ATM (Asynchroneous Transfer Mode)

• Basic unit of an ATM network consists of a high speed electronic switch. This dedicated copmmunication computer can serve 16-32 host computers via a user to network interface UNI (a pair of fiber optics cables - one in each direction - includes hardware to sense light pulses and a minilaser to send the same).
• ATM switches are connected via UNI or NNI (designed to connect networks with different operators)
• Connection oriented interface
  • Switched virtual circuit: For each frame, a path from the source to the destination is established by
    • Each switch on the path commintting the necessary resources and sending the request to the next switch along the path
    • Each participating switch identifies the path so that the frames have no addresses
    • Finally, the destination node accpts the path by sending a signal to the source
  • Permanent virtual circuit: Configured by network administrator for special purposes (special service types, debugging, maintenance)

• Frames (cells)	48 octets of data
  	5 octets control

  a circuit identifier consists of
  8 bits of virtual path ID
  16 bits of virtual circuit ID
  for traffic control and accounting.

  Flow control	1/2 octet
  Circuit ID	3 octets
  Payload type/prio	1/2 octet
  CRC	1 octet

  followed by data.

• ATM adaptation layer if the network interface interacts with the application software. Source and destination hosts must agree on the adaption layer protocol used for the entire transmission.
  • AAL1 For audio and video data requiring a fixed rate service.
  • AAL5 To send conventional data in packets

    	1 to (2^16)-1 octets long	followed by
    an	8 octet long trailer	that includes
    	the Length 16 bit field Checksum 32 bits

    This packet is	divided into cells
    	transmitted through the network
    and	recombined at the destination
    before being	delivered

X25Net

• Runs internet protocols over public data networks (instead of networks of expensive leased lines)
• PDNs operate under a telecommunication standard X.25 protocol which treats TCP/IP packets strictly as data
• The tunneling technique of using an underlying protocol as integral part of the network requires the end_points of a TCP/IP transmission to establish the protocol beforehand
• X.25 provides virtual circuits for the duration of a transmission; X25net can (and does) send packets of a message simultaneously through several connection
• Addressing on X.25 similar to telephone numbers: area code based on location
• Gateways between X.25 (any PDN) and ARPAnet: VAN (Value Added Network gateway, see, for instance note)

Network abstraction

• Hiding network details for the user (application software)
  • Changes in a network topology would otherwise necessitate modifications to application
  • Machine specific applications require code duplication across the network
    
• Separating the application programs from data communication activities
• Network interconnection via routers (previously called gateways): computers physically attached to networks they connect.
  

Guest lecture

Internet addressing

• All interconnections among physical networks are provided by dedicated computers.
  Routers use networks (and not individual hosts) when routing a packet.
• An Internet address specifies uniquely
  an access route to a host
  multihomed hosts (for instance, routers) have multiple addresses
• The anatomy of an IP address
  • 32 bits in three fields
    class; network ID; host ID

  • Class	A	network	More than 65,000 hosts
    	B		Between 256 and 65,000
    	C		Fewer than 256 hosts

  • Few huge networks (7 bit net ID)
    Plenty medium sized (14 bits)
    Innumerable small nets (21 bits)

  Figure 4.1

• Notations & conventions

  Dotted decimal	Example: 128.223.8.31 (Majestix)
  Dotted hexadecimal	80:DF:08:1F
  Big-Endian	most significant bit first

• Internet addressing problems
  • Changing network size: class transition
  • Changing network membership
  • Accessibility depends on the address used (in the figure below, addres I_3 used by A to reach B will not work if B's link to Network 1 fails, even though the two hosts could ycommunicate through address I_5.)

  Figure 4.2

• Addressing policy and assignment control lies with Internet Assigned Address Authority(IANA)
  New internet members assigned addresses by INTERNIC

Special (reserved) addresses

• Broadcasting (one-to-all)
  • Host id: All 1s
  • Local (limited) broadcast: all 32 bits 1
  • Hardware broadcast (common medium)
• Multicasting (class D; prefix 1110)
  • Hardware multicast
  • Multicast groups:
    • "Well-known"
    • Transient (joining protocol IGMP)
  Default addressing (all 0 field):
  This network, this host
• Loop-back address
  (127.x.x.x = 01111111.x.x.x, here, x -- arbitrary)

Address resolution

• Mapping Internet address to physical address.
  • In the same netowkr:
    Which host corresponds to the given IP address?
  • In different networks:
    Which router?
• Direct mapping (in networks with small address space)
  Efficiency!
  • A few machine cycles (for instance, when masking the host ID field of the IP address)
  • Tables of corresponding IP and physical addresses (EG. hash tables)
• Dynamic binding through Address Resolution Protocol ARP: Finding the phusical address of a host on the same network via control messages.
  • Host A (the sender) broadcasts request for the receiver's (host B) physical address (B identified by its IP address)
  • B responds with its IP and physical address (in one frame sent directly to A)
• Refinements
  • Hosts maintain cache (fast local memory) of recently requested (and received) bindings. (Cache strategies, limited size, problems)
  • Each ARP request includes sender address binding (Other hosts can cache or update its entry if they so choose)
• ARP is considered a part of the physical system (network interface layer).
  In some network technologies it is necessary, in some others, not.
• Implementation of ARP

  ---

  Sender

  ---

  • The sender either:	has the binding cached, or
    	broadcasts the ARP request.

    
    (The packet is stored while the request-response goes on; it may be discarded if timed out, or joined by other packets to the same destination)
  • Upon receipt of a reply, a cache entry is created (or updated); the reply is matched with a request and the corresponding packets are sent.

    ---

    Receiver

    ---

  • Receivers of an ARP request
    • Update or create cache entries for the sender
    • The destination host replies with its binding directly to the host

Anatomy of an ARP frame

• ARP messages are transmitted in special frames with identifiable type (cf. two-obctet field in an ehternet frame header)
• Depending on network technology, ARP message header has
  • Hardware (data link) type
  • Protocol (network) type
  • Physical address length
  • Protocol address length
  • Sender's physical address
  • Sender's protocol address
  • Tarket's physical address
  • Target's protocol address