I am trying to improve a query-function inside an database (db) engine. To measure the performance gain, I calculate the operations per second (ops) between the old and new function.

The ops are depending on the number of entries inside the db and the number of values used in the query.

OPS of new and old function.

As you may can see from the plot: If the number of entries is greater than around 100, the new functions does more ops than the old function.

My goal is: To automatic select one of the two function depending on the values/entries count to perform best.

My current problem is: What mathematical approach do I need, to prove my visual result with real numbers?

My attempt is: To calculate the difference - but what now?

OPS difference between new and old function. Note: The difference is not mainly zero, the z-axis scale is just bad.

My dataset is in the form like this:

entries, values, ops
 1        1       3153480
 1        5       3153480
 5        1       2837753
 1       10       2283749
10        1       1978416
  • $\begingroup$ Why not use the sign of the difference as your decision variable? $\endgroup$ Nov 24, 2016 at 10:17
  • $\begingroup$ Well, this would be a simple solution but my dataset is just a fraction of possible combinations. I do not want an exact solution - more an approximate one. $\endgroup$
    – Viatorus
    Nov 24, 2016 at 11:45


You must log in to answer this question.

Browse other questions tagged .