I wanted to obtain more metrix in my applications, by the use of my framework, and obtain some interesting results. Basically, every data access goes into the data class. So, that makes it simple to implement such mechanism. The framework knows under which platform it runs, such as Web, Web Service and Desktop. The goal here was to know how many hits we are sending to SQL Server in regards to SQLExec, Update, Insert and Delete. This is basically all we do in our applications, at the exception of a few isolated scripts that are executed very rarely.

So, in my daily statistic table, I added four fields for those four types of data access, a total for them, and did the same for the other two platforms. Then, I added a grand total. So, basically, a site receives 100,000 hits in a day. In here, the average response time is 0.1 to 0.2 second. But, what is going on exactly between those 0.1 and 0.2 second is interesting. As, this is where everything is going against the CPU to SQL Server. So, out of those hits, I have an average of about 20 times of SQLExec, double for Update, about one to one ratio for insert and a few isolated delete commands. So, for the day, I can have something like this:

Web site hits: 100,000
Average response time: 0.1 to 0.2 second
SQLExec: 1.2 millions
Update: 80,000
Insert: 45,000
Delete: 100

So, basically, for processing 100,000 hits, I hit SQL Server about 1.4 million times.

This was in regards to the Web access. As I mentioned above, I have the same thing for the Web Service and the Desktop application, which is a robot processing a lot of things for the Web application at regular intervals. So, totaling all that provides some very good request usage to the backend. This was easy to accomplish because everything goes through the data class. I have seen so many environments where all developers are coding everything for every data access that it becomes a real headache to answer such a request when time comes. In 2011, I would say most environments are now using centralized classes for their data management. Doing it like that thus provides some very good tools such as the one I just described.

Anyone else who did that before? I would be interested in the techniques being used and some numbers.