[Blog recommendations] some misunderstandings about Server Load balancer Technology

[Blog recommendations] some misunderstandings about Server Load balancer Technology

This blog post is from Bkjia's blog Li yunxing. If you have any questions, please go to the blog page for an interactive discussion! Blog blog address: http://quetzal.blog.51cto.com/3313934/1631459

Nowadays, Server Load balancer is no longer a new term, nor is it a new technology. It is mainly used to solve the limitations of Single-host load capacity. But the problem is that your application has reached the single-host load limit, not necessarily. Many people who do not know how to infer the bottleneck and how to solve the problem start to blindly add machines. It seems that as long as they can add machines, performance is not a problem, the server Load balancer technology has become a popular white horse. You have used 10 servers for a single machine. This is obviously a cost issue.

In fact, many problems can be easily solved. refer to the following scenarios.

1. A single database server runs two databases, and in the production environment, the load of these two databases is similar, and the tool monitoring shows that hard disk I/O has encountered a bottleneck, what should I do? I believe many people will immediately think of a solution, such as adding a server, configuring read/write splitting, or synchronizing replication, and running two databases separately on a separate server. In fact, this problem is very simple. Since the bottleneck is I/O, we will solve the I/O problem, add a hard disk, and put the data of the two databases on different hard disks. Isn't it solved, in some scenarios, increasing the memory can also reduce disk I/O and easily solve the problem.

2. People familiar with databases know the importance of indexes in queries. However, few developers I have met who are actually able to make good use of indexes. In many cases, there are considerable performance differences caused by indexes and no or improper indexes, if this is the case, you need to add machines to solve the problem that you don't know how many machines are enough to use.

3. For File Download servers, the bottleneck of such servers usually occurs on the network throughput of a single machine. The network throughput of a single machine depends on the network throughput of the network adapter, it is hard to say that you need to add machines because of NIC problems. In fact, many people ignore the problem that one machine can add multiple NICs.

In fact, there are many examples. However, I believe many people have found problems in the above three scenarios, but we are too concerned about Server Load balancer and cluster technologies, however, some problems can be easily solved at a low cost.

Conclusion: Do not blindly apply various technologies, learn to think, and discover the essence of the problem is the key to better solve the problem.