Analysis of the correspondence between throughput and network traffic

Throughput is an indicator of the network traffic that can be processed by network devices per second. In reality, network administrators often use the term "traffic" to measure the performance of devices, what is the relationship between 1-to-1 and 1-to-2? It has plagued many people.
First, let's clarify the concept:
Traffic: it is a vague term. It is not a device indicator term. In this case, it is compared with throughput, we understand it as the data volume per second that flows through a device (unit: bps ).
Throughput: the maximum data rate (bps) that the device can process (accept and forward) per unit of time ).
The English word "throughput" is throughput. Both Chinese and English words are very relevant. "through" means "pass". "going from one end is called" pass "." throughput "means" pass: swallow and spit it out. It is easy to analyze the concept after we have enough.
Most people confuse traffic and throughput from the full duplex of the physical port of the device. Assume that one physical port is a gigabit Full Duplex (each cable has two wires, each port can run 1 GB of data at the same time. The total data volume that a port can transmit is 2 GB. Is the real data traffic 2 GB?
Let's analyze it as follows:


Now let's test the traffic and throughput of switch1 (to simplify the process, we suppose switch1 has only two ports), and switch1 is connected in the network, each port is connected to switch2 and switch3 (we know that each port has two internal cables), and the upstream data is from switch2 to swit3 and then to switch3, the downstream data is from switch3 to switch1 and then to switch2.
In a certain second, assume that 1g of data enters the switch1 device and goes out from the switch3 end in the direction of the red line, we call it switch1 that processes 1 GB of data (1 Gbps of throughput), also known as forwarding 1 GB of traffic. In the same second, if 1 GB of data, such as blue lines, goes from the uplink switch3 to the switch1 3 end and goes out from the downlink switch2 end, therefore, switch1 also processes 1 GPS data (and 1 GB of throughput) and forwards 1 GB of traffic down. At this time, the two ports of switch1 are full, but we can see that the throughput and traffic are calculated as 2 Gbps.
The above method is a string connection, which is more intuitive. Let's take a look at the bypass access. Is it the same?


Swit2. 2 is mounted next to Switch2. both upstream and downstream data are sent from switch2 to swit2., and then processed by swit2. and sent back to switch2.
Assume that the uplink data is 200 M in one second, Switch2 is sent to swit2 through the red line, swit2 2 is forwarded back to Switch2 through the blue line, downstream data is 800 M, and Switch2 is also forwarded to Switch1 through the red line, swit2 forwards the data back to Switch2 through the Blue Line. At this time, the upper and lower rows add up to 1 GB of data, while the data flowing through the red line and the Blue Line is 200 M + 800 M = 1 GB, both lines are full. Therefore, in the bypass mode, a port is full and only 1 GB of data traffic (that is, 1 GB of throughput) exists ). Instead of computing 2G as the full port duplex.
 
Why?
The most fundamental reason is that the two lines in the full-duplex port cable actually run the same data.
As shown on the right, the red line and the Blue Line are the two lines in the port cable. 1g data passes through the red line from switch2 to swit2, and then goes through the blue line, the full duplex of a port is occupied, while the two lines actually transmit the same content, so the traffic is 1 GB, And the throughput is calculated as 1 Gbps.
Therefore, for a full-duplex gigabit port, the actual data transmitted per second is at most 1 GB (for the serial connection mode, we can see from the first figure that, 1G data simultaneously occupies one line of Port 1 and one line of Port 2 within one second. In turn, the two ports actually transmit 2G data ).
Computing traffic in reality clearly does not add duplicate data. Finally, we can conclude that the throughput that a device can process is the same as the traffic that a device can process per second, the two can be the same. However, it is wrong to add inbound and outbound data on all ports on a device as the traffic forwarded by the device. However, for the serial connection mode, it should be correct to check the total incoming and outgoing data of the half port as the traffic. For example, the ports in the downlink Intranet Add the incoming and outgoing traffic to the port, the total traffic of the device.
(Wyl)
This article is from the ADC technology blog