1 Chapter 7—packet-switching Chapter 5: data-link layer protocols (some protocols are nearly same for transport layer) Chapter 6: MAC sublayer (for broadcast networks). This chapter: network layer –Telephony networks (i.e., circuit-switching) for stream of real time voice –Packet-switching networks for short messages, burst information, as well as real-time applications.

2 Two views of networks External view: what services provided to transport layer by a network (i.e. network layer) –Need connection setup or not –What QoSs are provided –Services should be independent from underlying networks so that transport layer can run over any networks as long as the networks provide the services Internal view: –physical topology, – datagram message transfer or virtual circuit information transfer, –addressing and routing, congestion control. In this chapter, networks indicate systems to transfer information from one end to the other. It is reasonable to consider networks to be network layer here.

3 Comparisons of two views by examples Broadcast networks and packet-switched networks From external view, both networks provides transfer of information between users, not too much different But from internal view, very different: –A broadcast network (such as LANs) is small, addressing is simple, frame transferred in one hop so no routing is needed –In a packet-switching network, addressing must accommodate large-scale networks and routing is necessary.

4 What we will discuss Network services and internal network operations Physical view of networks, Datagram and virtual circuit Routing Shortest path algorithms ATM networks Traffic management and congestion control (if have time)

5 Figure 7.1 t0t0 t1t1 Network Transfer of a block of info. VS. a sequence of blocks

6 Services provided by network layer Two basic kinds of services –One view: Transfer of a block of information VS. a sequence of blocks –Second view: connectionless VS. connection-oriented. –IP provides connectionless service VS. ATM provides connection-oriented service Other services such as: –Best-offer connectionless services –Low-delay connectionless service –Connection-oriented reliable stream service –Connection-oriented transfer of packets with delay and bandwidth guarantees End-to-end argument: –Functions should be placed as close as possible to the application since the application is in best position to determine whether a function is being carried out completely and correctly, –Therefore, network layer should provide minimum functions required to meet application requirements and performance. Leave more functions to upper layers such as transport layer.

7 Figure 7.2 Physical layer Data link layer Physical layer Data link layer End system  Network layer Network layer Physical layer Data link layer Network layer Physical layer Data link layer Network layer Transport layer Transport layer Messages Segments End system  Network service Network service protocol stack: network provides minimum required services to transport layer

8 Discussion of connectionless vs. connection-oriented Both connection-oriented and connectionless services in the upper layer can be implemented over a connectionless network (layer) (such as IP), similarly, can be implemented over a connection-oriented network (layer) (such as ATM). In network layer, connectionless service is also called datagram service, connection-oriented service called virtual-circuit service. Which is preferred? From end-to-end argument, connectionless network (IP) is preferred because IP makes intermediate nodes as simple as possible and put the burden on end systems. Moreover, it makes the network grow easily (more scalable). Why ATM which is connection-oriented? Main reason is QoS because connection-oriented network can easily guarantee bandwidth, delay, etc. Detail motivation behind ATM will be discussed later. Network layer: connectionless(IP) connection-oriented (ATM) Transport layer: connectionless connection-oriented Thinking: differences (or implications) of connection-oriented in transport layer and in network layer. What the connection between two ends in transport layer means?

9 Most basic functions of a network (layer) Routing and forwarding. Priority and scheduling (to provide QoS) Congestion control (also in transport layer) Segmentation (to deal with different frame sizes of underlying systems) Alternatively, network (layer) sends error message to let edge system to do segmentation. Addressing 1. to deal with different address formats when interconnecting networks. 2. Hierarchical address for scalability.

10 Figure MUX Network access Node Packet network topology --Access multiplexer Multiplexer combines multiple bursty flows into one aggregated flow. Access node forwards packets into backbone packet networks.

11 LAN Bridge LAN 1 LAN 2 (a) (b) Figure 7.5 Packet network topology --LAN Now nearly every LAN is connected to packet-switching network (the Internet) to share information globally and to make the Internet extremely large. LANs were introduced for sharing of resources in a organization. They are basic building components for Wide Area Networks.

12 R R R R S SS s s s s ss s ss s R s R Backbone To internet or wide area network Organization Servers Gateway Departmental Server Figure 7.6 Campus network LANs  extended LANs  subnetworks (by backbone networks)  campus network  university network, finally connect to Internet, bridges, routers and gateway,… play key-role in networks Backnone network may be connected by links, LAN, or ATMs. Servers provide various services.

