(Chapters 2 3 in Huitema) E7310/Internet basics/comnet 1

Transcription

1 Introduction to routing in the Internet Ethernet, switching vs. routing Internet architecture IPv4 Addressing Routing principles Protocols: IPv4, ICMP, ARP (Chapters 2 3 in Huitema) E7310/Internet basics/comnet 1

2 Background: Local Area Networks E7310/Internet basics/comnet 2

3 Ethernet Most widespread LAN technology Shared medium: Carrier sense multiple access with collision detection (CSMA/CD) Host 1 Host 2 Host 3 Host 4 Shared Ethernet segment Everyone receives everyone s traffic! Supports broadcast (=a host can send a packet to everyone) Limited length! (divide data into shorter pieces - frames) Each frame contains source and destination addresses and error-checking data so that damaged data can be detected and re-transmitted. E7310/Internet basics/comnet 3

4 Interconnecting Ethernet segments with repeaters Repeaters repeats traffic of one segment on the other segment Host 1 Host 2 Host 3 Host 1 Host 2 Host 3 Host 4 Repeater Host 5 Hubs are multi-port repeaters Hub Still everyone receives everyone s traffic! Limited number of repeaters! E7310/Internet basics/comnet 4

5 Bridges learn where hosts are and forward packets only to the segment of the destination Prior to discovery bridges/switches pass all traffic between segments Host 1 Host 2 Host 3 Allows different speeds in different segments Network size not physically limited Host 4 Host 5 Broadcast traffic sent Bridge/switch to all devices (if there is no FIB entry, broadcast/when bridge sees a frame, it learns addresses = MAC learning) A bridge/switch is a improved repeater. E7310/Internet basics/comnet 5

6 Example Host 1 Host 1 Host 2 Host 3 Host 4 Switch Host 2 Host 3 Host 4 Switch Host 1 has a packet to send to Host 4 The switch receives the packet and learns that Host 1 is on interface 1 The switch does not know Host 4 The packet is sent on all interfaces Host 4 has a packet to send to Host 1 The switch receives the packet and learns that Host 4 is on interface 4 The switch now knows that Host 1 is on interface 1 The packet is sent on interface 1 E7310/Internet basics/comnet 6

7 A spanning tree connects all segments without loops Spanning tree protocol Only one possible path between two devices loops are impossible Some bridges are redundant Redundant bridge 1 Note that the packet from host 1 to host 2 must be sent via two bridges, not one! 2 Bridge Bridge Active bridge Bridge A layer 2 network Bridge E7310/Internet basics/comnet 7

8 IP allows building large networks and interconnecting various technologies Internet is a network of networks IP adds a layer, the network layer, on top of the underlying network technologies A common packet format, the IP packet A common address, the IP address Routing end-to-end Routers implement packet forwarding on the network layer Ethernet Router ATM Router Token Ring Router Router Router Frame relay Router Ethernet A layer 3 network E7310/Internet basics/comnet 8

9 Issues to be addressed for the usage of Ethernet in WANs Avoid broadcasting of packets to an unknown destination Limit use of broadcasting Use the shortest path instead of a spanning tree Use all available links for load balancing Faster convergence (rapid spanning tree) Forwarding table size (flat topology) Peering, policies, address hierarchy Do we re-invent IP when these are implemented in Ethernet? E7310/Internet basics/comnet 9

10 Internet Architecture E7310/Internet basics/comnet 10

11 Internet Architecture Principles IP over everything by Vinston Cerf Internet connects different types of networks Each with different framing, addressing, Interconnection based on translation Mapping through a gateway Never perfect Detecting loops is difficult Translation still needed in many cases E.g. signaling interworking, IPv4 to IPv6 mapping Interconnection based on overlay Approach used by IP Single protocol over all underlying networks Simple to adapt to new technologies Define framing or encapsulation Define address resolution: IPaddress! network address Unique IP-address E7310/Internet basics/comnet 11

12 Internet Architecture Principles IP over everything HTTP, FTP, IMAP, SMTP,... TCP, UDP,... IP IEEE-802, ATM, X.25,... This is what has made IP so powerful! This is also what has made it so difficult to replace IP (e.g. with a newer IP)! E7310/Internet basics/comnet 12

13 Routed IP vs. switched Ethernet Routed IP Interworking between network technologies Complex, expensive One route per destination shorter routes load distribution Address allocation Hierarchical address space Switched Ethernet Works with Ethernets Simple, cheap, less layers All destinations use the same spanning tree longer routes congestion on the tree Fixed addresses Flat address space E7310/Internet basics/comnet 13

