Computer Networks. College of Computing. Copyleft 2003~2018

Transcription

1 Computer Networks

2 Computer Networks Prof. Lin Weiguo Coege of Computing Copyeft 2003~

3 Attention The materias beow are avaiabe for use by others. Instructors are wecome to use them in their own courses, downoad them to their own cass' web site, or modify them to suit. However, you must acknowedge the source of the origina and not attempt to pace your own copyright on this materia. Thanks to 3

4 Roadmap Introduction Physica Layer Data Link Layer Medium Access Subayer Network Layer Transport Layer Appication Layer 4

5 Chapter 5 The Network Layer

6 Now I have come to the crossroads in my days, and I have aways known the right path, aways, without exception, I knew. But I never took it, you know why? Because it's too damn hard. ---A Pacino, scent of a woman inwei@cuc.edu.cn 6

7 Do you Taobao? ogistics :42: :59: :03: :04: :18: :32: :29: :35: :36: :57: :24: :01:11 B B,,, B,,,,,, :03: :36: :02: :05: :53: :50: :51:02,, inwei@cuc.edu.cn 7

8 What you wi earn: The Network Layer is concerned about getting packets from source to destination, no matter how many hops it may take. It s a about routing. 5.1 Network ayer design issues 5.2 Routing agorithms 5.3 Congestion contro agorithms 5.4 Quaity of service 5.5 Internetworking 5.6 The Networking ayer in the internet inwei@cuc.edu.cn 8

9 5.1 Network Layer Design Issues Store-and-Forward Packet Switching Services Provided to the Transport Layer Impementation of Connectioness Service Impementation of Connection-Oriented Service Comparison of Virtua-Circuit and Datagram Networks 9

10 OSI Reference Mode 10

11 Communication subnet 11

12 Store-and-Forward Packet Switching ISP s equipment xdsl The environment of the network ayer protocos. inwei@cuc.edu.cn 12

13 Store-and-Forward Packet Switching Mechanism A host with a packet to send transmits it to the nearest router, either on its own LAN or over a point-to-point ink to the carrier. The packet is stored there unti it has fuy arrived so the checksum can be verified. Then it is forwarded to the next router aong the path unti it reaches the destination host, where it is deivered inwei@cuc.edu.cn 13

14 Who-Does-What The network ayer is responsibe for routing packets from the source to destination. The routing agorithm is the piece of software that decides where a packet goes next (e.g., which output ine, or which node on a broadcast channe). For connectioness networks, the routing decision is made for each datagram. For connection-oriented networks, the decision is made once, at circuit setup time. inwei@cuc.edu.cn 14

15 Design Goa of the Network Layer 1. Services independent of router technoogy. 2. Transport ayer shieded from number, type, topoogy of routers. 3. Network addresses avaiabe to transport ayer use uniform numbering pan - even across LANs and WANs inwei@cuc.edu.cn 15

16 The Discussion Does error/fow contro beong in network or transport ayer? (Whether the network ayer shoud provide connectionoriented service or connectioness service) 16

17 Internet community viewpoint The subnet is inherenty unreiabe no matter how it is designed. Thus, hosts are forced to do error/fow contro anyway. Given that they perform error/fow contro, why have the network ayer dupicate the same function? The TCP/IP Internet is connectioness in its impementation but provides connections to its users. 17

18 Teephone Companies Viewpoint The connection oriented approach is the right way. Users don't want compex error/fow contro protocos in host machines. User's want reiabe, troube-free service. Quaity of Service is the dominant factor. 18

19 Datagram vs. Virtua Circuit Datagram for connectioness service: Each packet sent is routed independenty of its predecessors. Decisions are made "on the fy", so more computing required, however this method is more robust. Virtua Circuit for connection-oriented service: A path from the source router to the destination router must be estabished before any data packets can be sent. Avoids choosing a new route for each packet. A virtua circuit is a state -- it remembers how to send a packet from source to destination. This state is hed in the subnet, in the source DLL or in each of the DLL ayers aong the route. inwei@cuc.edu.cn 19

20 Impementation of Connectioness Service ISP s equipment A s tabe (initiay) A s tabe (ater) C s Tabe E s Tabe Every router has an interna tabe teing it where to send packets for each possibe dest. Each tabe entry is a pair consisting of a dest and the outgoing ine to use for that dest. Routing within a datagram network. inwei@cuc.edu.cn 20

