Shulou(Shulou.com)06/03 Report--

In our last article, we talked about the modeling method when the performance test load model landed, and here we start with the first and most commonly used method:

Raw data analysis

Business scenario modeling is mainly to explain how users use the system, so it is the most effective and accurate method to analyze and model according to the original data used by the system, that is, logs. The so-called log refers to the traces of the user's operating system, which is generally divided into two categories according to the different perspectives from which the information is recorded: one is based on the user, which we call the operation log, and this kind of log mainly takes the user as the observation entity and records every action of the user from logging in to leaving the system. Another kind of perspective is based on the system, which we call access log. This kind of log is mainly for the application system as the observation entity to record the input and output of the system. The input is each received request information, and the output is the response status of the request.

Based on operation log

Corresponding to the concept of our previous model, when we focus on data such as user volume and business volume, we usually use operation logs for analysis.

This kind of log information is generally stored in the physical table of the system, so most of them can be analyzed through statistical SQL. For example, for each product line within the company, we can use our own UMT gadget to directly count out the key information needed for modeling. For products outside the company, this method is also applicable, because the log information of any product includes at least 4W1H, that is, what user (WHO) and which machine (WHERE) do what (HOW) by accessing which function (WHICH) at what time (WHEN). Statistics for each W or combination, we can roughly get the three elements of information we mentioned earlier.

The disadvantage of this analysis method is that for restoring the accuracy of the user usage model, it mainly depends on the complete preparation of the system operation log record.

Based on access log

When we focus on data such as PV and throughput, we usually analyze access logs, which are generally stored in the log files of middleware or WEB servers.

The data recorded in this kind of log are generally regular data, so it is convenient to use regular statistics to sort out the data. For temporary statistics, you can use tools such as log analyzer to make statistics. There are many such open source gadgets, and the basic principles are all the same. For some systems that need long-term observation, the recommended way is to establish a physical table according to the fields in the log file, write the log records in the log file into the physical table continuously through the shell script, and then analyze them in the physical table. The analysis method is basically the principle of 4W1H.

The analysis scenario is accurate, but because the log is based on the information recorded by the http request. Therefore, in order to establish a highly practical analysis model, we need to be familiar with each page and function of the system, and to establish well-prepared metadata to establish the mapping of pages and functions.

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.