1. Fall 2005 EE 543 Packet Switched Networks Fall 2005

2. Circuit switching Route through the network is reserved end-to-end for exclusive use over duration of call Call set up: establish route, reserve links, set switches Call tear down: release resources Where is the intelligence? Network mgmt and control Data base switch

3. Fall 2005 Packet switching Packets have source and destination addresses in headers Packets stored in buffer at each node Node reads headers routes packet to next node How do routers know where to send the packet? Routing table Routing table Routing table Routing table router

4. Fall 2005 Switching: circuit versus packet Circuit Well suited for continuous flows (CBR) “easy” to implement Network has “state” Utilization can be low Multiplex traffic to increase efficiency Packet Well suited for bursty flows – datagrams Storage needed at each node No “state” – each node functions independently High utilization Queuing delays Packet loss due to congestion

5. Fall 2005 Packet – circuit hybrid: virtual circuits ATM switch Path through the network is the same for all packets in a connection Capacity along a virtual circuit is not reserved for a connection

6. Fall 2005 Delays in networks Transmission: (packet size)/(bit rate) –How long does it take to get the packet on to the link? Propagation: (path length)/(signal speed) –How long does it take to get from the sending node to the receiving node? Queuing (buffers) –How long does the packet sit in a buffer? Processing –How long does it take the switch/router to move packet from input buffer to output buffer?

7. Fall 2005 Delays in networks: examples Transmission: (packet size)/(bit rate) –10kbit packet, 10Mb/s link speed Propagation: (path length)/(signal speed) 100km path, fiber link (3.3  sec/km) Queuing (buffers) –Typically controlled to 4x T trans Processing delay –Typically negligible

8. Fall 2005 Delay-bandwidth product How many bits are enroute from source to destination? –Prop delay * bandwidth delay bandwidth Bits in Network as a “pipe”

9. Fall 2005 Delay-bandwidth product Example- long distance, high speed network: –Distance = 5000 km, delay = 5000*5*10 -6 = 25 msec –Bit rate = 10 gbs = 10 10 bps –Bw*delay = 10 10 * 25 * 10 –3 = 25 * 10 7 = 250 MILLION BITS! delay bandwidth Bits in Network as a “pipe” Very large buffers needed at source and destination!

10. Fall 2005 Delay-bandwidth product: buffer management The prevalent delay-bandwidth product rule of thumb is that a router port should have a buffer capacity B equal to the average Round Trip Time (RTT) of the TCP sessions flowing through the link times the link bandwidth (BWlink in bps). B = RTT x BWlink Since the propagation delays for transcontinental links and transoceanic links are large, routers and switches with high bandwidth WAN interfaces need large buffers. For example, according to the delay-bandwidth rule of thumb, a switch/router with 10 GbE WAN interface linking two research campuses 10,000 miles apart would require approximately 200 ms x 10 Gbps = 250 MB of buffering available to that interface. Ref: http://www.force10networks.com/applications/buffermgmt.asp?content=1

11. Fall 2005 Error, flow and congestion control Interrelated functions: –Depend on delay-bandwidth product Several alternatives for error control: –Detect errors –Detect and correct errors “Best” approach depends on application requirements and available bandwidth –Detect/retransmit: fine when latency is not a concern –Detect/correct: Better where timing is critical (but adds overhead to bandwidth)

12. Fall 2005 Several retransmission protocols available to correct errors Alternating bit protocol (ABP): –Receiver sends ack for each correct packet –Sender waits for ack before sending next packet –Sender resends packet if ack not received before a time out Time out interval Corrupted ack Corrupted send

13. Fall 2005 Error control and flow control protocols Some special cases: –Satellite communications links: High bandwidth (100 Mbps) long delays (250 msec) –Deep space communications links: Low bandwidth (1-10 kbps) Very long delays (days, weeks, months…) –High loss paths: Non-line of sight radio (error rate high and variable) Meteor propagation (low bandwidth, high error rates)

14. Fall 2005 Principles of layering Layer should be created where a different level of abstraction is needed Each layer should perform a well-defined function Layer boundaries should be chosen to minimize information flow across interfaces Each layer provides a service to the layer above, receives a service from the layer below Separation of concerns

15. Fall 2005 Internet Architecture Defined by Internet Engineering Task Force (IETF) Hourglass Design Application vs Application Protocol (FTP, HTTP) ■ ■ ■ FTP TCP UDP IP NET 1 2 n HTTPNVTFTP

