Question regarding Memory Consumption in Collections

Question

I Read this on some website -
Memory for collections is stored in the program global area (PGA), not the system global area (SGA). SGA memory is shared by all sessions connected to Oracle Database, but PGA memory is allocated for each session. Thus, if a program requires 5MB of memory to populate a collection and there are 100 simultaneous connections, that program causes the consumption of 500MB of PGA memory, in addition to the memory allocated to the SGA.
My Question is -
If i use Collections in my Procedures and they consume 5 MB space, is the memory still allocated to the session even if the Procedure completes its execution ??
I mean is memory released for other sessions after procedure completion or session termination ??

Answer 1

user9276238 wrote:
If i use Collections in my Procedures and they consume 5 MB space, is the memory still allocated to the session even if the Procedure completes its execution ??Yes. How does the session know that your very next instruction to it is not run that very same procedure again (with different parameters). It will again need 5MB of memory. And if releases, it will need to reallocate it. If all processes work like this, immediately releasing memory and then grabbing memory again - that could cause contention and other issues.
Memory management is more complex than this. For example, let's say your session required 1 few KB of space. So the process's memory looks something as follows:
xxxx
Next your session needs that 5MB for the (enormous) collection:
xxxxyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy
And during the processing, it needs a few KB more:
xxxxyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyzz
Okay, so now it can release the 5MB memory (as indicated by the y ) back to the kernel? It cannot shrink it as there is trailing memory allocated and used (as indicated by the z ). It will need to re-org that memory in order to release that 5MB worth of y's.
So it is not as straight forward as one would like to think. The basic rule for PL/SQL is to tread softly using memory. Why? Because this is server code. Memory used in PL is allocated as private memory to that process. Unlike shared memory, no-one else can share in its use. Thus the memory must be considered expensive.
There's also the issue of scalability. If your PL code needs 5MB of private memory from the server.. what happens when a 100 client sessions all run your code? How well does your code scale in a multi-process and multi-server environment?
So what memory should you ideally then use in PL? Shared memory. The SGA. This means it is better for scalability and server resources to rather store large data structures for PL as GTT's (global temporary tables - where a temp table is created per session).
I mean is memory released for other sessions after procedure completion or session termination ??Session termination is a "sure thing" ito memory being released back to the kernel as free memory - assuming this is a dedicated server session. The actual server process will terminate and its PGA will be released. Shared server processes services multiple sessions during their life time. And only when the the shared server process terminates, will its PGA be released.(though PGAs can be shrunk as and when decided feasible by their memory managers)