An adapter is a board that typically contains RAM, DSP, a host bus interface, and a link interface (these are often called Network Interface Cards or NICs)

MAC layer addresses

It is not the node, but the node's interface card that has the address - three terms MAC layer, Physical layer, LAN address

Typically 6 bytes Unique on the planet - companies purchase blocks of 24 bit addresses ARP - address resolution Protocol (RFC 826) How does the sending node learn the MAC address of the receiver when it knows the IP address? Each node has an ARP table, which contains a mapping of IP addresses to MAC addresses. It also has an expiration date (TTL, shelf life). When the sender wants to send a packet, it first looks in its own ARP table. If not, it sends a special packet, an ARP packet, this has a number of fields Sender IP address
Sender MAC address
Receiver IP address
Receiver MAC address (initially set to broadcast FF-FF-FF-FF-FF-FF)

Request and response have the same format

Reply is send only to the sender.

ARP is plug and play. It does not require the sys admin to put data in a config table.

RARP - BOOTP - how does a diskless client learn its IP address. Similar to DHCP

Ethernet

Preamble     8 bytes 10101010 x7, then 10101011  This alows the 
   receiver to synchronize  Note that if the receiver is initially
   not synchronized and misses a byte, it hasn't lost anything.
   When it sees two 1s, it knows that the important stuff is coming.

Dest addr    6 bytes
Source Addr  6 bytes
Type         2 bytes (ethernet can multiplex several protocols,
             So this says that it is an IP packet
CRC          4 bytes
Data 46 to 1500

CSMA/CD (Carrier sense multiple access with collision detection)

The padding is necessary to prevent a station from completing the transmission before the first bit has reached the far end, where it may collide with another frame

Collision detection. Each station detects if there is more power on the cable than they are transmitting, and if they are they send a noise burst.

Round trip time on modern ethernet is about 50 usec in the worst case (including repeaters) so a frame has to take at least that long. At 10 Mbps a bit takes 100 nsec, so 500 bits is the smallest frame that is guaranteed to work. This is rounded up to 512 bits (or 64 bytes) which explains the padding.

Performance is very good under low congestion, but collapses with heavy congestion.

Ethernet Technologies

Interconnections: Hubs and Switches

But usually all are in a common collision domain.

Hubs are physical layer devices. They know about volts but not headers. A switch operates on ethernet frames and thus are layer 2. A full-fledged packet switch forwards frames based on LAN addresses.

The switch needs a switching table, which is a mapping of MAC addresses and interfaces.

Self-Learning (Plug and Play)

1. The switch table is initially empty
2. When a frame arrives with a dest addr not in the table, the switch forwards copies to all of the interfaces.
3. For each incoming frame, the switch stores in its table The MAC address of its source field.
   The Interface from which it arrived.
   The current time
   
4. Items are deleted from the table if no frames are received from that source after a period of time (the aging time). Thus if a user replaces his adapter card with a new one, the old card is eventually dropped.

Here is a link to a good description of VLANs.

Token Ring

Performance is somewhat worse than ethernet if there is little congestion, but in times of high congestion, ethernet collapses completely while token ring continues to work, although with decayed performance

PPP the point to point protocol

PPP data framing
A one byte flag 01111110
Address field - ALWAYS 11111111
Control field - always 00000011
Protocol - IP is 21, Appletalk is 29, DECnet is 27
The data
The checksum