Shulou(Shulou.com)06/02 Report--

Today, I will talk to you about the reasons for upgrading the kernel of linux, which may not be well understood by many people. in order to make you understand better, the editor has summarized the following for you. I hope you can get something according to this article.

Due to the need to use docker, some features are not supported in lower versions of linux, such as namesapce,cgroup. Therefore, the kernel needs to be upgraded.

Why upgrade the kernel?

Yes, a lot of people do not allow such a view. Perhaps they feel that it is enough to use the most durable and stable interface provided by the kernel, and most custom features need to be developed in user space. Of course, this is also in line with UNIX's philosophy, and they feel that users write their own so instead of bash. Python scripts will improve the competitiveness of your software. Because you can't see the source code. A so can achieve load balancing of OpenVPN, and you don't know how it does it, so behind this sense of wonder, people will think that its technology is very advanced. On the contrary. Assuming that they use the features provided by the kernel, they will feel that such a technology can be used by anyone and is not exclusive to them. What is done loses its competitiveness.

Such a view is completely wrong and extremely wrong and childish.

The point is not how many technical details you keep that others don't know. The key is how quickly you can launch the simplest version that you can use, and how often it can be continuously updated and optimized, and you don't have an epiphany behind closed doors. It's feedback from users. Hurry up. That's what's fundamental!

How can I be quick?

Simple, can not write their own directly take it, can make use of ready-made features directly. In this way, you can save a lot of work. Focus on how to simply expand functions to meet your needs, rather than delving into a technical detail.

When you write so in C, don't you also call the API provided by other people's C libraries and third-party libraries (even written by your colleagues)? Isn't this here, too? What you hide is the details of how to assemble all this, and this is only one way, no one cares how you assemble the components into functions, what users care about is the ability to use them.

For example.

I spent weeks talking on the Linux 2.6.32 kernel and couldn't figure out how to get multiple OpenVPN processes to share a TUN network card. In fact, I rewrote the TUN driver, but I didn't pass the stability stress test, so I don't know if I can use it. What I have finished may be nothing more than a feasibility pre-study, and later, due to no project or personal reasons, such results can only stay on paper. But after all, it consumed my time and energy, although I didn't take up my work time (it was different at a later stage), when my wife and daughter were away for a long time. I'm alone at home. I have plenty of time after work or on weekends. But after all, because of this thing, I still owe several documentaries about ancient Rome. I spent a very long time studying how to achieve the concurrency of UDP services. The solution in UNIX Network programming has also seen that it is simply impractical for services like OpenVPN. I have thought of letting the child process inherit the UDP description descriptor of the parent process through fork, but there will be surprise problems. I know that there is an option of SO_REUSEPORT, but the implementation of Linux does not meet my needs (this is an implementation problem, not an interface problem). So it took me a very long time to write the UDP multi-port load balancing kernel module, and at the same time I came up with a random DNAT scheme using iptables. However, as long as I upgrade the kernel version to 3.9.6 or above, I can solve all the problems. I just need to set a SO_REUSEPORT option on UDP socket, and I no longer have to write code.

The premium version of the kernel has three benefits, more efficient, and many other new features.

Previous bug fixes, but for newly introduced features, new bug may be introduced. But you just need to be careful with the new features. For example. The features I will actually use, such as TCP's Fast open, network namespaces, and multi-queue TUN NICs, are not supported on the 2.6.32 kernel (namespaces are available. However, it is too inconvenient to use the interface. Do you want to spend a month writing a multi-queue TUN driver yourself? Or do you want one that can be used directly? Admittedly, the TUN driver that comes with the premium version of the kernel is not written by you and does not represent your strength. It won't make you competitive, but your goal is nothing more than using a multi-queue TUN, your goal may be to build an alternative OpenVPN. Easy-to-use features that no one else has thought of, possible new requirements that you extract yourself. These are the competitiveness, not the realization. I reached Rome through Appiah Avenue. And you go through the Alps, yes, you are more able to embody strength and perseverance than I do. however. I got there first.

After reading the above, do you have any further understanding of the reason why linux upgraded the kernel? If you want to know more knowledge or related content, please follow the industry information channel, thank you for your support.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.