The Basics

• Open System Interconnect

• collection of conceptially similar functions

• layers offer services to layers above

• and uses services offered by layer below

• no it only specifis which services a layer might offer

• no, only functions of layers in an abstract way

• Thorugh layering, the communication process is divided an conquered

• -> smaller and widely independently portions

• through the more or less independent layers, the implementaiton of one should theoretically be able to change without affecting other layers…

• sender and receiver exchange Protocol Data Units (PDU) at the same layer (e.g. TCP packet)

• communication between layers in a single system -> e.g. Transport with network layer

• higher layers provide PDU -> which is the SDU of lower layer (payload)

• lower layers strip PDU and transmit SDU to higher layer…

protocol specific information

-> how to proceed with the SDU…

• no, can also be used for e.g. conceptualizing communication in single system like USB periphery

• no, sometimes layers may act very closely, -> merging through cross layer approaches

• not neccesarry

• interface between running application on a node and protocols offered by OSI model

• HTTP

• -> bworser itself not part of the OSI model

• -> but HTTP as interface to the OSI transport stuff…

• others: SMTP, telnet

• not always implemented as dedicated protocols

• presentation: formatting the data such that it is readable at the device at receiving side (e.g. JSON.…; -> serialization)

• presentaiton might also provide encryption and compression servicves

• session layer: establish and maintain logical session s

• -> controls mode of operation e.g. simplex, duplex, …

• => e.g. used for RPC…

• end-to-end communicatoin between applications -> using port numbers…

  
  

• 0-1023

65535

• TCP (slow reliable)

• UDP (fast unreliable)

• ensures that data at receiver can be fully reassembled

• offer mechanisms to detect and revocer from transmisison errors

• offers flow control and congestion control

• flow control: receiver is not overwhelmed by sender

• congestion control: data sent as fast as possibnle without exceeding network capacity

• none -> just sents it and hopes it arrives…

• => but minimal overhead -> minimal delay

• e.g. video streaming…

• segment

• datagramm

• to allow addresation of nodes in a wide-spread network

• enable routers to decide the next hop

• detect and report errors

• IP protocol

• IPv6 protocol

• ICMP

• ICMPv6

• interwork packet exchange (IPX)

• datagram delivery protocol (DDP)

• address the sender and receiver of the current hop

• network -> address the general sender and receiver…

• frame

• network layer -> hierarchical

• data link -> flat

• control access to the physical medium used for data transmission (CSMA/CD)

• interaction between network hardware and physical medium

• how are bits represented?

  • coding, modulation, voltage levels

• pin layout of connectors

• characteristics of cables

• …

• transmission technology for LAN

• first ethernet standard

• following standard

• thick vs thinner coaxial cables (thicknet vs thinnet)

• 10 Mbit/s

• half duplex since single medium shared between devices

• => whole medium is single collision domain / segment

• portion of network where collision can occur when two stations send at the same time

• -> corrupts data

• control when station is allowed to send

• e.g. CSMA/CD

• since the medium broadcasts messages

• -> specify for whom the message is… with medium access control

• 48bit

• 100m due to signal athenuation (wird schwächer…)

• repeater

• fiber optic cables

• CSMA/CD -> at some point does not work correctly anymore…

• 1990 standard T

  • twisted pair cable (4 copper cables)

  • one pair for receiving (RX) and one for transmitting (TX)

• solely difference -> 100Base-TX 100Mbit/s

• two statoins connected directly

• RX and TX exchanged

• -> RX on one is TX on other and vice versa

• -> allows for full duplex…

• star

• only macro level

• -> funcitonality same

• -> straight through (no cross cable) to central hub

• hub connects TX of each to RX of all others

• -> same as bus as transmitted frames visible for all nodes…

• on respective ethernet standard

• 100Base-T4

• Gigabit Ethernet over twisted pair cables

• 100Base-TX (and 1000Base-TX)

• Line Code

• group of 4 bits (nibble) mapped to 5 bits

• e.g. ensure that at leas one 1 is present (0000 -> 1110…)

• with MLT-3 guarantees change of voltage at least every 5 symbols

• -> important for synchrnization (detect stard and end of symbol intervals)

• -> allow to define command characters which do not represent data words but used for signaling (16 singaling characters possible…)

• overhead -> 1.25 symbols per bit needed

• transmitted electrical 0 or 1

• Line Code

• multi level transmit

• -> reduce frequency

• cycle through -1,0,+1

• change when 1, stay when 0

• 125MHz

• 31.25MHz

-> the one takes 1 0 for a cycle

-> MLT-3 takes 1 0 -1 0

• no

• 250MBit/s

• use all four pairs of cables to send or receive -> 4x 250MBit/s -> 1GBit/s

• does not

• -> sender knows what he sent

• -> able to reconstruct the received stuff…

• carries sense multiple access / collisoin detection

• -> control how nodes access shared medium and how collision are detected and handeled

• each node listens to medium if there is already data sent

• if medium idle -> start sending

• assumed to be successful (only assumed) if no errors (i.e. no collision)

• when collision -> stop current transmission and send jam signal

• involved stations notified about collision

• exponential back-off

• each station choose random element from {0,1,2,4,8,…,2^i-1} wher i is number of successive collisions

