Currently I've been assigned the task to research ways to speed up an existing project. It's going to be rebuild so we can start from scratch, the current implementation has been build by another company and is lacking in performance.
The data isn't all that big except for one table, which is going to grow quite rapidly. I've been looking into sharding and partitioning tables and I can find enough info regading those topics.
Another possible solution would be lowering resolution of data gather for old data.
So for example, the table would contain answer of employees, these employees anser 1-30 questions a day, answers have a numerical part over which averages are made.
I could store only the averages for data of a couple of years old, ie average per month, for data of a couple of years old.
However I can't find any info because I don't know what to google for, does anyone recognize this pattern, or know if this solution is already flawed for whatever reason.