TCP optimization: Three acceleration methods for TCP/IP network traffic

Many companies are facing an interesting WAN planning problem. An organization with 3/4 said it would not increase the budget, but the overall planned budget growth was only 5 ~ 8%. The budget is very limited. However, nearly 50% of projects significantly increase bandwidth consumption.

One way to solve this problem is to optimize existing technical solutions. Many network traffic is still based on TCP/IP. TCP provides reliable and ordered data packet transmission, which is used by most Web applications, emails, and files. However, the stream management algorithm of TCP is not advanced: if the network or the receiving end cannot process the transmission speed, it may cause packet loss, timeout, or excessive unordered packets, the transmission speed of the network stream will be halved. Later, the speed will increase slowly, but it is still significantly slower than before.

Therefore, the performance of TCP applications (such as large file transfers) that require long data transmission over the WAN may be severely affected: the longer the transmission time on these links, this means that the sending end may not be able to perceive the performance degradation for a longer time, and then it will take a longer time to respond to the speed reduction. TCP optimization can alleviate these problems by avoiding packet loss or disordered transmission on the network, or directly correcting TCP streams and adjusting the speed.

The following three tips can help IT managers and engineers optimize TCP:

Tip 1: Do not want to increase the bandwidth, but pay attention to the congestion problem. The problem is not necessarily how much data needs to be transferred from point A to point B, but how quickly all independent senders and receivers can complete data transmission. Fast Speed up/fast speed down/slow speed up/speed up and so on will accelerate the application transmission speed of the non-congested network, but it will increase the probability of congestion when the new sending end is added to the transmission. For some Optimizers that slow down the initial or reduce the initial speed by changing the speed-up behavior, they can prevent the first congestion, in addition, when a new traffic stream starts to increase the transmission speed, the transmission speed is slowed down in advance to better deal with threats.

Tip 2: Pay attention to the network priority. The traffic forming optimizer is designed to ensure that the Organization can control the consumed bandwidth. Control can be positive, that is, to ensure that a specific application, device, or user obtains a certain amount of bandwidth. It can also be negative, that is, to limit the bandwidth used by a specific user, device, or application. To use them to optimize TCP traffic, you need to define the traffic type in the system, and which applications and users have the priority to obtain Network Resources in the case of congestion.

Tip 3: Leave TCP alone. Nowadays, more and more important traffic (video conferencing, VoIP) does not use TCP/IP. On the contrary, it uses UDP/IP (User datatime Protocol over IP ). However, UDP does not have the same flow control mechanism as TCP, so TCP may be more suitable for robust optimization. The UDP sender is only responsible for sending data packets, while the receiver receives data. Considering that enterprises generally set a higher priority for UDP-based Real-time session traffic, the TCP optimizer can separate TCP traffic from UDP traffic, in addition, limiting TCP traffic to transmission through a small number of channels allows UDP traffic to pass first, which can alleviate this problem. (Of course, there are more features in the optimizer running on the IP layer)