21 Impementation of Connection- Oriented Service ISP s equipment A s tabe C s Tabe E s Tabe Incoming packets with connection id 1 is to be sent to router C and given id 1 Routing within a virtua-circuit network. inwei@cuc.edu.cn 21

22 Labe Switching Virtua Circuit Ca setup, teardown for each ca before data can fow Each packet carries VC identifier (not destination host ID) Every router on source-destination path maintains state for each passing connection Transport-ayer connection ony invoved two end systems Link, router resources (bandwidth, buffers) may be aocated to VC MPLS(MutiProtoco Labe Switching) Used within ISP networks in the Internet, with IP packets wrapped in an MPLS header having a 20-bit connection ID. inwei@cuc.edu.cn 22

23 Comparison of VC and Datagram Networks 23

24 Trade-offs between VC and datagram Setup time versus address parsing time Using VC requires a setup phase, which takes time. In a datagram subnet, a more compicated ookup procedure is required to ocate the entry for the destination. Overhead and bandwidth If the packets tend to be fairy short, a fu dest. address in every packet may represent a significant amount of overhead and wasted bandwidth. Amount of tabe space in router memory Datagram network needs entry for each possibe destination Whereas a VC network just needs an entry for each VC inwei@cuc.edu.cn 24

25 5.2 Routing Agorithms The Optimaity Principe Shortest Path Routing Fooding Distance Vector Routing Link State Routing Hierarchica Routing Broadcast Routing Muticast Routing Anycast Routing Routing for Mobie Hosts Routing in Ad Hoc Networks 25

26 Routing and Forwarding One can think of a router as having two processes inside it. Forwarding process: handes each packet as it arrives, ooking up the outgoing ine to use for it in the routing tabes. Routing process: responsibe for fiing in and updating the routing tabes. --that is where the routing agorithm comes into pay. inwei@cuc.edu.cn 26

27 Routing Agorithms Routing is concerned with the question: Which ine shoud router J use when forwarding a packet to router K? Goa: determine good path (sequence of routers) through network from source to destination The routing agorithm is that part of the network ayer software responsibe for deciding which output ine an incoming packet shoud be transmitted on. inwei@cuc.edu.cn 27

28 Goas for Routing Agorithms Correctness and Simpicity Robustness networks are never taken down; individua parts (e.g., inks, routers) may fai, but the whoe network shoud not. During years of continuous operation, being abe to hande a kinds of hardware and software faiures. Being abe to hande changes in topoogy and traffic patterns. Stabiity A stabe agorithm reaches equiibrium and stays there. It shoud converge quicky too. Fairness and efficiency Often contradictory inwei@cuc.edu.cn 28

29 Fairness vs. Efficiency Network with a confict between fairness and efficiency. inwei@cuc.edu.cn 29

30 Types of Agorithms Non-adaptive agorithms Routes never change once initia routes have been seected. Aso caed static routing. Because it does not respond to faiures, static routing is mosty usefu for situations in which the routing choice is cear. Adaptive agorithms Use such dynamic information as current traffic, topoogy, deay, etc. to seect routes. Differ in where they get information(ocay, adjacent routers, or a routers), when they change the routes(when changes happen, every Δt), and what metric is used for optimization(distance, number of hops, estimated transit time). inwei@cuc.edu.cn 30

31 The Optimaity Principe This simpy states that if router J is on the optima path from router I to router K, then the optima path from J to K aso fas aong this same path. This means we can form a sink tree (optima paths to B) as in the next Figure. I J K inwei@cuc.edu.cn 31

32 Sink tree Sink tree: the set of optima routes from a sources to a given destination form a tree rooted at the destination. The goa of a routing agorithms is to discover and use the sink trees for a routers. (a) A network. (b) A sink tree for router B. inwei@cuc.edu.cn 32

33 Graph abstraction Graph abstraction for routing agorithms: 1.graph nodes are routers 2.graph edges are physica inks ink cost: deay, $ cost, or congestion eve To choose a route between a given pair of routers: find Good path typicay means minimum cost path other definitions possibe inwei@cuc.edu.cn 33

