reliable protocols

Summary

RDT protocols

Concept

Reliable data transfer

• in the transport layer
• network layer is best-effort, not completely reliable
• packets can be:
  • dropped
  • corrupted
  • delayed arbitrarily
  • reordered

Requirements

• guarantee packet delivery and correctness
• deliver packets(to app) in the same order they are sent(from server)

Reliable Delivery Transfer(RDT)

RDT 1.0

• assume channel is reliable, no loss, no errors

TODO find a better renderer for FSMs

RDT 2.0

• channel might flip bits
• use a checksum to detect bit errors
• ACK -> packet was received well
• NAK -> packet has errors

ACK and NAK corruption

• sender won’t know what happened at the receiver
• if corrupted, just retransmit -> but packet might be duplicated, receiver has no way of knowing

RDT 2.1

• add sequence numbers to packets
• allow duplicates to be discarded

RDT 2.2

• remove NAK
• replace with ACK numbered with last correctly recieved packet
• duplicate ACK will trigger sender to retransmit

RDT 3.0

• packets may be lost
  • if no ACK by timeout, retransmit
• packets may be delayed for an arbitrary time
  • assume loss, retransmit -> receiver will detect duplicates
• order is preserved