14 Classical Ethernet vs. IP Flat addressing! each FIB entry is for a single node Spanning tree! no loops Broadcast zone for frames with unknown destination-mac! broadcast storms! limited scalability to nr of network nodes MAC learning populates FIB = adaptive method Hierarchical addressing! aggregate routing entries TTL! packet killed if it loops Route to default route when destination unknown/ broadcast limited to special cases Routing protocols populate RIB and!fib = distributed method Carrier grade Ethernet: FIB populated by management layer, No-MAC-learning and no-broadcasting of frames with unknown dest MACs.! Scales better, more predictable service, but still does not scale as well as IP because of 1) flat addressing and 2) slow management layer. Routed Ethernet improves on the last argument but still just single domain. E7310/Internet basics/comnet 14

15 Internet Architecture Principles End-to-end principle Hop-by-hop control vs. Intelligence in the network Error and flow control on each hop Used in X.25 by David D. Clark End-to-end control Intelligence in end station Error and flow control in end station Used in IP The network can not be trusted The user must in any case check for errors Network control is redundant Error checking and flow control by TCP in the end stations No state information in the network The network is not aware of any connections Packets routed independently If a link fails, another route is used However, NATs (Network Address Translators) break this principle Interconnection; ambiguity caused by address hiding (private IP addresses) E7310/Internet basics/comnet 15

16 Internet Architecture Principles Connectivity is its own reward The value of a network increases in proportion to the square of the number of nodes on the network Be liberal with what you receive, conservative with what you send Try to do your best to understand what you receive Maximum adherence to standard when sending Robert Metcalf's law by Jon Postel Snowballing effect keeps all interested in connectivity, thus, keep adhering to standards E7310/Internet basics/comnet 16

17 Fundamentals, addressing, IPv4, ICMP E7310/Internet basics/comnet 17

18 The IP address defines the interface (not the host) Host 3 IP address C IP address F IP address B Router IP address D IP address E IP address A Host 2 Host 1 E7310/Internet basics/comnet 18

19 On the link layer, every interface has a media specific MAC address Host 3 IP address C MAC c IP address B MAC b Router IP address D MAC d IP address E MAC e IP address F MAC f Host 1 IP address A MAC a Host 2 This represents a subnet (e.g. an Ethernet segment), in which all hosts can send to each other without the help of a router. E7310/Internet basics/comnet 19

20 The Internet layer model in hosts and routers Host 1 Router Host 2 Application Layer URL, SIP address, Application Application Transport Layer port number TCP/ UDP TCP/ UDP Network Layer IP address IP IP IP Link Layer MAC address MAC MAC MAC Physical Layer Network 1 Network 2 Note that a router only operates on the network layer E7310/Internet basics/comnet 20

21 Layers and message forwarding Host 1 Router IP address B MAC b IP address A MAC a IP address C MAC c Host 2 IP address D MAC d Application Application TCP/UDP Router TCP/UDP IP A B IP C D IP MAC a b MAC c d MAC network 1 network 2 Encapsulation: a b, IP Ethernet header A D, TCP IP header TCP header Data Encapsulation: c d, IP Ethernet header A D, TCP IP header TCP header Data E7310/Internet basics/comnet 21

22 IPv4 address formats Originally a two-level (network, host) hierarchy bits MSB Network 0 7 bits 24 bits bits Host 16 bits bits 8 bits bits - multicast address For experimental and future use Class A B C D E E7310/Internet basics/comnet 22

23 IPv4 address formats A new level for easier network administration 1984 Example: Network Subnet Host Address: Mask: Network: first bit = Subnet: address* AND mask = Host: address AND NOT mask = address* = address with network part zeroed NB: Also written as / 14 Later updated by CIDR (Classless Inter-Domain Routing) (discussed later) E7310/Internet basics/comnet 23

24 IPv4 address examples Mask IP address Network Subnet Host /16 /23 /23 / A: B: C: Mask in traditional notation Mask in CIDR notation High order bits: ! A-class ! B-class ! C-class E7310/Internet basics/comnet 24

25 IP address classes A, B, C Network Type Max No. Of Networks Scope of IP Addresses Max No. Of Hosts Private Address Scope A 126 2^7-2) B 16384(2^14) C (2^21) E7310/Internet basics/comnet 25

26 The key functions of routing: 1. Packet forwarding On which interface should this packet be forwarded? Which is the following destination on that network? Host 3 address C address B Router address D address E address F address A Host 2 Host 1 Look in the routing table! E7310/Internet basics/comnet 26

27 The key functions of routing: 2. Collection of information Routers exchange routing information with routing protocols (e.g. RIP, OSPF, BGP) Routing Information Protocol (RIP) Open Shortest Path First (OSPF) Border Gateway Protocol (BGP) Router Router Router E7310/Internet basics/comnet 27