13 Interdomain level Intradomain level LAN level Autonomous system or domain Border routers Figure 7.7 Internet service provider Intradomain and interdomain levels Domain: indicate the routers run the same routing protocols. Autonomous systems: included domains under a single administrations Various networks are connected through ISP. Tremendous small office and home (SOHO) users connect to the Internet through ISP, using dynamic IPs.

14 RARA RBRB RCRC Route server NAP National service provider A National service provider B National service provider C LAN NAP (a) (b) Figure 7.8 National service providers are connected by NAPs (national access point) Route server distributes routing information to routers.

15 Packet-switching topology What a big picture of topology for picket-switching networks!!! – Hierarchical structure, from LANs  campus, university and organization networks  Internet –ISPs provide backbone networks to connect tremendous different networks and enormous home PCs –Switches (bridges, routers, gateways): key elements –Various servers provides rich services to users –Domain, intradomain, interdomain for easy administration and network management.

16 Network Packet switch Transmission link Figure 7.9 Simplified picture of switched networks 1.Transmission lines and packet switches 2.They provide connectivity between any source and destination dynamically 3. The resources are allocated when needed, so can be shared among multiple users 4. Information can be transferred in connection-oriented or connectionless manner.

17 Brief discussion of switches Switches are generic term for switching devices in switched networks –Circuit-switches in telephony network –LAN switches in LAN connection, i.e. bridges, Ethernet switch,… –Packet switches in pack switching networks, i.e. routers and gateways.

18 Control N Line Card Interconnection Fabric Line Card N Figure 7.10 … … … … Structure of a packet switch/router Components: input ports, output ports, interconnection fabric, controller A line card handles several pairs of input/output ports and implements physical layer and data link layer functions: symbol timing, line coding, framing, physical addressing, error checking, MAC protocol, data link protocol, and buffering (speed mismatch between lines and fabric).

19 CPU N NIC Card Main Memory I/O Bus Figure 7.11 … … A PC or workstation can be configured to become a switch 1.Several NICs are inserted into expansion slots, 2. Install routing and other protocols in the computer Operations: 1. NIC gets frames from networks and de-encapsulate packets 2. Packets are transferred into memory using I/O bus 3. CPU performs required routing and protocol processing 4. Packets are transferred from memory to an appropriate NIC 5. The NIC forms new frame and sends out

20 Three basic resources and bottlenecks in switches Processing, memory and bus bandwidth –Processing implements protocols, hence processing capacity places a limit on maximum rate at which switch can operate –Memory stores packets, hence the amount of memory determines the rate at which packets are lost, placing another limit on switch load, moreover memory bandwidth also place limit on switch rate –I/O bus bandwidth places a limit on total rate at which information can be transferred between ports.

N 1 2 N Figure 7.12 … … Input port demultiplexes incoming packet stream; packets are routed to output port; output port multiuplexes outgoing packet stream Switches/routers play a key role in controlling packet flows, thus efficiently using network bandwidth and optimizing performance.

22 Network nodes Message Subscriber B Subscriber A Message Figure 7.13 Connectionless packet switching — originated from message switching message has header with source & destination address CRC check bit are used to detect errors Each switch check error, if yes, ask retransmission, if not find next hop. Message enter into a QUEUE to wait for line free to transmit Increased utilization of line is at the expense of queuing delay Loss of message may occur because of insufficient buffer. End-to-end error recovery is needed

23 t t t t Delay Source Destination T p Minimum Delay = 3p + 3T, ( remember it,will use it later) Switch 1 Switch 2 Figure 7.14 Delays in message switching Back p: propagation delay T: message transmission time

24 Packet switching VS. Message switching Why packet switching? 1. Long message has more chance to incur error than short packet 2. Long message is not suitable for interactive application because it causes very long waiting delay on other messages. Moreover the delay for one message is generally longer than the total delay of packets the message is divided into. Examples: L=1M=10 6 bits over two hops. Bit error rate: p=10 -6 the probability without error: P c =(1-p) L =( ) 10 6  e Lp =e -1  1/3. So on average, 1/(1/3)=3 tries for 1st hop. Similarly, 3 tries for 2 nd hop, Total six tries, 6M. Suppose divided into 10 5 bit packets, the no-error probability  0.9. So each packet needs 1/0.9=1.1 times on average. So entire message Get transmitted over each hop using 1.1M bits transmission. So total 2.2M bit transmission over two hops. Example: come soon.

