Reliable data transfer over a channel bit error rdt : 20, Computer Networking

Reliable  data  transfer over a channel bit error  rdt : 20

A more realist model  of the underlying  channel  is one  in which  bits in packet may be  computed. Such  bit errors typically  occurs  in the physical   components  of a network as a packet  is transmitted propagates or is buffered.

Before  developing a  protocols  for reliably  communication over  such a channel  receiver might  say ok after  each  sentence has bee beard understood an recorded. It the receiver hears a garbled sentence he asked to  repeat the garbled sentence, this message  dictation protocols uses both positive acknowledgment and negative acknowledgments  these control  messages allows  receiver correctly and what has  been received in error  and thus  requires repeating.

Figure  shows  the FSM representation of rdt 2.0 a data  transfer  protocols  employing  error   detection positive  acknowledgments  and negative  acknowledgments.

545_Reliable  data  transfer.PNG

                                                     Figure  rdt 2.0 FSM specification

The send  side of  rdt 2.0  has two states. In the left  state  the send  side  protocols  is waiting  for data  to be passed  down  form the  upper  layer  when  the rdt send event occurs the sender will create a packet containing  the data  to be sent along  with  a packet  checksum  and then  send  the packet  via  duct send operation. In the  right most  state the sender protocols is waiting  for an ACK or a NACK  packet  from the  receiver. If an ACK  packet  is received. The sender  knows that the most recently transmitted  packet  has been  received correctly  and thus  the protocols  returns to  the state  of waiting  for data  from the upper layer. If  a NACK is received the  protocols retransmits  that last  waits  for an ACK or NACK to be returned by the receiver in response to the  retransmitted data  packet. It is  important to note that when  the sender  is in the  wait for  ACK or NACK state it  cannot get  move data from the  upper layer that is  the rdt send event  cannot occur that will happen  only after the sender receives an  ACK  and leaves this state. Thus  the sender will  not send  a new  piece  of data until. It is sure  that the  receiver has correctly received the current  packet  because  of this  behaviour  protocols  such  as rdt 2.0  are known as stop  and  wait  protocols.

The receiver  side FSM for rdt 2.0  still  has a single  state on  packet  arrival the  receiver replies  with either an ACK  or a NACK. Depending  on whether  not  the  received  packet in corrupted. In  figure the notation rdt rev and corrupt  corresponds to the event in which  a packet  is received and is  found to be in error. Protocols rdt 2.0 many  look  as if  it works  but  unfortunately it has  a fatal flaw  or may be possible  that the  ACK or NACK  packet  could be corrupted.

Possibilities  for handling corrupted ACKs or NACKs

a.For  the first  possible it the sender didn't  understand  the ok or please repeat that reply  from the  receiver the sender  would  probably ask what did you say? The  sender  would  then repeat the reply. But  what if the  speaker what did you  say? Is corrupted  the receiver having  no idea whether difficult to understand the garbled sentence was part  of the dedication or a request to repeat he last reply, would  probably  then respond  with  did you  say?  And then  of course  that  response  might be garbled.

b.A second  alternative  is to  enough checksum bits to allow  the sender not only  to defect but also  to recover from bit errors. This solves  the immediate problem  for a  channel  that can  corrupt  packets  but not  lose them.

c.A third  approach  is for the  sender simply  resend the current  data packet when it  receives a garbled ACK or NACK  packet. This  approach  introduces duplicate packets  into  the sender to receiver channel. The fundamental problem with  duplicate packets is that the  receiver doesn't  know  whether the ACK or NACK  it last  sent was  received  correctly at eh sender. To  handling  the duplicate packets sender. Sender  adds sequence number  to each packet  retransmits  current  packet  if ACK / NACK  garbled  receiver discards  duplicate  packets.

Posted Date: 3/11/2013 7:12:29 AM | Location : United States

Related Discussions:- Reliable data transfer over a channel bit error rdt : 20, Assignment Help, Ask Question on Reliable data transfer over a channel bit error rdt : 20, Get Answer, Expert's Help, Reliable data transfer over a channel bit error rdt : 20 Discussions

Write discussion on Reliable data transfer over a channel bit error rdt : 20
Your posts are moderated
Related Questions
What is WDM? WDM is conceptually the similar as FDM, except that the multiplexing and demultiplexing includes light signals transmitted by fiber optics channel.

Which NetWare protocol works on layer 3--network layer-of the OSI model? Ans)IPX

Q. Permutation representation through Matrix? Figure: Clos Network The functioning of this above permutation is displayed in Figure. This permutation is able to

Hosts are generally, individual machines at a specific location. Normally Resources of host machine are shared and can be used by any user on Internet. Hosts & local networks are g

Discuss an example of threats in computer software Assume only one attack out of the above list i.e., the virus attack. Though the present anti-virus software solutions can det

What is Beaconing? The process that permits a network to self-repair networks problems. The stations on the network notify the other stations on the ring when they are not rece

The following information was obtained from a host computer: 00:05:17.176507 > S 2688560409:2688560409(0) win 16384 (DF) (ttl 46, id 2096

QUESTION (a) Briefly describe three security requirements of an agent (b) List four threats that may be caused by malicious hosts to mobile agents (c) Describe the forwar

Write the definition of SAN A SAN is a separate network to handle storage requirements. The SAN detaches storage tasks from specific servers and makes a shared storage facility