Severe DML Performance Issue using Oracle RLS on 11.2.0.3

Question

Hi Experts/Gurus,
We are trying to implement Oracle RLS in one our Corporate Application
Purpose of using RLS is to convert Single Company Application into Multiple Company Application
by adding COMPANY_ID Column and Index to make Application Multi-Company.
Each user would be allowed only to view/edit records from his own Company
RLS function we defined is very simple and static.
When we did some DML Tests using 8 parallel sessions, find DML Performance is
severely degraded (response time is twice our existing Application without RLS)
Top 5 Timed Foreground Events
Event Waits Time(s) Avg wait (ms) % DB time Wait Class
db file sequential read 1,312,910 6,197 5 43.08 User I/O DB CPU 2,976 20.69
library cache: mutex X 80,808 1,139 14 7.91 Concurrency
WCR: replay lock order 1,705 1,107 649 7.69 Application
gc current grant 2-way 309,886 550 2 3.83 Cluster
Would like to know why we encounter library cache: mutex X
Checked Mutex Sleep Summary section of AWR and it has following:
Mutex Sleep Summary
ordered by number of sleeps desc
Mutex Type Location Sleeps Wait Time (ms)
Library Cache kglpin1 4 26,849 0
Library Cache kglpnal2 91 25,230 0
Library Cache kglpndl1 95 21,382 0
Library Cache kglpnal1 90 12,823 0
Cursor Pin kksfbc [KKSCHLFSP2] 5,196 0
Would like to know how to interpret
each of the Library Cache entries such as kglpin1 4.
What does the keyword after Libary Cache (starting with kglpin mean)
Thanks in advance

Answer 1

Welcome to the forum!
Please see the FAQ #3 at the top of the thread list for how to post a tuning request and the information you need to provide
SQL and PL/SQL FAQ
>
We are trying to implement Oracle RLS in one our Corporate Application
Purpose of using RLS is to convert Single Company Application into Multiple Company Application
by adding COMPANY_ID Column and Index to make Application Multi-Company.
Each user would be allowed only to view/edit records from his own Company
RLS function we defined is very simple and static.
When we did some DML Tests using 8 parallel sessions, find DML Performance is
severely degraded (response time is twice our existing Application without RLS)
What does the keyword after Libary Cache (starting with kglpin mean)
>
The short answer is that those are pinned objects in the library cache.
The short question is why you are focusing on trivia instead of collecting, and posting, information that might shed some light on your problem?
Read the FAQ thread above and post the information needed.
Your RLS function may be 'very simple and static' but we can't see it. And we can't comment on what we can't see.
We also can't see:
1. your table or index DDL
2. the query you are using that you say is non-performant
3. any data about statistics and whether they are current
4. any data about row counts for the table, the (unknown) query predicates, the expected result set
5. any data about the query predicates
6. any execution plans for the query from the standard table and from the current table
7. any info about the RLS security you are using and how you implemented and configured it.