Performance test load model (v)

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

Raw Data analysis method

Business scenario modeling is mainly about how users use the system, so it is the most effective and accurate method to analyze and model the original data used by the system, i.e. log. The so-called log refers to the user's operating system traces, according to the different perspectives of the recording information, generally divided into two categories: a kind of perspective is based on the user, we call the operation log, this kind of log mainly to the user as an observation entity, recording the user from the system to leave the system of each action, another perspective based on system, We call it the access log, which is the main application system for observing entities, recording system input and output. The input is the request information received by each receiving, the output is the response status of the request.

based on the operation log

In response to the concept of the model we presented earlier, when we focus on the load of data such as user volume, business volume, we usually use the operations log for analysis.

This type of log information is generally stored in the system's physical tables, so most can be statistically SQL for analysis, such as individual product lines within a company, can be used to directly count the critical information needed for modeling using our own UMT gadgets. This method is also applicable for products outside the company, because the log information for any product includes at least 4w1h,which is what user (who) is at what time (when) by which machine (where By accessing which feature (which) does what (how), for each W or combination of statistics, you can generally get the three elements of the information we mentioned above.

The disadvantage of this analysis method is that it relies primarily on the complete preparation of system operation log records for restoring the user's accuracy of using the model.

 based on access logs

when we focus on the load is mainly PV, throughput This type of data is typically analyzed for access logs, which are generally stored in the log files of the middleware or WEB server.

This kind of logging data is generally the rules of the data, so it is easy to use the regular data collated after the statistics, for the temporary statistics, you can use similar to log Analyzer such tools statistics, such open-source gadgets are many, The rationale is the same, and for some systems that require long-term observation, the recommended way is to create a physical table based on the fields in the log file, and then write the log records in the log file to a physical table through a shell script, and then analyze them in the physical table. The analysis method is basically also 4w1h principle.

The analysis scenario is highly accurate, but because the log is information that is logged based on the HTTP request. Therefore, to establish a practical analytical model, it is necessary to be familiar with the various pages and functions of the system, and to establish a complete set of metadata to establish the mapping of pages and functions.

This article from the "Performance Test" blog, reproduced please contact the author!

Performance test load model (v)