HI Lila.
in Persian, we call you : Leila (لیلا)
lets to talk about sqlqlchemy instead elixir.
SA is so fast in relationships but i suggest to test some situations,
and put the results in this group for all members.
test your case in two phases:
1.normalized db design(
http://en.wikipedia.org/wiki/
Database_normalization):
2.denormalized db design(
http://en.wikipedia.org/wiki/Denormalization)
i suggest test this case with approximately 900,000,000 records in
each table.
after ensure your architecture, (one of mentioned db design
architectures: NORMALIZATION or DENORMALIZATION)
,try to optimize your case with indexes and all other options depends
on witch RDBMS backend you are using with elixir(sqlalchemy).such as
oracle:select optimization
NOTE: SA has a powerful caching mechanism in memory.and elixir can
load relationships in two modes: lazy or not lazy, see elixir
documentation. but the performance is depends on your RDBMS, i
recomend Postgresql, open source, powerful, very very very fast in
linux.
######################################################################
Some notes about my project:
my db design is so complex and usually i design databases strongly
normalized!.
my records are over 1,000,000,000
RMBMS: postgresql-8.4
OS:Ubuntu 10.4 server
Processor:Quad core ....
Memory 4G
Average requests per minutes: 1500
AND Everything is so good.
#######################################################################