Common network performance metrics-network latency and response time

 

Common network performance metrics-network latency analysis

Network latency includes transmission latency, transmission latency, queuing latency, and router execution time.

Propagation delay is the time that an electrical signal or optical signal takes to transmit from one point to another. Latency is defined as a function of the speed of electricity/light in copper or optical fiber. Essentially, data signals are transmitted in the telecom network at a speed of about 2/3 km per second at a vacuum speed. In fact, with the expansion of the network in the national or international scope, the propagation delay has become the largest part of the total latency and cannot be reduced. Propagation latency is the main reason why WAN performance is always inferior to LAN performance. The larger the distance between sites or nodes, the slower the WAN speed.

The transmission delay (transmission delay) is the time it takes to transmit a certain amount of data at a given transmission rate. The delay can be expressed in mathematics directly. For example, for a 1000 Kbps link, it takes 10 seconds to transmit kbps of data, while for a Kbps link to transmit data in 2 seconds. You can create a Transfer Rate Table in the network. You can see that the transmission delay is particularly obvious when a large amount of data is transmitted.

Note:
Increasing the bandwidth of the WAN link can reduce transmission latency, but cannot reduce transmission latency. If you use a pipe with a certain width and length for analogy, increasing the width of the pipe is similar to increasing the bandwidth, but this does not reduce the length of the pipe. Therefore, if an application is limited by bandwidth, increasing the bandwidth will improve the performance. However, our experience shows that the vast majority of Interactive Client/Server Applications send a relatively small amount of data, which is mainly affected by transmission latency.

Queue delay (queuing delay): when many devices (including customers and servers) in the network want to send data on the network at the same time, a queue delay occurs. In the case of media access latency, each network device stops sending its packets before receiving a sending opportunity. In the case of queuing delay, all data packets are sent to the same device (usually a router), which can be connected to the WAN ). The transmission bandwidth is the main factor that determines the time required for the router to send all queued data packets. For example, if the backlog on the vro is 10 KB, it takes more than one second to send the queue on the 56kbps link. On the other hand, it takes less than 1/10 seconds on the T1 link with a transmission rate of 1544kbps. The router execution time (latency) is the time required by the router to transmit a data packet from one interface to another. This is usually a small value. The execution time of the Cisco router can be reduced by using different switching paths. Although the latency itself is small, if there are many hops, the execution time of the end-to-end router will become a significant factor. How to minimize the execution time has become a major challenge for large-scale network design. The router execution time is more problematic on the campus network than on the company intranet.

Transaction usually takes three aspects: data packets pass through the network, the server retrieves data, and the client receives data and requires more data.

Assume that the client sends data to the server request. Therefore, measurement at least two points is required: the local network of the client and the local network of the server. On the client side, the client latency will be measured. At the data packet level, this is the latency between the client receiving a frame of data and requesting the next frame of data. Similarly, the client network must measure the composite (Network + server) latency. The second measurement is performed on the server network to measure the server latency.

The more interactions between the client and the server, the more data packets are transmitted back and forth, and the more sensitive the application is to network latency. This sensitivity is partly due to a turnaround time related to each exchange. For example, after the client sends the message, it waits for the response from the server to send the message again. In this way, any increase in network latency will result in a 2x increase in the turnaround time. First, the increase in network latency from the client to the server, and second, the increase in the returned latency.

In a LAN, the turnaround time may be several seconds. However, in a company's Wan that spans a country or a whole world, the turnaround time may be dozens or hundreds of seconds. Obviously, this increase in turnaround time may be disastrous for applications. This latency is one of the main reasons why many application development projects fail. For example, 50 ms of network latency (MS between San Jose and NC Riley is a typical network latency) will increase the application response time by 2 seconds.

When analyzing the response time, you need to ask the following questions:

How many times does the application require turnover? This check of application traffic can be performed using protocol analysis tools or network detection tools. This is the first step to understand the application traffic diagram. The results may show that a single application requires dozens or hundreds of network turnaround times.

What is the network latency? As mentioned above, this latency measure is basic to the application response time performance. Knowing the network latency also facilitates network design decisions. For example, the latency of the 1000-mile T1 link is smaller than that of a satellite.

Does the sum of values T1, T2, and T3 fall within the response time required by the Service Level Agreement or operation target? If you answer "no", you may also have the opportunity to adjust the application, reduce the turnover, and improve performance. It is also possible that the application cannot run as expected.

Another way to view this is that when you evaluate a program for a cross-Wan application, the more data it sends, the worse the data. This is especially true for relatively slow Wan. They will be unable to bear the need for applications to send a large amount of data and process it in large quantities. Therefore, when checking the application response time, it is necessary to check the application protocol behavior, network performance, and the performance and scalability of the Client/Server.