28 The key functions of routing: 3. Generation and selection of routes Routers implement a routing algorithm to generate the routing tables based on the collected information. The routing algorithm and routing protocol are tightly linked. Destination Interface Next hop Distance / / / / / E7310/Internet basics/comnet 28

29 Routers maintain routes to networks (not to hosts) Example Network / 16 Destination Interface Next hop Dist / / / / / Host 3 Network / Network / 16 Router Host 2 Host 1 E7310/Internet basics/comnet 29

30 Aggregation describes several destinations in a single entry to reduce size of routing tables Example Host 3 Network / 16 Network / Network / 16 Router Host 2 Host 1 Network / 8 E7310/Internet basics/comnet 30

31 Special purpose addresses An unknown network is replaced by 0 Only used as source address (e.g. a booting host) = this host in this network 0.X.Y.Z = host X.Y.Z in this network Limited broadcast address To all hosts in the local network Directed broadcast addresses A , B.B , C.C.C.255 To all hosts in a specified network Loopback-address 127.X.X.X (usually ) Internal in one host Multicast-addresses (e.g = all routers on this subnet) E7310/Internet basics/comnet 31

32 IPv4 packet header Version IHL Type of service Total length Identification Flag Fragment offset Time-to-live (TTL) Protocol Header checksum Source IP Address Destination IP Address Optional Padding 32 bits RFC- 791 E7310/Internet basics/comnet 32

33 IPv4 packet header Version IP version number. The current version is 4. IHL Type of service Identification Flag Fragment offset 32 bits Total length Time-to-live (TTL) Protocol Header checksum Internet header Contains 2 fields: length. Expressed Source packet IP Address priority and as number of Destination service IP Address type. 32-bit words. Optional Differentiated the Padding header Services reuses the field for DSCP. Expressed as number of octets in the payload and in RFC- 791 E7310/Internet basics/comnet 33

34 IPv4 packet header Version IHL Time-to-live (TTL) Type of service Identification Flag Fragment offset Protocol Source IP Address Total length Time-to-live. An upper bound on the Destination IP Address Expressed Used when as large time that an IP datagram can Optional exist in number packets are of octets an Internet system. The TTL field is in fragmented the payload when set by the sender of the datagram, and and underlying network reduced by every router on the route 32 bits the has header maximum packet to its destination. The packet length. is deleted when TTL reaches 0. Header checksum Padding E7310/Internet basics/comnet 34

35 IPv4 packet header Protocol, that the receiving host should use to process the packet, e.g. 6=TCP, 17=UDP, 1=ICMP, 89=OSPF Version IHL Time-to-live (TTL) Type of service Identification Flag Fragment offset Protocol Optional Protect the header of IPv4 data packets against data corruption. The header checksum is calculated as 16 bit one s complement sum Source IP Address Destination IP Address Total length Header checksum Padding IP address of the sender of the packet. IP address of the receiver of the packet 32 bits Used for special types of information or tricks. One packet can carry many option fields. E7310/Internet basics/comnet 35

36 The most important fields in routing are the destination address and the time-to-live Version IHL Type of service Total length Identification Flag Fragment offset Time-to-live (TTL) Protocol Source IP Address Destination IP Address Options Header checksum Padding Every router decrements the TTL! must calculate new checksum Options (e.g. source routing, record route, timestamp) rarely/never used in practice. E7310/Internet basics/comnet 36

37 The original Type of service field Precedence D T R C Precedence The packet with the highest precedence is first taken from the queue to be routed. Route selection criteria D minimization of delay T maximization of transmission capacity R maximization of reliability C minimization of cost Only one can be selected. In practice, these were not used E7310/Internet basics/comnet 37

38 DiffServ reuses the Type of service field DSCP ECN Differentiated Services Code Point (DSCP) Indicates the traffic class Each traffic class receives different behavior in case of congestion. Classes defined by the operator Explicit Congestion Notification (ECN) Used to mark packets when congestion occurs (instead of dropping them) 00 - Non ECN-Capable Transport 10 - ECN Capable Transport ECN Capable Transport Congestion Encountered Requires support from upper layer protocols (TCP) E7310/Internet basics/comnet 38

39 Options: Source routing allows the sender to determine the route Type Length Pointer Address 1 Address 2... Address N Common to all options Implemented with the source routing options Loose source routing (type 131, ) The packet is sent to the next address in the list using normal routing. Strict source routing (type 137, ) The packet is sent to the next address in the list. If there is no direct link to the address, the packet is destroyed. Slow and unsecure Rarely used Can be replaced by encapsulation: A!C, IP-IP A!B, TCP TCP Data E7310/Internet basics/comnet 39

