Link Layer Protocols

• e.g. have single link and two senders wanting to sent frame a and b

• => transmit single modified packet combine(a,b)

• => receivers can reconstruct individual packets…

• different type of routing

• -> nodes in network combine packets possibly from different sources and generate groups of encoded packets

• => network coding allows to acheive maximum possible information flow in a network

• not really

• e.g. retransmission schemes exist on different layers

• => basically was designed by a commitee once…

• benefits:

  • need layers to manage complexity

  • common functionality

• disadvangages:

  • layer N may duplicate lower layer functinlaity

  • different layers may need same information

  • layer N may need to peak into layer N+x

• transferring packet from one node to adjacent node over a link

• framing, link access

• reliable delivery between adjacent nodes

• flow control

• error detectoin

• error correctoin

• half-duplex and full-duplex

• encapsulate packet into frame -> adding header and trailer…

• channel access if shared medium (CSMA)

• MAC address used to identify src and dst node

• low bit error links -> rarely used

• but for high loss links e.g. wifi, L2 retransmission scheme in place

• point to point

• broadcast (shared wire or medium)

• distributed algo that determines how nodes share channel (when to transmit)

• communicatoin about channel sharing uses channel itself (no out-of-band)

• channel partitioning

  • divide channel in pieces (e.g. time slots)

  • allocate pieces to node for exclusive use

• random access

  • channel not divided -> allow collisions

  • mechanism to recover from colisions in place

    • ALOHA, CSMA…

• taking turns

  • e.g. token ring

  • nodes take turns

  • nodes with more to send can take longer turns

  • either polling from central site or sth. like token passing used…

• 6B dest MAC

• 6B Src MAC

• 2B Type (e.g. IP, ARP,…)

• 46-1500B L3-PDU

• 4B FCD (CRC-32)

• 4bit Version

• 4bit IHL (header length)

• 1B Type of service

• 2B total length (multiple of 32 bit)

• 2B identification

• 3bit flags (reserved, DF, MF)

• 13 bit fragment offset

• 1B TTL

• 1B Protocol

• 2B Header checksum

• 4B Src Address

• 4B Dst Address

• Options/Padding (optional)

• Base2

  • thin ethernet

  • 10Mbit/s

  • ma 185m

• Base5

  • thicknet

  • 500m

• scalability

  • flat addresses

  • no hop count

  • missing additional protocols (such as ICMP)

  • perhaps missing

    • fragmentation

    • error messages

    • congestion feedback

• human friendly notation (separate by : or -)

• multicast and broadcast addresses

• OUI

• Locally administered MAC adres (similar to private IP / e.g. VM)

• flat address portability (can easily ove one LAN card somwhere else…) -> implication on privacy?

  
  

• 2nd least significant bit of first byte has value 0

• least significatn bit of first byte has value1

• coresponds to convention of transmitting least significant bit of each byte first in serial data communicatoins

• -> first bit of each byte maps to least significant bit in memory…

• physical (unicast) -> first bit 0

• logical (multicast/broadcast) ->first bit 1

=> can be used to configure NICs to capture frames for unicast or one of a set of multicast addresses

• seletive forwarding plug and play

• -> not forward to incoming interface…

• prevent loops

• => bridges gossip among themselves and compute loop-free subset of network

• => forward data on spanning tre while other links are backup

1. select root bridge (with lowest bridge ID)

2. determine least cost path to root brige

   1. each bridge determines leas cost path to root

   2. each bridge chooses least cost path

   3. port connecting to the path becomes root port (multiple root ports if there are multiple paths…)

   4. bridges onnetwork segment determine bridge port with least cost path to root -> designated port (best root port)

3. disable all other root paths

4. => regularily send Bridge Protocol Data Units (default 2s) to STP multicast address

• configuratoin BPDUs transmit bridge_ids and root path costs

• topology change notification (in case change in network topology…)

• topology change notificatino acknowledgemetn

• blocking

• listening

• learning

• forwarding

• disabled

bridges = switches…

• no, uses most significant bit first

• record link associated with senditn host (source link)

• index switch table using MAC destionatino address (lookup MAC dst)

• if entry found:

  • from same link the frame came from -> drop

  • else: forward to interface indicated by switchgin table

• if no entry found:

  • flood (to all interfaces except the one we received the frame from)