34 Shortest Path Routing Often used because simpe and easy to understand Shortest Path Metrics (Path Length) Number of Hops Physica Distance Mean Queuing and Transmission Deay Bandwidth Average Traffic Communication Cost 34

35 Dijkstra's agorithm Use Dijkstra's agorithm (or variation) (SPF, Shortest Path First agorithm) Basic idea is: Choose the source, and put nodes connected to source in ist to consider. From the ist to consider choose the nearest node. Note: Edsger Wybe Dijkstra ( ; [ˈɛtsxər ˈwibə ˈdɛɪkstra]) 35

36 The Agorithm on a Weighted, Undirected Graph The first 6 steps used in computing the shortest path from A to D. The arrows indicate the working node. inwei@cuc.edu.cn 36

37 Dijkstra s Agorithm 1 Initiaization: 2 N = {A} 3 for a nodes v 4 if v one-step reachabe from A 5 then D(v) = c(a,v) 6 ese D(v) = infinity 7 8 Loop 9 find w not in N such that D(w) is a minimum 10 add w to N 11 update D(v) for a v one-step reachabe from w and not in N: 12 D(v) = min( D(v), D(w) + c(w,v) ) 13 /* new cost to v is either od cost to v or known 14 shortest path cost to w pus cost from w to v */ 15 unti a nodes in N inwei@cuc.edu.cn 37

38 Dijsktra s Agorithm 5-8 top Dijkstra's agorithm to compute the shortest path through a graph. inwei@cuc.edu.cn 38

39 Dijsktra s Agorithm 5-8 bottom Dijkstra's agorithm to compute the shortest path through a graph. inwei@cuc.edu.cn 39

40 Fooding Fooding is a form of isoated routing. Does not seect a specific route. When a router receives a packet, it sends a copy of the packet out on each ine (except the one on which it arrived). To prevent packets from ooping forever, each router decrements a hop count contained in the packet header. Whenever the hop count decrements to zero, the router discards the packet. inwei@cuc.edu.cn 40

41 Fooding Reduce Looping Add a sequence number to each packet. Each router maintains a private sequence number. When it sends a new packet, it copies the sequence number into the packet, and increments its private sequence number. For each source router S, a router: a) Keeps track of the highest sequence number seen from S. b) Whenever it receives a packet from S containing a sequence number ower than the one stored in its tabe, it discards the packet. c) Otherwise, it updates the entry for S and forwards the packet on inwei@cuc.edu.cn 41

42 Fooding uses Fooding has severa important uses: In miitary appications, the network must remain robust in the face of (extreme) hostiity Sending routing updates, because updates can't rey on the correctness of a router's routing tabe. Theoretica-chooses a possibe paths, so it chooses the shortest one. inwei@cuc.edu.cn 42

43 Distance Vector Routing Aso known as Beman-Ford, used in the Internet under name RIP Each router maintains a tabe (vector) giving the best known distance to a destination and the ine to use for sending there. Tabes are updated by exchanging information with neighbors. Each router knows the distance (cost) of reaching its neighbors (e.g. send echo requests). Routers periodicay exchange routing tabes with each of their neighbors. Upon receipt of an update, for each destination in its tabe, a router: Compares the metric in its oca tabe with the metric in the neighbor's tabe pus the cost of reaching that neighbor. if the path via the neighbor has a ower cost, the router updates its oca tabe to forward packets to the neighbor inwei@cuc.edu.cn 43

44 D-V Routing Exampe destination Metric is deay JàA: 8 ms JàI: 10ms JàH: 12ms JàK: 6ms cost Outgoing ink to use (a) A network. (b) Input from A, I, H, K, and the new routing tabe for J. inwei@cuc.edu.cn 44

45 The Count-to-Infinity Probem This agorithm was used in the origina ARPANET. Unfortunatey, it suffers from the probem: convergence takes time, good news traves quicky, bad news traves sowy (count-to-infinity probem). Convergence: the setting of routes to best paths across the network is caed convergence. 45

46 The Count-to-Infinity Probem Suppose A is down initiay and a the Others know this. When A comes up, the other routers Learn about it via the vector exchanges. A routers are initiay up, and Suddeny A goes down The count-to-infinity probem. inwei@cuc.edu.cn 46

