? Too much memory in the intermediate nodes is as harmful as too little memory. 7.Could the congestion problem be solved with high-speed links?
? Introducing high-speed links without proper congestion control can lead to reduced performance the same speed
(a) The time to transfer a particular file was five minutes.
(b) The link between the first two nodes was replace by a fast 1 Mbits link, the transfer time increased to 20 Minutes!
? With the high-speed link, the arrival rate to the first router became much higher than the departure rate, leading to long queues, buffer overflows, and packet losses that caused the transfer time to increase.
? The protocols have to be designed specifically to ensure that this increasing range of link speeds does not degrade the performance. 8.Could the congestion problem be solved with high-speed processors?
? Similar to that for links. Introduction of a high-speed processor in an existing network may increase the mismatch of speeds and the chances of congestion.
? Introducing high-speed links without proper schemes Congestion occurs even if all links and processors are of the same speed. ? An example of the balanced configuration
? - Assume all processors and links have a throughput capacity of 1 Gbits. A simultaneous transfer of data from nodes A and B to node C can lead to a total input rate of 2 Gbits per second at the router R while the output rate is
only 1 Gbits per second, thereby, causing congestion. 6. solutions?
? Congestion in networks is a dynamic problem. It cannot be solved with static solutions alone.
? We need protocol designs that protect networks in the event of congestion. ? The explosion of high-speed networks has led to more unbalanced networks that are causing congestion.
? In particular, packet loss due to buffer shortage is a symptom not a cause of congestion.
? Solution: proper protocols and mechanisms design, e.g. Admission Control, Scheduling, et. al
7. “end to end” and “point to point”
a) end to end communications: i.
Data communications on a path between the source node and the destination node. The path possibly comprises multiple links
b) point to point communications: i.
Data communications on a link connecting the adjacent nodes
8. Performance Issue ? Stop-and-wait
? Without error t0 =2tprop+2tproc+tf+tack=2tprop+2tproc+nf/R+na/R
? Let Pf be the probability that a frame transmission has errors and needs to be retransmitted.
? tsw=t0/(1-Pf)Efficiency is: ηsw=((nf-no)/tsw)/R
? Where nf, nano are the total number of bits in the frame,number of bits in the ack, and number of bits in the overhead. R is the bandwidth. 9. Constrains on Windows’ Size(a sliding window protocol)
n bits sequence numberThe size of sending window: WsThe size of receiving window: WrWs+ Wr≤2n 10. Binary exponential backoff ? When a collision occur
each station waits 0 or 1 slots before try again
? In general, after i collision, a random number between 0 and 2i -1 is chosen. ? Maximum is 1023
? After 16 collisions, the controller gives up and reports failure. 11. Three types of switching fabrics
12. Input Port Queuing r Fabric slower than input ports combined ->queueing may occur at input queues
r Head-of-the-Line (HOL) blocking: queued datagram at front of queue prevents others in queue from moving forward r queueing delay and loss due to input buffer overflow!
16IP Fragmentation and Reassembly
相关推荐：
loading