[network counting] 1.4 network delay, packet loss and throughput

This article is based on 《 computer network —— top-down approach 》

Type of delay

1. Processing delay ： Check the packet of the message （packet） The first part of and the time to decide where to direct the group is Processing delay Part of
2. Queuing delay ： In the queue , The time when a packet is ready for transmission on the link is called Queuing delay , The queuing delay of a particular packet depends on the number of packets that arrive in advance and are waiting for transmission over the incense link . If the queue is empty and no other packets are being transmitted , Then the packet delay is 0.
3. Transmission delay ： It refers to the delay when all packets are sent to the link , For example, a 10Mb The packet of is sent to a 100Mbps It takes time on the link 0.1 second , This is the transmission delay , Please be sure to distinguish from the propagation delay below .
4. Propagation delay ： Propagation delay refers to the time taken for a packet to propagate from its origin to its destination , The rate of transmission depends on the physical media of the link , Such as twisted pair 、 Optical fiber, etc .

The difference between transmission delay and propagation delay lies in , Transmission delay is the time taken by the router to upload packets to the link , It has nothing to do with the distance between two routers . The propagation delay is the time of packet propagation on the link , It is related to the distance between two routers .

The total delay is the sum of the above four delays

Queuing delay and packet loss

There is a problem about packet loss , First of all, we need to define a concept ： Flow intensity

The traffic intensity is the ratio of the average packet arrival rate to the routing transmission rate . If the flow intensity is greater than 1, That is, the packet arrival rate is faster than the transmission rate , Then the sending speed will not catch up with the receiving speed , Thus, the waiting queue can be extended indefinitely , in other words , Queuing delay tends to infinity ！ therefore , When designing the system, the flow intensity shall not be greater than 1.

In theory , At a flow intensity less than 1 In the system , If the packet arrives periodically , That is, if the arrival interval of the group is the same , In theory, there is no need to queue . But in fact, the arrival time of each packet is random , If multiple packets arrive at the same time , A large delay may occur . In the case of random arrival of groups , If the flow intensity is closer 1 The larger the average queuing delay , Because the less time the route is idle .

And the extension of queuing delay means that there are more packets waiting to be sent , But the packet cache that a link can hold is limited , If there are so many waiting packets that the cache is full , The next arriving packet will have no place to store , The router will discarded （drop）, This is it. Packet loss

Packet loss ratio , That is, the proportion of packet loss increases with the increase of traffic intensity , Therefore, the performance of a node can be measured according to the delay , It can also be measured according to the packet loss probability .

throughput

Suppose there is a router connecting the server and the client , The rate of routing to clients is R_c, The rate of routing to the server is R_s, Because the maximum throughput of a whole link depends on the link with the lowest transmission rate （ Similar to water pipe ）, Therefore, the maximum throughput on this link is min{R_c,R_s}. The promotion is , There is a problem between the server and the client N A network of links , Then the file transfer throughput from the server to the client is min{R₁, R₂, R₃,…, R_N}