47 The Count-to-Infinity Probem The core of the probem is: When X tes Y that it has a path somewhere, Y has no way of knowing whether it itsef is on the path. Attempts to sove: Spit horizon with poisoned reverse in RFC 1058 inwei@cuc.edu.cn 47

48 Probems with D-V The `od' Arpanet routing agorithm was repaced in Probems with od agorithm incuded: High-priority routing update packets were arge, adversey affecting traffic. Network was too sow in adapting to congestion, too fast to react to minor changes. Average queue ength was used to estimate deay. This works ony if a ines have the same capacity and propagation deay. Doesn't take into account that packets have varying sizes. inwei@cuc.edu.cn 48

49 Link State Routing Each router maintains a database describing the topoogy and ink deays between each router. That is, each router keeps track of the fu graph of inks and nodes. Avoids the count to infinity probem since a routers get each other router s information. Each router uses an SPF agorithm to cacuate shortest paths based on the current vaues in its database. Because each router makes its cacuation using the same information, better routing decisions are made. inwei@cuc.edu.cn 49

50 Five Steps Each router must do the foowing: 1. Discover its neighbors, earn who its neighbors are. 2. Set distance/cost metric to each of its neighbors. 3. Construct a packet teing a it has just earned. 4. Send this packet to, receive packets from a other routers. 5. Compute the shortest path to every other router. inwei@cuc.edu.cn 50

51 Learning about the Neighbors When a router is booted, its first task is to earn who its neighbors are. It accompishes this goa by sending a specia HELLO packet on each point-to-point ine. The router on the other end is expected to send back a repy giving its name. These names must be gobay unique. inwei@cuc.edu.cn 51

52 A graph mode of a network Modeing the LAN: introduced a new, artificia node, N, to which A,C, and F are connected. One designated router on the LAN is seected to pay the roe of N in the routing protoco. (a) Nine routers and a LAN. (b) A graph mode of (a). inwei@cuc.edu.cn 52

53 Setting Link Cost The cost to reach neighbors can be set automaticay, or configured by the network operator. A common choice is to make the cost inversey proportiona to the bandwidth of the ink. If the network is geographicay spread out, the deay of the inks may be factored into the cost so that paths over shorter inks are better choices. The most direct way to determine this deay is to send over the ine a specia ECHO packet that other side is required to send back immediatey. By measuring the round-trip time and dividing it by two, the sending router can get a reasonabe estimate of the deay. inwei@cuc.edu.cn 53

54 Buiding Link State Packets ID of the sender Sequence number (32-bits, one per second, wrap around after 137 years) Age for the packet (decrement once per second) A ist of neighbors and deay for it (a) A network. (b) The ink state packets for this subnet. inwei@cuc.edu.cn 54

55 Distributing the Link State Packets Basic distribution agorithm: Fooding To keep the food in check, each packet contains a sequence number that is incremented for each new packet sent. Routers keep track of a the (source router, sequence) pairs they see (in a ink state database). When a new ink state packet comes in, it is checked against the ist of packets aready seen. If it is new, it is forwarded on a ines except the one it arrived on. If it is a dupicate, it is discarded. If a packet with a sequence number ower than the highest one seen so far ever arrives, it is rejected as being obsoete since the router has more recent data. inwei@cuc.edu.cn 55

56 The Age fied Probems: Sequence number confusion Router crashes and ose track of its sequence num. A sequence number is corrupted. Soution: The Age fied Incude the age of each packet after the Seq. and decrement it once per second. When the age (in the router s ink state database) hits zero, the information from that router is discarded. The age fied is aso decremented by each router during the initia fooding process, to make sure no packet can get ost and ive for an indefinite period of time (a packet whose age is zero is discarded). inwei@cuc.edu.cn 56

57 Agorithm Refinements When a ink state packet comes in to a router for fooding, it is not queued for transmission immediatey. Instead it is first put in a hoding area to wait a short whie. If another ink state packet from the same source comes in before the first packet is transmitted, their sequence numbers are compared. If they are equa, the dupicate is discarded. If they are different, the oder one is thrown out. To guard against errors on the router-router ines, a ink state packets are acknowedged. inwei@cuc.edu.cn 57

