When there is a problem with Web Access performance, how do I go deep into it?

For operations or development engineers, when encountering access performance problems, the first need to locate is the problem in which link, is the network problem, the problem of the service side, or the problem of the client?

Often technicians like to focus on the availability of back-end services, and whether the front-end interface can be loaded properly, whether the full rendering is not too concerned. But from the point of view of business, the interface is the ultimate business, and the business is realized through human-computer interaction.

What scenarios do we encounter on a daily basis that require targeted access performance bottlenecks?

• Different geographic access to the same application speed when the difference is large

• When accessing the same application speed at different times, the difference is large

• A low-version browser causes the interface to load slowly or fail

• Different operating systems (especially mobile phones) cause the interface to load slowly or fail

• External chain resources or third-party services slow down the loading speed of the interface

There are more scenarios that need to be positioned. Do we have an easy way to locate the problem of the network, the problem on the server, or the client? The answer is yes. Modern browsers provide us with some convenient conditions to get from the start of the request to the page rendering to complete each link of time, note that this must be a modern browser, they have a common feature, that is to follow the TIMINGAPI.

I've previously communicated with you, "monitor the real user experience, start with one line of code." Through a line of code to achieve data acquisition, Web Access performance positioning is relatively simple, see.

Let's look at how to interpret the data.

Measure user experience from a user's point of view

The first is the name of the operation, such as "Open Home", which is a redirect operation, the corresponding URL is http://

How does the "open home" Operation come here? Here first sell a Xiaoguanzi next, this is the superior cloud Web and any other product is not the same place, other products are often give you a URL address, content is what, context operation is what you need to guess, non-technical ape class can not understand.

• Focus on key metrics

Oapdex

O Average Response time

o Average user-operable time

O Throughput Rate

Where the average response time can be decomposed into

O Client Time

O Network Time

O Service-time

In this example, we found that the client time accounted for 15 seconds, and the network and server accounted for very little time, the two only 1 seconds less than. The description page takes a lot of time to load and render at the front end. When this happens, users tend to think that the server is slow, but in fact, we understand that 70% of the access performance problems in the front, this and our cognition is very different. So who's going to carry this pot? About the rate, 70% of the pot should be a developer to the back, operation and maintenance of data to quickly hit the past it.

• In-depth analysis of key indicators

o Correlation between response time and operand trends: to analyze whether there is a positive correlation between the two

o Refinement of response time: average, median, slowest 5%

o Response time by satisfaction distribution: what is satisfactory, what is tolerable, and what is not

• Sample Surveys

We will sample the dissatisfied users, by sampling we can see which areas, which browser users will be more prone to access performance issues, and can be detailed to the page's resource load time. What is the use of this data? You can know which page resources can be further compressed optimization, reduce the number of HTTP requests, you can let us check whether the CDN is really effective, or no CDN-enabled applications can consider CDN acceleration and so on.

The above are some of the personal experiences of accessing the performance experience, or it is recommended that you personally feel it is not the same.

When there is a problem with Web Access performance, how do I go deep into it?