reas for Discussion Direct Link Networks Joseph Spring School of Computer Science Sc - Computer Network Protocols & rch s ased on Chapter 2, Peterson & Davie, Computer Networks: Systems pproach, 5 th Ed Ethernet (Section 2.6) Introduction Physical properties ccess Protocol Frame format ddresses Transmitter algorithm Lecture - Ethernet 1 Lecture - Ethernet 2 Ethernet - Introduction Developed in mid 70 s by Xerox PRC (Palo lto Research Center) Introduction Original 10 Mbps standard defined in 1978 by Dec, Intel and Xerox This later formed the basis of IEEE standard 802.3 Lecture - Ethernet 3 Lecture - Ethernet 4 Ethernet - Introduction CSM/CD Carrier Sense Multiple ccess with Multiple ccess set of nodes share a single physical link Carrier Sense ll nodes can distinguish between an idle and a busy link Each node listens as it transmits and can therefore detect when the frame it transmits has been corrupted Lecture - Ethernet 5 Physical Properties Lecture - Ethernet 6 Lecture - Ethernet 1
Ethernet Physical Properties Transmission Rate 10 Mbps 50 Ohm Coaxial cable Maximum length 500 metres Host Connections Taps at least 2.5 metres apart Terminators Placed at end of Ethernet cable to match line impedance absorb signals and minimise reflections Lecture - Ethernet 7 Ethernet Physical Properties Repeaters Maximum of four repeaters between nodes so maximum reach of 2500 metres Signal roadcast over entire network it Representation Manchester Encoding Number of Hosts Up to 1024 Hosts Note: There are both cheaper and faster versions of Ethernet Lecture - Ethernet 8 Ethernet Transceiver and daptor Ethernet Repeaters (Fig 2.27) Transceiver daptor Ethernet Cable (Fig 2.28) Host Host Repeater Lecture - Ethernet 9 Lecture - Ethernet 10 lternative Ethernet Technologies 1. Original 10 Mbps Ethernet developed by Xerox et al. 2. IEEE 802.3: standardisation of above (10ase5) Ethernet Hubs 10aseT (twisted pair Ethernet) Often employs multiple point-to-point connections in a multi-way repeater known as a hub 3. 10ase2: based on thinner coaxial cable 4. 10aseT: based on twisted pair wire Hub Hub 5. Fast Ethernet: 100Mbps 6. Gigabit Ethernet: 1000Mbps Lecture - Ethernet 11 Hubs are also used with 100Mbps Ethernet Note: frames from one host always reach all other hosts on the Ethernet, irrespective of the configuration Lecture - Ethernet 12 Lecture - Ethernet 2
Ethernet - Frame Format (Fig 3.21: original Dec/Intel/Xerox format pre IEEE format) ccess Protocol 64 48 48 16 32 Dest Src ody 8 Postamble (Fig 2.30: IEEE 802.3 format Ethernet format p115, 3rd Ed.) 64 48 48 16 32 Dest Src ody field now used to define length of frame Lecture - Ethernet 13 Lecture - Ethernet 14 Ethernet Frame Format 64 bits for synchronisation; alternating ones and zeros Serves as demultiplexing key Identifies higher-level protocol associated with the frame Data Minimum of 46 bytes to allow collision detection Maximum of 1500 bytes to prevent hogging 32 bit Postamble 8 bits to detect end of frame Ethernet Frame Format 2 Frame Formats The second Ethernet frame format shown in 2 nd, 3 rd and 4 th edition of the text Earlier texts imply that first Ethernet frame is the pre IEEE format Difference Only difference is that the type field is used to define the length of the frame Hence postamble is discarded Note: Since original type field values always exceed 1500, the maximum length of an Ethernet frame, both frame formats can coexist in a single Ethernet system Lecture - Ethernet 15 Lecture - Ethernet 16 Ethernet ddresses Source ddress 48 bit ess Destination ddress 48bit ess Every Ethernet node world-wide has a unique ess Each manufacturer is allocated a unique block of esses Ethernet ddresses n Ethernet daptor receives all frames and accepts: Frames essed to its own (unique) ess Frames essed to the broadcast ess (all 1 s) Frames essed to certain multicast esses ddresses that start with one Could therefore use a specific multicast ess to send messages to all file servers ll frames In promiscuous mode Lecture - Ethernet 17 Lecture - Ethernet 18 Lecture - Ethernet 3
Transmission lgorithm Sending a frame: If line idle daptor transmits the frame immediately Since no arbitration is required, access time is excellent Upper data limit of 1500 bytes avoids hogging If line busy daptor waits until the line is idle and then transmits immediately good way to encourage collisions!!! Transmission lgorithm If a collision is detected: (Own message not read back successfully) 1. daptor transmits a 32-bit jamming sequence and dthen stops t least 96 bits are therefore transmitted, a 64 bit preamble plus the jamming sequence This ensures that all other units also detect a collision Lecture - Ethernet 19 Lecture - Ethernet 20 Transmission lgorithm If a collision is detected: (Own message not read back successfully) 2. Wait a random time and try again Initially the adaptor waits between 0 and 51.2 ms 3. The waiting interval used doubles after each successive failure called exponential backoff 4. Give up eventually, typically after 16 tries, and report an error Lecture - Ethernet 21 Ethernet adaptors must always transmit at least 512 bits of each frame Every Ethernet frame must therefore be at least this long 14 bytes header 46 bytes of data 4 bytes of 512 bits takes 51.2 msec to transmit Enough time to get from one end of a maximum size Ethernet to the other and back (see Fig 2.31) Lecture - Ethernet 22 (Fig 2.31 p123) Suppose daptor starts to send a frame from one end Just before the frame reaches, also starts to transmit If doesn t transmit for one RTT the signal from will fail to reach before completes transmission will therefore fail to detect the collision Lecture - Ethernet 23 Lecture - Ethernet 24 Lecture - Ethernet 4
(Fig 2.31 p123) Lecture - Ethernet 25 Lecture - Ethernet 26 Works very well surprisingly well under lightly loaded conditions However, utilisation over 30% is considered heavy There are then too many collisions for Ethernet to work effectively Note: Ethernets do not saturate gracefully eyond a certain point the more frames the nodes try to send, the lower the effective bandwidth Lecture - Ethernet 27 Hosts therefore typically provide end-to-end flow control mechanisms to avoid excessive use of the Ethernet by one node See pages 123-124 for more details Lecture - Ethernet 28 Summary Ethernet (Section 2.6) Introduction Physical properties ccess Protocol Frame format ddresses Transmitter algorithm Lecture - Ethernet 29 Lecture - Ethernet 5