If u expert Please tell me

Question

i have one doubt
i have 4 table this table have record
a is 11,b is 3,c is 1 and d_city is 0
and field is
a(city_code,name ,dept)
b(dept,dname)
c(dept,salary)
d_city(city_code,city_name)
all table dont have index,con. ect just simple create..
my query is
Select a.city_code,a.name,b.dept,c.salary
From a,b,c,d_city
Where a.dept=b.dept
And a.dept=c.dept
And a.dept=d_city.city_code
in explain plan some time it scant d_city first,some time it scan b first
i want to know how oracle desided which table scan first

Answer 1

The CBO (Cost Based Optimiser) takes a number of things into consideration when deciding how "expensive" a query is in order to arrive at some cost figure.
One of the prime considerations is the amount of I/O to do. I/O is very expensive. Why? Simply because it is such a slow operation - there are inherant latency in I/O. Disk heads have to move and read data, data has be shipped from disk into memory, etc.
So the CBO will attempt to reduce I/O by selecting an execution path where I/O can be minimised. E.g. using an index instead of scanning the entire table.
Why would Oracle choose one table over another table when doing a join? Ideally it would want to use the "smaller" table as that means less I/O.
However, it's decision is based at a certain point in time on the available information at the time. Table sizes can change. Thus an execution plan may not look the same day after day and week after week.
What is important on your part though is enabling the CBO to make the proper decision by supplying it with accurate information. You achieve that by analysing the tables (and indexes).