Joining large tables in SQL - sql

Question

I have a table called "calls", there are columns:
a_imei_number
b_imei_number
a_phone_number
b_phone_number
call_start_time
call_end_time
if a specific phone called x calls y then imei number of x is in a_imei_number column if y calls x then imei of x in b_imei_number. Shortly difference between a_imei_number and b_imei_number is incoming and outgoing calls for an imei. same for phone_number columns.
I am searching calls for a specific imei that happen in same time (cloned imei numbers) so i thought if i find a call whose call_start_time between other's call_start_time and call_end_time then i would find the cloned phones. so imei numbers must be same and phone numbers must be different logically.
so i wrote
select * from calls c1 , calls c2
where (c1.a_imei = 1234 or c1.b_imei = 1234)
and
c1.call_start_time between c2.call_start_time and c2.call_end_time
table has maybe 500M data. so this query is not returning and result maybe in 1 week it returns. Is there anyother way to find the result without joining same table like this ?

Answer 1

If I understand correctly, you are looking for calls that occur at the same time as calls to or from a specific number. The following query expresses this idea:
select c2.*
from (select c.*
from calls c
where c.a_imei = 1234 or c.b_imei = 1234
) cbase join
calls c2
on cbase.call_start_time between c2.call_start_time and c2.call_end_time;
The performance is going to depend greatly on the number of matches of the first query.
Sometimes, the database engine has a hard time optimizing or in a condition. I would suggest having indexes on calls(a_imei, call_start_time) and calls(b_imei, call_start_time) and rewriting the query as:
select c2.*
from ((select c.call_start_time
from calls c
where c.a_imei = 1234
) union all
(select c.call_start_time
from calls c
where c.b_imei = 1234
)
) cbase join
calls c2
on cbase.call_start_time between c2.call_start_time and c2.call_end_time;
For the final join, a third index would be useful: calls(call_start_time, call_end_time).

Answer 2

This probably won't help completely, but will hopefully give someone with more knowledge something to start with:
Improving the join
SELECT *
FROM calls c1
INNER JOIN calls c2 ON c1.call_start_time BETWEEN c2.call_start_time AND c2.call_end_time
WHERE (c1.a_imei = 1234 or c1.b_imei = 1234)
Other Comments:
SELECT * will be inefficient as it is, especially as it will be returning non unique column names, you should only select the columns relevant to the query in question.

Answer 3

There are several things you can do to improve your query.
Indexes
It seems that you should have indexes defined on a_imei and b_imei. Perhaps you also would want to include call start and end times in those indexes as well, this depends.
Specify columns
Don't use select *, instead specify the list of columns you want to return.
select
a_imei_number,
b_imei_number,
call_start_time,
call_end_time
Proper Join
This depends on exactly what you are looking for in results. If you want to report on all possible duplicates, you would structure it one way.
select c2.a_imei, c2.b_imei, c2.call_start_time, c2.call_end_time
from (select c.a_imei, c.b_imei, c.call_start_time, c.call_end_time
from calls c
where c.a_imei = c.b_imei
) cbase join
calls c2
on cbase.call_start_time between c2.call_start_time and c2.call_end_time;
If you have a known imei_number and want to search for it, the query would be structured differently.
select c2.a_imei, c2.b_imei, c2.call_start_time, c2.call_end_time
from (select c.a_imei, c.b_imei, c.call_start_time, c.call_end_time
from calls c
where c.a_imei = 1234 or c.b_imei = 1234
) cbase join
calls c2
on cbase.call_start_time between c2.call_start_time and c2.call_end_time;