COLLECTION REAL PURPOSE AND BENEFITS

Question

Hi, Kindly anyone can give the real reason behind the existance of Collections in Oracle.
Because all things we can do through database tables/views/procedures etc.
Please help me to understand and give me some idea where use of collections become important.
Thanks in advance.

Answer 1

ALI wrote:
One my query is returning results very slow because it is joining four tables A,B,C and D. and each of three tables A,B and C is having around 1 million rows and D table is having less rows around 100000 rows.
So I was thinking that may be some Collection should be used to load data in the data block of the oracle form. for example, by using bulk collect and limit operators.
Any idea ?How will this make the query faster?
By putting that data into memory? That is what the SQL engine also does, using the database's buffer cache. And it does it with far more sophistication and better utilisation of memory.
Keep in mind that your collection sits in the most expensive server memory in Oracle - the PGA. This is private process memory. And unlike the buffer cache that is memory shared by all, PGA memory can only belong to a single process and be used by that process.
How do you get that data into the collection (memory)? Using I/O to read the data from table (disk).
And this is also exactly what the SQL engine will do.
So how does your approach do it better and faster than the SQL engine?
And who says that the SQL engine is slow? It is actually very powerful and quite capable. And able to apply more sophisticated data processing algorithms, then many developers can hope to code themselves.
What makes it slow? The same thing that makes any s/w slow. I/O. Reading data from disk is a slow and expensive operation.
That is why a RDBMS has features like index organised tables, partition tables, table clusters, indexes and so on - as these allow one to reduce I/O and thus increase performance.
Performance is not governed by the number of rows or size of the table. It is governed by how much I/O is needed to be used to process that data and deliver the results you want. And the following SQL clearly demonstrates that:
SQL> set timing on
SQL> select count(*) from daily_xxxx;
  COUNT(*)
5298194698
Elapsed: 00:00:41.055.29 billion rows counted by that SQL statement in just over 40 seconds. Several 100 million rows counted per second.
How? Because it uses local partition bitmap indexes to determine the number of rows. Tiny indexes that means little I/O - in comparison to actually reading each and every row in the table and counting them.
Can you design and code better then what the SQL engine is capable of? Very unlikely.
Instead, learn how to make SQL part of your development arsenal and use it. It is an awesome weapon for crunching data - when wielded correctly.
You need to implement your data model correctly in the database's physical layer. You need to provide the SQL engine's CBO with the data it needs, in order for it to choose the optimal data processing method.