PERFORMANCE
The charts use light, medium, and heavy bins to show the distribution of user activity. Each bin represents a group of databases and their metric values. These bins reflect the "bin sizes" values specified in the Server Profile Options dialog box. View the distribution of activity before it is balanced (Current Profile), and then view it again to determine if your goals have been met. Resources that are not well balanced show a disproportionate amount of activity in the heavy bin. After resource balancing has been applied, the recommended distribution in bins should be relatively even across the servers, if your goals were achieved. The higher the accuracy of resource balancing, the more evenly activity is distributed.
Example
The following chart shows database transactions on each server. The overall height of the bar represents the sum (total) of the database transactions. The three bins represent the light, medium, and heavy modal distribution of the database metric -- in this case, transaction. In this example, heavy is the first 30% of databases; middle is the next 40%; and light is the top 30%, all adding up to 100%.
To understand how bin values are calculated, assume there are 20 databases, each with a varying number of transactions. Five is the lowest number of transactions on any database, and 420 is the highest number of transactions on the most active database. The total transactions per database is represented as follows:
Middle = 250,300,310 (3 databases account for 860 transactions; 1154 is the target)
Heavy = 350,400,420 (3 databases account for 1170 transactions; 866 is the target).
When you view these charts, you see that 29% of the chart is light blue; 30% is medium blue; and 40% is dark blue. Hovering over the bar on the chart, the pop-up shows that most transactions on the server occur on relatively few (three) databases. In this case, 15% of the databases account for about 40% of the transactions. If the bars for the other servers on which you are balancing resources have different proportions for light, medium and high bins, then resource balancing would better spread the load across the system and probably result in better server performance.
Related topics