40 ICMP Internet Control Message Protocol Purpose: Give feedback about the network operation. An ICMP packet is sent to the source if e.g. The destination is unreachable TTL expires All hosts and routers must support ICMP. ICMP messages are transported in IP packets If a ICMP message is dropped, a new one is not generated to avoid the snowballing effect. RFC- 792 E7310/Internet basics/comnet 40

41 ICMP messages Types: 0 - Echo reply (used for ping ) 3 - Destination unreachable 4 - source quench (= slow down ) (dropped from recommendations) 5 - Redirect 8 - Echo (used for ping ) 9 - Router advertisement 10 - Router solicitation 11 - Time exceeded (TTL reached 0) 12 - Parameter problem 13 - Timestamp 14 - Timestamp reply 15 - Information request 16 - Information reply Codes for Destination Unreachable 0 - network unreachable 1 - host unreachable 2 - protocol (TCP/UDP/OSPF) unreachable 3 - port unreachable 4 - fragmentation needed and the DF-bit (Don t Fragment) is set 5 - source route failed Type Code Header checksum 0-field IP header + leading 8 octets of original datagram 32 bits E7310/Internet basics/comnet 41

42 Packet sending how to determine the next hop The sender checks if the destination address is in the same sub-network by comparing the masked values of the source and destination address. If same, the destination is in the same subnet (next hop = destination). Otherwise, the packet must be sent to a router (next hop = router). It then obtains the media address (MAC-address) of the next hop using the ARP-protocol. (ARP: Address Resolution Protocol) Sender ARP request (broadcast) ARP reply The destination recognizes its own address The media address is stored in the cache. Note: All hosts in the same subnet store the address in their cache. E7310/Internet basics/comnet 42

43 ARP Address Resolution Protocol Purpose: Translate an IP-address to a MAC-address ARP maps IP to the underlying protocol Each network technology requires its own ARP adaptation. Easy if the network supports broadcast or multicast. E.g. Ethernet, Token Ring, FDDI Broadcast to everyone, the indicated destination replies ATM requires a special ARP-server Manually defined address for point-to-point links E.g. X.25, ISDN, Frame-Relay Works on top of Ethernet (not on top of IP) RFC- 826 E7310/Internet basics/comnet 43

44 Router discovery How does the host know the address of the router? Configure manually Obtain with DHCP Dynamic Host Configuration Protocol Configured by administrator, still needs manual work Listen to routing protocols Uses resources of the host, too many routing protocols! not used today Automatic router discovery with ICMP network Y X LAN1 LAN2 B E7310/Internet basics/comnet 44

45 ICMP router discovery (1) The routers send router advertisements to all hosts periodically (e.g. in 7 minute intervals) network ICMP messages: Y Router advertisement (to all hosts) X LAN1 LAN2 B List of the router s addresses. The preference of the addresses, which are used to identify the normal, reserve, etc. router or router address (the preference of the default router is highest) Lifetime of the information (e.g. 30 min) RFC E7310/Internet basics/comnet 45

46 ICMP router discovery (2) The host would have to wait up to 7 minutes before it can send packets outside its sub-network. Using a router solicitation, the host gets the advertisement immediately network X ICMP messages: Router advertisement Y Router solicitation (to all routers) LAN1 B LAN2 E7310/Internet basics/comnet 46

47 ICMP router discovery (3) The host chooses the router with the highest priority as its default router (default gateway). All packets for destinations outside the sub-network are then sent to the default router. Any advertisement from a router outside the sub-network is discarded E7310/Internet basics/comnet 47

48 A network may have many routers, the closest to the destination should be used A packet sent through the default router reaches the destination, but may waste resources A Packet A! B (a!x) Y network X Default router Packet A! B (x!y) Packet A! B (y!b) B E7310/Internet basics/comnet 48

49 A network may have many routers, the closest to the destination should be used The router can send a redirect to indicate a shorter route to the destination A Packet A! B (a!x) Y Type Code Header checksum IP address! router=y IP header + 8 octets of the original network datagram Type X Default Code router 5 redirect 0 redirect for network (no mask!) 1 redirect for host 2 for type of service and network 4 for type of service and host Packet A! B (x!y) Packet A! B (y!b) ICMP redirect use router Y instead Packet A! B (a!y) Packet A! B (y!b) B E7310/Internet basics/comnet 49

50 Host must have feedback from the first router to avoid sending to a black hole Feedback may be TCP acknowledgements Router advertisements ARP-replies ICMP echo reply (ping) Between routers, routing protocols provide similar feedback and help in detecting failed router neighbors. E7310/Internet basics/comnet 50

