Index Search Performance

Here is my problem: Let's say I had a table A (a1 integer,a2 integer, ....ak integer) and Table B (b1 integer,b2 integer,...,bm integer) with non-unique indexes that consists of (a1,...,ak) and (b1,...,bm) on each tables. Page size for each table is 2K. And I'd like to perform such a select: Select * from A,B where a1=b1 and a2=b2 and ... ai=bi my questions is: 1. What will be the statisticscal (if that select will be performed to much times) compexity of such operation? 2. What will be the statistical complexity of such operation if table A has X% of appropriate rows and B table has Y% of approprotate rows ?. 3. How this complexity depends of SQL plan ? Its seems very worst problem in informix cause all the time optimizer set the smallest table before the biggest. Sincerely, Alexander