25 Packet 2 Packet 1 Packet 2 Figure 7.15 Datagram packet switching Operations: 1. Address information is included in header 2. CRC for error recovery 3. switches inspect destination address in header to determine next hop 4. Packets are put in QUEUE to wait for line becoming available 5. Sharing lines among multiple packets, high utilization is at the expense of queue delay 6. Packets travel independently and may along different paths 7. Route may be detoured,thus bypassing failure and congestion 8. Packets may arrive out of order, resequencing may be required

26 t t t t p + 2(T/3) first bit received 3p + 3(T/3) first bit released 3p + 5 (T/3) last bit released Lp + (L-1)P first bit received Lp + LP first bit released Lp + LP + (k-1)P last bit released where T = k P 3 hops L hops p p + P Source Destination Switch 1 Switch 2 Figure 7.16 Delay in packet switching networks Assume three packets along the same path and transmitted in succession

27 Delay comparison of message switch vs. packet switch Support two switches and 3 packets /message –p: propagation delay, T: message transmission time, P: packet transmission time, and T=3P –Delay in message switching: 3p+3T see detailsee detail –Delay in packet switching: 3p+T+2T/3 Support L-1 switches and k packets/message: –T=kP –Delay in message switching: Lp+LT =Lp+LkP –Delay in packet switching (k packets): Lp+LP+(k-1)P Results: message switching involves additional delay of (L- 1)(k-1)P. Note, in the above, we neglect queuing and processing at each switch.

28 Destination address Next hop Figure 7.16 Routing table in connectionless packet switching Designing routing table is a key issue in packet-switching networks Which requires: knowledge of network topology and traffic levels. Moreover, size of table will become very large when network size increase.

29 Example –IP internetworks Objective of IP is to provide connectionless transfer service across heterogeneous networks Read it through textbook.

30 Virtual-circuit packet switching Establishment of connection between source & destination prior to the transfer of packets See figureSee figure Connect request is sent by source and connect confirm is replied by destination See figureSee figure Signal exchanges is performed in virtual-circuit setup See figure See figure All packets from a connection follow the same path. Admission control could be used to limit load on networks or specific links Routing is easier once connection was setup routing tablerouting table

31 Packet Figure 7.17 back Virtual-circuit switching Question: why called virtual-circuit? Because resources, e.g., switches, buffers and lines, are shared by packets from multiple connections, not dedicated to one specific connection.

32 t t t t Release Connect request CR Connect confirm CC Figure 7.19 Back Beginning latency and delay of packets in virtual-circuit packet switching

33 SW 1 SW 2 SW n Connect request Connect confirm Figure 7.20 … Back Signaling message exchanges in virtual-circuit setup Parameters such as buffer, bandwidth, delay requirements were set in every switch along the path during setup

34 Identifier Output port Next identifier Entry for packets with identifier 15 Figure 7.21 Example of virtual-circuit routing table for an input port Each connection is identified by VCI (virtual-circuit identifier). VCI may be different along the path for the same connection: the input VCI in a switch is different from output VCI for the same connection, called local VCIs. VCI is in the header of a packet, which is much smaller than IP address, as in IP protocol. Table lookup will find the output port # and output VCI based on input VCI, which is faster.

35 Pro and con about virtual-circuit switching Pro: – Header is shorter, resources allocated during setup, admission and congestion are easier. –Minimum delay reduces further. See figureSee figure Con: –Every router needs to maintain state information about all connections –Once there is a failure, all affected connections must be set up again. Example –ATM networks

Minimum Delay = 3p+T t t t t Source Destination Switch 1 Switch 2 Figure 7.22 Cut-through packet switching Forward a packet as soon as the header is received and table lookup is carries out. Assumptions: no error check, all lines are available. Back

37 t t t t p + 2(T/3) first bit received 3p + 3(T/3) first bit released 3p + 5 (T/3) last bit released Lp + (L-1)P first bit received Lp + LP first bit released Lp + LP + (k-1)P last bit released where T = k P 3 hops L hops p p + P Source Destination Switch 1 Switch 2 Figure 7.16 Delay in packet switching networks Assume three packets along the same path and transmitted in succession

38 Example –ATM networks Connection-oriented networks All information must be divided into fixed-length very small packets called cell (53 bytes). (Why?) The setup gives a chance for negotiating parameters between user requirement and network commitment, resources are allocated along the path The connection is defined by a chain of local identifiers VCIs. Assume low-error rate optical channel, so error control is done only end-to-end Is intended to support a large range of applications