Overview

• Data Link protocols must provide
  • Frame Synchronization
  • Flow Control
  • Error Control
  • Addressing (on a shared link)
  • Control and Data on the same link
  • Link Management

Flow Control

• We will see this repeated again and again.
• Make sure that the receiver is not overwhelmed
  • The sender is given data in a buffer, it must encapsulate this data and send it.
  • Perhaps it needs to compute a crc, but that is mostly it.
  • The receiver must do the crc
  • And place the data in the proper place
  • And make sure all is there.
  • And hand it off to the upper layer, which may not be ready for it.
• The transmission time is the time it takes to emit the bits onto the wire. This is proportional to the length of the data.
• bits * seconds/bit = seconds
• The propagation time is the time it takes for a bit to travel over the wire from source to destination.
• meters x seconds/meter = seconds.
• Data is sent in a frame.
• Assume (for now)
  • Frames arrive without error
  • Frames arrive in error.
• Stop and Wait flow control
  • Each frame must be acked.
  • Sometimes Large data blocks are broken into smaller blocks
    • Limited buffer size on sender/receiver
    • P_{frame error} = (P_{bit error})^{frame length in bits}
    • IE the larger the frame the more likely it will have an error.
    • And it is more expensive to retransmit.
    • Freight train vs VW bug at intersection.
  • Bit length of a link
    • Let B be the bit length of a link.
    • B = R x d/V
    • R = data rate in bits per second
    • d = length, in meters, of the link
    • V = propagation velocity, in m/sec

             R bits      x   d meters x sec
      B =   ---------        --------------
               sec             V meters
      	     

    • This is how many bits can be in a link at one time.
    • Let L be the number of bits in a frame.
    • Let a = B/L
    • if a < 1, then there are more bits in the frame than in the link.
    • ie, the receiver will start to get the message before the sender finishes transmitting.
    • if a > 1, then there are fewer bits in the frame than in the link.
    • ie, the receiver will not start to get the message until the sender is done transmitting.
    • Stallings gives us a nice pictorial representation
    • 
    • Note that when a>1, the line is seriously underutilized.
      • Most of the time is spent waiting for the frame to traverse the media.
      • Most of the time, the media is empty.
      • You should work your way through example 7.1 it is not too hard and it is really illustrative of what is going on.
  • Sliding-Window flow control
    • In this model, the receiver allocates space for n frames.
    • Multiple frames can be in flight at one time.
    • Each frame must have a sequence number.
    • Sequence numbers present a bit of a problem.
      • Longer numbers mean more frames in flight.
      • But also more bits in the data field.
      • Wraparound is a problem.
    • If k bits are used for a sequence number.
    • Only allow n < k frames in flight at a time.
    • This is called a window.
    • The sender
      • Keeps track of the last sequence frame acked
      • Also how many frames it has sent.
    • The receiver
      • Can ack any number of frames with a single ack.
      • This slides the window.
    • The ack actually contains the sequence of the next frame it is ready to receive (RR)
    • Stallings gives us a nice image again.
    • 
    • If the receiver becomes overwhelmed, it can send a RNR
      • Receive Not Ready
      • Sequence number of last received frame
      • Sender does not send any additional frames until a RR is received.
    • If data is being sent both ways, a second sequence number can be added to the header information.
      • Assume B sends a data packet to A
      • The second field is a RR for the data sent from A->B
      • This is known as piggybacking.
    • Stallings calls Sliding Window a form of pipelining.