB and physical addresses PA and PB, respectively
• A mapping: IA => PA and IB => PB is needed - the Address Resolution Problem
• If the physical addresses can be chosen (e.g. in proNET), a direct mapping is possible: PA = f(IA)
• In the general case, a direct mapping is not possible
Address Resolution Protocol (ARP)
• In LANs fixed 48-bit MAC-level addresses are used
• A 32-bit IP address cannot be directly mapped into a 48-bit MAC address
• The Address Resolution Protocol (ARP) is a mechanism,
through which a host can find the physical address of another host residing in the same network based on its IP address
• The source host sends the inquiry to the local network in a broadcast message, which the target host answers to
• The source host caches the physical addresses into a lookup table in order to avoid unnecessary use of the ARP
• At the same time the target host can cache the physical address of the source host
• ARP is a low-level protocol that helps hide the physical network addresses
Routing IP Datagrams
• We have already gone through the datagram service of internet, IP addresses and the Internet Protocol (IP)
• We shall now see, how routers route IP datagrams and deliver them to the right recipient
• Routing means the process of choosing paths over which to send packets
• Router is a computer that performs routing
• In wide area networks (WANs, such as X.25) there happens routing within the network that is invisible to the outside
• The purpose of IP is to interconnect various networks into an internet providing a uniform datagram service
IP Routing
• A router is connected to at least two networks and it routes between them
• In practice, a router may also have just one physical network interface (e.g. Frame Relay or ATM), in which case routing takes place between virtual circuits or logical networks
• Both the sending host and the routers between the sender and receiver participate in routing
• Any computer connected to more than one network may act as a router
• Internet standards define clear and separate roles for a
computer and a router - a routing computer needs to keep these roles separate
Direct and Indirect Delivery
Routing can be divided into two forms:
• Direct delivery
• An IP datagram can be sent directly (without involving routers) to another host on the same physical network
• Because all IP addresses of the same network have a
common prefix (net-id), it is easy to determine, what hosts reside on the same network
• Indirect delivery
• When the recipient is on another network, the IP datagram is delivered indirectly through router(s)
• In indirect delivery, the sender has to determine, what router the datagram shall be sent to
• The routers of an internet cooperate to form a routing infra- structure that passes datagrams from router to router until they reach a router than can deliver them directly
IP Routing (continued)
• Normally IP routing is based on a routing table, on each
machine, that stores information about possible destinations and how to reach them
• Because both computers and routers participate in routing, they both have routing tables
• According to the Information hiding principle, we prefer to store information about routes to networks rather than
individual hosts
• Typically a routing table entry is of the form: (N, R), where N is the IP address of the destination network and R is the IP address of the ”next” router
• The router R is called the ”next hop” and the type of routing described above ”next-hop routing”
• As implied by its name, the next-hop router can be reached directly (with one hop) and is connected to the same network
The Need for Routing Protocols
• We have already gone through the functioning of the Internet, which is based on the IP protocol and routers
• If the sender and receiver are not on the same network, the IP packet gets sent to a router, that sends it on to the right direction based on the contents of its routing tables
• Questions:
• What values should the routing tables contain?
• How are these values born?
• Routing tables shall have initial values which may be empty
• Routing tables need to be updated, either manually or automatically
• In practice the routing tables of a large network can only be efficiently updated automatically
• For this purpose, routers exchange routing information by using a routing protocol
Growth of the Internet and its Consequences
• In the early 1980’s, the whole Internet was a uniform network, whose routers (gateways) exchanged routing information
using the Gateway-to-Gateway Protocol (GGP)
• The routing tables of all the routers contained the routing information of all network numbers of the Internet
• As the Internet kept growing, the following problems appeared:
• The overhead caused by routing increased rapidly
• The interoperation of different routers caused difficulties
• When the version of the routing protocol was changes, all the routers had to be updated at the same time
• A solution was needed, where not all routers exchange routing information
Dividing the Internet into Autonomous Systems
• It was decided, that the Internet should be divided into Autonomous Systems (AS)
• One AS if formed by the routers and networks under the same administration
• ARPANET and SATNET together formed an AS called by the name ”core” and forming the backbone of the Internet
• Other AS’s where called by the name ”stub AS”
• Every stub needs to have at least one router connected to at least one core router in order to achieve global connectivity
• Inter-AS traffic normally flowed through the core AS
• Routers connected to other AS’s were called ”exterior gateways” and the Exterior Gateway Protocol (EGP) was developed for them
Autonomous Systems
• To make routing possible in an increasingly complex network, the Internet was divided into autonomous systems (AS)
• An autonomous system is responsible for its internal routing
• Router Rn of AS #n tells the core network router RCn, what networks can be reached through it
Core Network
...
R1
Autonomous System #1
RC1
R2
Autonomous System #2
RC2
Rn
Autonomous System #n
RCn
Interior Routing Protocols
Within an autonomous system an Interior Routing Protocol is used, such as:
• Routing Information Protocol (RIP)
• A distance vector protocol
• Traditional and easy to implement
• RIP v2 (RIP2) supports sub- and supernetting
• Doesn’t scale well to large networks
• Enhanced Interior Gateway Routing Protocol (EIGRP)
• The proprietary Cisco distance vector protocol
• Widely used and powerful but entirely Cisco-dependent
• Open Shortest Path First (OSPF)
• A link state protocol
• Based on Dijkstra’s Shortest Path First (SPF) algorithm
• The recommended interior routing protocol
Exterior Routing Protocols
Between autonomous systems an Exterior Routing Protocol (a reachability protocol) is used:
• Exterior Gateway Protocol (EGP)
• A distance vector protocol
• Old and now obsolete
• Border Gateway Protocol (BGP)
• A path vector protocol
• “The EGP of the 1990’s”
• BGP-4 is defined in RFC-1771 (March, 1995, the latest version) Classless Inter Domain Routing (CIDR)
• A supernetting technique
• An intermediate solution to the exhaustion of class B network numbers
• Requires the use of BGP-4
On Choosing Routing Protocols
• Most companies never have to deal with exterior routing, only interior routing
• Telcos and some large organizations have their own
autonomous systems and need to use an exterior routing protocol
• In non-redundant networks static routing often is adequate
• Cisco has delivered well over 50% of all backbone routers and pushes its proprietary EIGRP routing protocol
• Choosing a proprietary routing protocol is not to be
recommended because it ties the organization into one vendor
• Openness is worth some effort
• The routing protocols recommended by the Internet Architecture Board (IAB) are OSPF and BGP-4