• i limited to max value of 10

• host waits x time slots before allowed to send data again

• 100MBit/s 5.12 microseconds

• 10MBit/s 51.2 microseconds

• wait till its free…

• => basically exp backoff -> timeout before sense the carrier again…

• singals propagate with around 2/3 of light in vacuum

• -> 100 meter segment requires 0.5 micro seconds for propagation

• -> if longer -> node might have sent all and think it is no collision although node further away started to sent at the same time…

• collision avoidance

• -> has retransmission functionlaity

• -> especially good for unreliable links -> e.g. wifi

• MAC header

• compares destination to its own

• organizational unique identifier

• -> first 24 bit of MAC that identify vendor/manufacturer of network hardware

• assigned randomly

• -> prevent duplicate MAC from same vendor…

• => errors and inability on same networks…

• (preable (15 times 1010))

• (SFD 1011)

• Dst MAC 6 byte

• Src MAC 6 byte

• Type 2 byte

• Payload 46-1500 byte

• FCS 4 byte

• inter frame gap 9.6 microseconds

• stricly speaking not part of ethernet frame

• -> indicate start of frame and ensure spacing between subsequent frames (inter frame ga=

• e.g. IPv4 0x0800 ro IPv6 0x86DD

• frame check sequence

• -> constitutes a trailer

• contains CRC-32 checksum

• over whole ether frame (incl src and dst mac)

• discarded

• 64 byte (46 byte payload plus 18 byte header)

• otherwise cannot be guaranteed that sendng host recognizes collision in time

• -> serialization too short…

• 1518 byte

• yes, because it is backward compatible

• -> but slot size increases (carrier event -> in case frame only 64 bit -> padding to 512 bytes)

• -> called carrier extension

• 10Base-2 / 10Base-5

• all other standards if Hubs are used

• simply receive and amplify signals and forward to all ports except the one the frame was received from

• has switching table to decide to what port forward the packet to …

• (selective compared to hub)

1. receive packet

2. look at MAC

3. look up MAC in switching table

4. if MAC in switching table

   1. forward to associated port

5. if not

   1. broadcast

• full duplex communication

• yes if A and B send to C in full speed

• -> might buffer one of the frames…

• if buffer full -> drop further frames…

• => no notification

• higher level protocols

• -> flow control and congestion contro….

• yes -> buffers message until medium is idle…

• describes a part of a network where a collision can occur

• each node forms its own collison domain

• -> two nodes can communicate in full duplex (either RX TX pair or sophisticated modulation scheme)

• include all nodes that can be reached with broadcast address

• -> area where layer 2 broadcast is visible

• MAC broadcast ff:ff:ff:ff:ff:ff

• -> routers break broadcast domains

• if hubs are used

• allow connecting hosts of different speeds

• -> negotiate speed on per-host basis

• translate them

  
  

• tranlate e.g. WLAN to Ether

no

• flat hierarchy

• -> no hierarchical structure requiring us to take all existing nodes into consideration…

• hierarchical

• -> split up network in smaller chunks through subnetting…

• => withih… -> mac uniquely identify

-> MAC hop to hop

-> IP end to end

• 32 bit

  
  

• yes -> n:1 mapping of IP:MAC

• network part

  • identifies network

• host part

  • identifies host in network

• 192.168.1.1/16

• -> 16 bits from left are network, 32-16 are host

• first: reserved network address -> identifies network itself

• last: broadcast address

• usually not -> prevent flooding

• Version (0x4 0x6)

• Header length

• DSCP

• ECN

• Total length

• Identification

• Flage

• Fragment Offset

• TTL

• Protocol

• Header Checksum

• Source and Destinatino IP

• 0x4; 0x6

• header length = total length / 32 bit

• differentiated service code point

• first 6 bit -> codepoint

  • set priority

• last 2 bit ->

=> QoS

• explicit congestion notification

• -> can be set by intermediate routers to indicate congestion

• => rarely used as TCP implements congestion control

• 16 bit value -> important when packet fragmented, as same ID means fragments belong together

• frst bit reserved -> 0

• second bit -> dont fragment (0:n, 1:y)

• third bit: packet is a fragment and more fragments follow

points the position of the packet in the fragmented whole

-> given in multiples of eight bytes

• each router forward decreases the TTL by one

• -> TTL = 0 drop packet

• => avoids packet-forwardin loops in case of routing loops

• 16 bit checksum (not CRC) protecting only header

• IPv4 address space exhausted

IPv6: 128 Bit

IPv4: 32 Bit

• eight groups

• of four hex digits

• seperated by colons

• leav out leading zeroes in groups

• concatenate longest group of 0000:0000:0000:…

• -> to ::

• !! but only one and must be longest!