58 Exampe LSP buffer for router B To reduce the numbers of ink state packets, wait for a short time and set 2 fags before forward the received ink state packets : the third packet, from E, is different. It arrived twice, once via EAB and once via EFB it has to be sent ony to C but acknowedged to both A and F The send fags mean that the packet must be sent on the indicated ink. The ack fags mean that it must be acknowedged there. The packet buffer for router B in the previous side (Fig. 5-12a). inwei@cuc.edu.cn 58

59 Computing the New Routes Once a router has accumuated a fu set of ink state packets, it can construct the entire network graph because every ink is represented. Every ink is, in fact, represented twice, once for each direction. The different directions may even have different costs. Now Dijkstra's agorithm can be run ocay to construct the shortest path to a possibe destinations. The resuts of this agorithm can be instaed in the routing tabes, and norma operation resumed. inwei@cuc.edu.cn 59

60 Limitations of Link State Routing For arge networks, the memory required to store the input data and the computation time can be an issue. Nevertheess, in many practica situations, ink state routing works we because it does not suffer from sow convergence probems. 60

61 Exampe Protocos using Link State Routing Agorithm IS-IS (Intermediate System-Intermediate System) in some Internet backbones. OSPF (Open Shortest Path First) in Internet (5.6.6) 61

62 Hierarchica Routing Hierarchica routing. 62

63 Broadcast Routing Sending a packet to a destinations simutaneousy is known as broadcasting. There are severa ways to impement broadcasting: For Broadcast Networks: The impementation is trivia: designate a specia address as the `a hosts address'. inwei@cuc.edu.cn 63

64 Broadcast Routing For Non-Broadcast Networks: 1. Send a unicast packet to each destination. However, this approach makes poor use of resources. 2. Food packets to a nodes. Fooding generates many packets and consumes too much bandwidth. 3. Use muti-destination routing: a) Each packet contains a ist (or bitmap) of a destinations, and when a router forwards a packet across two or more ines, it spits the packet and divides the destination addresses accordingy. b) This approach is simiar to sending uni-cast packets, except that we don't send individua copies of each messages. c) However, the copy operations sow down the abiity of a router to process many packets. inwei@cuc.edu.cn 64

65 Broadcast Routing For Non-Broadcast Networks: 4. Use a spanning tree. If the network can be reduced to a tree a) (There s ony one path between any two pairs of routers), copy a packet to each ine of spanning tree except the one on which it arrived. b) Works ony if each router understands the same spanning tree. 5. Reverse Path Forwarding (RPF): a) Use a sink tree (assume sink/source trees are the same). b) If a packet, originating from X, arrives on a ine of the sink tree eading to X, the packet is traveing aong the shortest path, so it must be the first copy we've seen. c) Copy the packet to a outgoing ines of the sink tree. If the packet arrives on another ine, assume that the packet is a copy - it didn't arrive on the shortest path - and discard it. RPF is easy to impement and makes efficient use of bandwidth. inwei@cuc.edu.cn 65

66 Reverse path forwarding Reverse path forwarding. (a) A network. (b) a Sink tree for router I. (c) The tree buit by reverse path forwarding. inwei@cuc.edu.cn 66

67 Muticast Routing (a) A network. (b) A spanning tree for the eftmost router. (c) A muticast tree for group 1. (d) A muticast tree for group 2. inwei@cuc.edu.cn 67

68 Muticast Routing (a) Core-based tree for group 1. (b) Sending to group 1. 68

69 Anycast Routing (a) Anycast routes to group 1. (b) Topoogy seen by the routing protoco. 69

70 Routing for Mobie Hosts Mobie Hosts - machines that are not currenty connected to their home ocation. Must get traffic to base host. Need intermediary agents. Look at Figures Base host uses encapsuation (tunneing) to send packet to mobie host. inwei@cuc.edu.cn 70

71 Routing for Mobie Hosts Packet routing for mobie hosts 71

72 Routing in Ad Hoc Networks (a)range of A s broadcast. (b) After B and D receive it. (c) After C, F, and G receive it. (d)after E, H, and I receive it. The shaded nodes are new recipients. The dashed ines show possibe reverse routes. The soid ines show the discovered route. inwei@cuc.edu.cn 72