16. Fall 2005 ISO Architecture

17. Fall 2005 Layering: the OSI model compared to the Internet application presentation session transport network link physical OSI = open system interconnect Internet Applications TCP Connection Oriented UDP Connection- less IP datagrams Network bitways OSI precedes commercial Internet OSI

18. Fall 2005 Protocols Building blocks of a network architecture Each protocol object has two different interfaces –service interface: operations on this protocol –peer-to-peer interface: messages exchanged with peer Term “protocol” is overloaded –specification of peer-to-peer interface –module that implements this interface

19. Fall 2005 Protocol interfaces

20. Fall 2005 Protocol machinery Protocol Graph –most peer-to-peer communication is indirect –peer-to-peer is direct only at hardware level Host 1Host 2 File application Digita l library application Video application File application Digita l library application Video application

21. Fall 2005 Protocol machinery (cont) Multiplexing and Demultiplexing (demux key) Encapsulation (header/body) Host Application program Application program RRP Data HHP RRP HHP Application program Application program RRP Data HHP RRP Data

22. Fall 2005 OSI: Physical Layer “Bitways” in the Internet nomenclature Hardware on which the network is built- carries the signals Fiber, copper, radio, etc. No assumptions of reliable delivery of bits Physical layer standards – specify Modulation scheme Characteristics of interfaces between xmtr, rcvr and medium Example protocol: 802.3/10BASEx (x=10, 100, 100, etc)

23. Fall 2005 OSI: Physical Layer Phy layer: converts bits into electrical or optical signals May add synch bits to synchronize rcvr with xmtr.

24. Fall 2005 OSI: Data link layer Transmitter and receiver execute a protocol to retransmit corrupted packets Transmitter adds error detection bits (and may add sequence numbers) to packet Issue: end-to-end error control versus link-layer error control Link-layer control adds overhead May be unnecessary if links are reliable Link-layer control preferable for high-loss links: Wireless! Satellite (long delays) Example protocol: Ethernet frames, ATM

25. Fall 2005 OSI: Data link layer Data link layer: Adds sequence number Adds error detection bits (CRC)

26. Fall 2005 OSI: Data Link Layer Sometimes divided into two sub-layers SAR: segmentation and reassembly (or LLC) MAC: media access control MAC SAR Segments data into smaller blocks, Adds tags for error detection, sequencing, etc. Controls media sharing- multiplexes Data from several sources onto common link

27. Fall 2005 Media access control (MAC) sublayer Required for a shared link: MAC address – designates destination for packet MAC protocol: rules for sharing the link resources

28. Fall 2005 Logical link control (LLC) sublayer LLC: Implements error detection/sequencing for shared link Same as data link protocol for point-to-point link LLC + MAC = data link layer

29. Fall 2005 OSI: Network layer Performs delivery of packets between source and destination Source-destination path may consist of multiple links Routing Network layer adds source and destination addresses Example protocol: IP

30. Fall 2005 OSI: Transport layer Delivers messages between ports in different computers Port numbers used to differentiate between different processes running on the computers Transport layer can be either connection oriented or connectionless connection oriented: delvers error free messages in the correct order Connectionless: Delivers messages one-by-one, no guarantee on order or correctness Breaks messages into packets for transmission by network layer Reassembles packets from network layer into messages Example protocols: TCP, UDP

31. Fall 2005 OSI: Session Layer Concern: supervision of dialog between end computers Sets up connections prior to exchange of user information Controls dialog during an information exchange May assist in synchronization of flow of large messages, insert checkpoints Example: SS7 (signaling system 7), MPLS

32. Fall 2005 OSI: Presentation Layer Concern: Syntax and semantics of information transmitted Encoding of data into standard application formats (e.g, ASCII) May include functions such as data compression, encryption Examples: ASCII, DPCM speech coding, MPEG

33. Fall 2005 OSI: Application Layer Concern: Provide commonly used functions for users Examples: file transfer protocols, terminal emulation, remote login, directory services Examples: FTP, Telnet User applications run on top of this layer: e-mail, www, etc.

34. Fall 2005 OSI: Summary of layers 4-7 functions

35. Fall 2005 Summary of services provided by layers in the OSI model

36. Fall 2005 The end-to-end principle: Minimize the functions at the lower layers; do as much as possible On an end-to-end basis application presentation session transport network link physical

37. Fall 2005 The end-to-end principle: Examples of functions best done end-to- end: –Encryption –Error detection/correction –Packet ordering –Deletion of duplicate packets –……………