• fe80::

• -> assigned to every network interface that has IPv6 enabled

• => can only be used for communicatoin in LAN -> routers must not forward if link lokal is source or dest

• => only valid in subnet the host is attached to

• ff00::/8

• SLAAC -> stateless address autoconfiguration

• Version 4 bit

• Traffic Class 8 bit

• Flow Label 20 bit

• Payload Length 16 bit (includes extension header -> seen as payload)

• Next Header 8 bit selector identifying header -> same values as IPv4 protocol field

• Hop Limit 8 bit TTL equivalent

• Source Address

• Destination Address

• (extension header)

• located between IP and upper layer headers

• -> can have 0, 1 or more extension headers

• => next header chain…

• Address Resolution Protocol

• -> host: i have IP what is the corresponding MAC?

• => usually only IP is known…

• no!!

• => offer service for translation between physical and logical address

• request

• replies

• MAC header with broadcast destination

• ========

• ARP header -> type ethernet (0x001)

• -> protocol 0x0800 IP

• Hardware Size, Proto Size,

• Opcode 0x0001 requestL 0x0002 reply

• ==========

• Sender MAC

• Sender IP

• Target MAC (00:00:00:00:00:00)

• Target IP

• Dest MAC -> sender from request

• Src MAC -> own MAC

• Type 0x0806

• ======

• Type ethernet, proto, hwsize, protosize

• opcode reply 0x0002

• ======

• sender MAC

• sender IP

• target MAC

• target IP

• -> sender own target receiver (the one who did the request)

• 
  

• compares IP contained with own IP

• -> no match => ignore

• -> match => reply

• yes -> only works on layer 2; offers service to layer 3

• Neighbor Discovery Protocol

• -> IPv6 equivalent to ARP

• yes!

• -> check wether SLAAC assigned link local addresses are unique

• find available neighbors or routers

• router solicitation (type 133)

• router advertisement (type 134)

• neighbor solicitation (type 135)

• neighbor advertisement (type 136)

• redirect (type 137)

• inernet control message protocol

• -> used for error signaling and management

• ICMP for IPv6

• ICMP is encapsulated into IP header

• ping -> echo reply

• traceroute (uses TTL incrementing by 1 until target reached -> node that discards sends ICMP time exceeded…)

• 8 bit type

• 8 bit code

• 16 bit checksum

• 32 bit data (optional)

type 0 echo reply:

1. echo reply

type 3 destination unreachable:

1. destination network unreachable

2. destination host unreachable

3. destination protocol unreachable

4. destination port unreachable

5. fragmentation required and DF flag set

1. Destination network unknown

2. destination host unknown

type 5 redirect message:

1. redirect datagram from the network

2. redirect datafram from the host

type 8 echo request(ping):

1. echo request

type 11 timeexceeded:

1. TTL expired in transit

2. fragment reassembly time exceeded

   
   

   
   

   
   

   
   

   
   

   
   

   
   

• TCP

• UDP

• reliable data transfer at cost of overhead and complexity

• minimalistic, low as possible overhead

• unreliable, no sequencing (order of arrival), no flow and congestion control

  
  

• connectionless, packet oriented service

• -> purpose is uniqely identify sender and reveicer of messages

• no acknowledgement of

• arrival

• receipt

• loss

• establishes logical connection

• between sender and receiver

• befor data transfer starts

• => connection oriented vs the connectionless in IP

• three way handshake (syn, syn ack; Ack)

• incrementing seq number for sender and receiver (different one)

• -> acknowledges always the next one -> kind of okay you can send seq xy…

• missing sequence number…

• receiver window

• -> satates max number of bytes sender is allowed to transmit before receiving an ack

• sliding windows

  
  

• IP-5 Tuple

• e.g. wb server has to be able to differentiate differnet nodes and also concurrent connections from same node

• physical representation of a bit or groups of bits

• baud rate times number of bits per symbol which carry informatino (e.g. 4b5b -> 4)

• baud rate: number of signal units transferred per second

• bit rate: number of information carrying bits transferred per second

• minimum frame size * serialization time (bit rate…) (in 100mbits 5.12 micro seconds)

• thus, RRT must be smaller than 5.12 micro seconds

• in 2/3 of speed of ligth propagatoni

• -> 512 meters…

• minimum frame size still the same

• -> but slot size is increased to 512 byte (padded with extesion symbols… -> symbols that cannot occur in the payload)

• => called carrier extension

• handle collisions

• when divice whats to send data -> checks if so else is sending alreead

• -> if so : waiting

• when collision occurs -> waits for a time determined by random choice from backoff interval and then tries again

• detec tthat received frame is (probably) correct

• yes

• -> only repeat incoming signals…

• reads dest MAC address and performs lookup to determine outgoing port

• does not touch source and dest MAC because fully transparent to nodes

• better support form mobile nodes with the binding update

• supports resource reservation (QoS) based on flow labels and traffic class

• standard header got shortened regarding the amount of different fields (faster processin in routers possible)

• new address calss: anycast (to one member of a group)

• NDP for configuration issues

• suppport for authentication and data integrity

  
  

• no -> only the destination optinos header might occur twice

• each header has next header field that indicates, wether a header is following…

• as the ARP cache still holds the old MAC for the respecitve IP

• -> tries to reach old MAC…

• ff02::1::ff:24 bit interface ID -> which is a solicited node multicast address…

• request / response

• NS -> What is the MAC of 2001:DB8::2?

• NA -> MAC of 2001:DB8::2 is MAC xy…