51 Virtual Router Redundancy Protocol (VRRP) network A X Master Y Backup Router advertisement from virtual MAC address E XX I m alive! I m alive! Purpose: Provide a backup for the default router Several physical routers advertise a virtual router to the hosts Only one of the physical routers (the master router) is active, doing the actual routing on behalf of the virtual router The master router periodically sends multicast packets to all VRRP routers Based on Cisco's Hot Standby Router Protocol (HSRP) RFC E7310/Internet basics/comnet 51

52 Virtual Router Redundancy Protocol (VRRP) network A X Master Y Backup Router advertisement from virtual MAC address E XX I m alive! I m alive! X Router advertisement from the same virtual MAC address E XX I m alive! Timeout Becomes Master If no multicast message is received within a timeout, the master router is considered to have failed and a backup router becomes the master router RFC E7310/Internet basics/comnet 52

53 DNS Domain Name Service Purpose: Map a host name to an IP address Why DNS? Easier to remember names than addresses Allows address changes without changing the name Several addresses per host Several names per host Extensions: service location (e.g. server), ENUM DNS does not affect routing, routers only deal with IP addresses RFC RFC E7310/Internet basics/comnet 53

54 Routing algorithms E7310/Internet basics/comnet 54

55 Internet routing is based on routing protocols, which collect information Routing is completely automatic and distributed No offline route planning Only dimensioning is made offline The routers communicate with a routing protocol The routing protocol implements a routing algorithm that finds the shortest (cheapest, fastest, etc.) route to every destination E7310/Internet basics/comnet 55

56 Routing in the Internet is generally dynamic, but static routing is used in some cases Dynamic routing is based on routing protocols which create and maintain the routing tables automatically examples of routing protocols are RIP, OSPF, BGP... Dynamic routing is cooperative in nature and assumes an implicit trust between the routers. Static routing is based on manually configured routing tables. Static routing is used when e.g. two peer providers do not trust each other To connect an organization to a service provider with a single connection Static routing is difficult to maintain E7310/Internet basics/comnet 56

57 Routing is divided into interior and exterior exterior neighbors Autonomous System (AS) = networks operated by a single organization and having a common routing strategy border router - at least one neighbor belongs to another autonomous system interior neighbors In this course we only deal with interior routing. Exterior routing in S Network Service Provisioning E7310/Internet basics/comnet 57

58 Routing is divided into interior and exterior Interior routing protocols Routing Information Protocol (RIP), RIP-2 Open Shortest Path First (OSPF) Interior Gateway Routing Protocol (IGRP), EIGRP Intermediate System-to-Intermediate System (IS-IS) Exterior routing protocols External Gateway Protocol (EGP) (historical) Border Gateway Protocol version 4 (BGP-4) E7310/Internet basics/comnet 58

59 Routing algorithms Proactive vs. reactive Proactive The router creates and maintains routes to all destinations The routes are available in advance The routing algorithms in the Internet are proactive Reactive Routes are created only when they are needed Used in e.g. ad hoc networks (discussed later in this course) E7310/Internet basics/comnet 59

60 Routing algorithms Distance vector vs. link state Distance vector Distance vectors are sent, until the state of the network is stable Routers only have a local view of the network The routers cooperate to generate the routes Example: RIP Link state Topology descriptions are sent periodically and nodes generate a map over the network Every router generates the routes independently of the other routers Example: OSPF E7310/Internet basics/comnet 60

61 Routing algorithms Distance vector vs. link state Distance vector Simple and lightweight + Slow convergence - Only one route per destination - Only one metric - Link state Complex and heavy - Fast convergence + Several routes per destination + Supports different metrics + E7310/Internet basics/comnet 61

62 Hop-by-hop routing vs. source routing Hop-by-hop routing (not to be confused with hop-by-hop error & flow control) The sender specifies only the destination Intermediate nodes use a routing table to forward the traffic closer to the destination Requires consistent routes in all nodes Internet: Used in the Internet PSTN: Corresponds to SOC (Sequential Office Control) Source routing The sender lists the nodes through which the traffic should be sent Requires that the sender has complete information (e.g. from OSPF) of the network Internet: Possible with source routing option but rarely used (if route inserted by a host, it can not be trusted) PSTN: Corresponds to OOC (Originating Office Control) Loose source routing can be thought as a mix between these E7310/Internet basics/comnet 62

63 Address prefixes in BGP routing table E7310/Internet basics/comnet 63

64 Number of Autonomous Systems Internet AS Allocated AS AS announced on the Internet AS Date E7310/Internet basics/comnet 64