Whats the best way to deal with big tables like 'thread_views'?

Question

I'm trying to make my mind up to add some statistics to my sites. For example: 'most viewed thread this day/week/year'.

I need a table that saves every view, related to an user (to avoid the same user adding many views) and thread, and with a timestamp.

But that would be one big table (in rows). Is this the way to go?

Answer 1

The answer is dependant on a number of things, such as the number of threads, the number of views, what hardware you have, the typical load, the read/write ratio, the required accuracy etc etc.

A resonable answer to your question without knowing the details of your specific scenario is to create the table you are describing:

thread_views(
   thread_id references thread(thread_id)
  ,user_id   references user(user_id)
  ,timestamp
  ,primary key(thread_id, user_id)
) 

The above approach gives you flexibility and good enough performance for the typical scenario.

I have recently answered two similar questions that you can have a look at to get some ideas.

Count article comments and Count visitor hits per day

A final point is that many of the major databases includes tools for aggregating data. Those tools lets you have a normalized data model while still gaining most of the benefits with a precomputed table of statistics.