Postam še malo interne diskusije o nosql, žal v angleščini, mogoče ti odgovori na katera vprašanja.
--
Ok so - statistics fast and relational databases are stuff don't go well together, so if we want effective statistic browsing and storing, we need to get rid of relational databases (read: get rid of *SQL servers).
Relational databases by design are not good for these kind of things, and therefore should not be used because there are alternatives that are 10 times more suited for this.
Alternative is using mongo db. See following URLs:
http://en.wikipedia.org/wiki/NoSQL
http://en.wikipedia.org/wiki/MongoDB
http://www.mongodb.org/display/DOCS/Tutorial
http://www.mongodb.org/display/DOCS/MapReduce
http://www.mongodb.org/display/DOCS/SQL+to+Mongo+Mapping+Chart
Which of course, would require rewriting part of the app to use it, instead of sql backend. Redesign of database also, since mongo db and generally non rdbms dbs are working quite differently.
I personally didn't use mongo db ever, I used pytables which is a python thing similar to this, 3 dimensional databases which are suited for things like statistics or whatever else which have huge amounts of data.
For example, in pytables i can do a complex query on tables with millions of rows and get data in less then a second, i'm thinking mongo db does similar thing, and if we want fast stats we need to get rid of the "standard" way.
Some nosql blog http://nosql.mypopescu.com/ ..
proposed to use mongoDB for this (he will install and configure when we need it), but as mongo represents noSQL and nonrelational database system, current stats will not work also so a whole application should be changed
Current mysql is 38.8GB one table and 8GB second table
but we discovered that nosql solution is no good
cause we would need to redefine complete data model
which is not wise
so we said that we will just optimize sql for you
Eden od razlogov za opustitev nosql je bil tudi, da je bila obstoječa aplikacija v sql in da je imela veliko podatkov. Če bi bil development denovo, bi bil mogoče outcome drugačen.
S tem problemom so se par tednov ukvarjali zelo izkušeni programerji in server admini in se potem odločili da nosql no go.