1. How using SQL backend would differ with using Perl for this particular solution (with regard of memory usage)? Given the amout of developers time spending on optimizing DB memory usage I would guess it would be far less than ad hoc solution made in Perl. Wouldn't you agree?
2. I always thought that views are not actual tables and more like indices. In the sense that they are rather collections of pointers to the actual records. Thus, your estimates for the vote_histogram and such are off, IMHO. (To tell the truth I cannot back this statement at all. At least at the moment. If somebody has a good insight how the views are actually constructed, I would really appreciate the pointers!)
3. As already pointed out, WHERE clause should eliminate most of the cases. We are looking for people with large number of votes for each other. Assuming that most of people are not 'bad' - we are left with relatively small number of suspects.
4. As for the efficiency of the actual query evaluation, I think this is kind of once-in-a-while problem that does not require everyday resolutions. I also doubt the quatity/amount of actual data to begin with. How many sites can you name, which have millions of unique users that actively participates in some sort of open forum and actually do vote? Or to this matter how many active topics can this site carry? Also, I'm sure that this data can be split in parts (like historic and actual) and dealt with separately.
Nevertheless, the discussion was really informative. Thank you.
BRIn reply to Re^6: Catching Cheaters and Saving Memory
by caelifer
in thread Catching Cheaters and Saving Memory
by hgolden
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |