Automatic In-Memory Management (AIM) hands-on
Table of contents
Preamble
Automatic In-Memory Management (AIM) is an extension to the In-Memory Enterprise Edition option that I tested back in 12cR1 (!!). As its name stand for this option put automatically objects in memory based on:
optimizes the column store through actions such evicting cold data, populating frequently accessed data, and compressing cold columns.
Or from product management team:
With Automatic In-Memory Sizing the IM column store can now be managed as part of Automatic Shared Memory Management (ASMM). This means that ASMM can grow or shrink the IM column store based on overall database workload requirements.
The In-Memory option is obviously not free but AIM is free. AIM is also not available with free In-Memory where you memory pool is below 16GB:

My testing has been done with Oracle AI database 26ai (23.26.3.0.0) running on Red Hat Enterprise Linux release 9.7 (Plow). My test server has 12 cores and 64GB of memory. I am obviously testing in a pluggable database as it is now mandatory with 26ai…
I initially tested in 23.26.2.0.0 and I have not been able to see the feature working. In 23.26.3.0.0 the feature works better but still the behavior is quite erratic and you can run the below example multiple times before seeing it working…
Automatic In-Memory Management (AIM) activation
There is a bunch on In-Memory parameters:
SQL> set pages 10000 SQL> show parameter %inmemory% NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ inmemory_adg_enabled boolean TRUE inmemory_automatic_level string OFF inmemory_clause_default string inmemory_deep_vectorization boolean TRUE inmemory_expressions_usage string ENABLE inmemory_force string DEFAULT inmemory_graph_algorithm_execution string DEFAULT inmemory_max_populate_servers integer 0 inmemory_optimized_arithmetic string DISABLE inmemory_optimized_date string DISABLE inmemory_prefer_xmem_memcompress string inmemory_prefer_xmem_priority string inmemory_query string ENABLE inmemory_size big integer 0 inmemory_trickle_repopulate_servers_ integer 1 percent inmemory_virtual_columns string MANUAL inmemory_xmem_size big integer 0 optimizer_inmemory_aware boolean TRUE
To activate AIM I define an In-Memory size (at CDB level) and change the automatic level from OFF to LOW on my pluggable database (then bounce the instance). I also change the parallel policy to benefit from auto parallel execution:
SQL> alter session set container=pdb1; Session altered. SQL> alter system set inmemory_size=1G container=current scope=both; alter system set inmemory_size=1G container=current scope=both * ERROR at line 1: ORA-02097: parameter cannot be modified because specified value is invalid ORA-64913: The Database In-Memory size cannot be increased more than 0 MB for PDB ID 3. Help: https://docs.oracle.com/error-help/db/ora-02097/ SQL> alter system set inmemory_automatic_level=low container=current scope=both; System altered. SQL> alter session set container=cdb$root; Session altered. SQL> alter system set inmemory_size=2G container=current scope=spfile; -- If you set only at PDB level then it is not initialized at container level, the parameter at PDB level is only to distribute the In-Memory CDB quota per PDB System altered. SQL> alter system set parallel_degree_policy=auto scope=both; System altered.
Remark:
Not fully clear to me in the official documentation but if you set INMEMORY_AUTOMATIC_LEVEL to LOW then the MEMCOMPRESS AUTO level of table is not valid. The MEMCOMPRESS AUTO works fine with INMEMORY_AUTOMATIC_LEVEL set to HIGH:
SQL> alter table app01.sales inmemory memcompress auto; Error starting at line : 1 in command - alter table app01.sales inmemory memcompress auto Error report - ORA-00922: missing or invalid option Help: https://docs.oracle.com/error-help/db/ora-00922/
If the INMEMORY_SIZE parameter has been set at CDB level then running statement like below at PDB level is working fine:
SQL> alter system set inmemory_size=1G container=current scope=spfile; System altered.
The “small” problem I have seen with INMEMORY_AUTOMATIC_LEVEL = HIGH is that ALL objects other than owned by Oracle maintained account are by default IM enabled and have MEMCOMPRESS set to AUTO… It means you would have to unflag (NOINMEMORY) all the objects you do not want to go in memory to keep only the focus on the most important ones…
Not very flexible and I would have expected the MEMCOMPRESS AUTO also available with INMEMORY_AUTOMATIC_LEVEL = LOW to flag In-Memory objects of my choice only…
Also, an important point, to have cold segments under memory pressure evicted from Automatic In-Memory, they must be with a PRIORITY equals to NONE…
Automatic In-Memory Management (AIM) test case
I create the same data model I add in my previous testing in APP01 schema. The script for the dimension city table:
drop table if exists city; create table if not exists city ( code number not null, descr varchar2(50), constraint city_pk primary key (code) using index (create unique index city_pk on city (code asc) tablespace users) enable ) tablespace users; insert into city values (1,'Paris'); insert into city values (2,'Lyon'); insert into city values (3,'Marseille'); insert into city values (4,'Bordeaux'); insert into city values (5,'Lille'); insert into city values (6,'Toulouse'); insert into city values (7,'Nantes'); insert into city values (8,'Montpellier'); insert into city values (9,'Dijon'); insert into city values (10,'Arbois'); insert into city values (11,'Nancy'); insert into city values (12,'Troyes'); insert into city values (13,'Bourges'); insert into city values (14,'Tours'); insert into city values (15,'Rennes'); insert into city values (16,'Grenoble'); insert into city values (17,'Annecy'); insert into city values (18,'Rouen'); insert into city values (19,'Brest'); insert into city values (20,'Poitiers'); insert into city values (21,'Metz'); insert into city values (22,'Reims'); insert into city values (23,'Amiens'); insert into city values (24,'Caen'); insert into city values (25,'Le Havre'); insert into city values (26,'Quimper'); insert into city values (27,'Saint-Nazaire'); insert into city values (28,'Carcassonne'); insert into city values (29,'Nimes'); insert into city values (30,'Cannes'); commit; exec dbms_stats.gather_table_stats(ownname => 'app01', tabname => 'city');
And a sales fact tables that is quite big, you are free to tune the parameter of the PL/SQL block to increase the number of rows, I set this table to NOLOGGING to avoid generating too much redo log:
drop table if exists sales; create table if not exists sales(invoice_id number, invoice_date date, city__code number, qty number, val varchar2(100)) nologging tablespace users; declare i number; nbrows number; begin i:=1; nbrows:=10000000; loop exit when i>nbrows; insert into sales values(i, to_date(trunc(dbms_random.value(to_char(date '2000-01-01','j'), to_char(date '2020-01-01','j'))),'j'), round(dbms_random.value(1,30)), round(dbms_random.value(1,10000)), dbms_random.value(1,10000)); if (mod(i,100000) = 0) then commit; end if; i:=i+1; end loop; commit; end; / exec dbms_stats.gather_table_stats(ownname => 'app01', tabname => 'sales'); create table sales2 nologging tablespace users as select * from sales; exec dbms_stats.gather_table_stats(ownname => 'app01', tabname => 'sales2');
The tables have already a good size, in fact a bit below the 1GB of my In-Memory pool but I cannot store them concurrently in memory::
SQL> select segment_name, sum(bytes)/(1024*1024) as "Size MB" from dba_extents where owner='APP01' and segment_name in ('SALES', 'SALES2') group by segment_name;
SEGMENT_NAME Size MB
_______________ __________
SALES 728
SALES2 728Automatic In-Memory Management (AIM) testing
My tables are initially not in memory:
SQL> select
table_name, inmemory, inmemory_priority, inmemory_distribute, inmemory_compression, inmemory_duplicate
from dba_tables
where owner='APP01'
and table_name in ('CITY', 'SALES', 'SALES2')
order by table_name;
TABLE_NAME INMEMORY INMEMORY_PRIORITY INMEMORY_DISTRIBUTE INMEMORY_COMPRESSION INMEMORY_DUPLICATE
_____________ ___________ ____________________ ______________________ _______________________ _____________________
CITY DISABLED
SALES DISABLED
SALES2 DISABLED
SQL> select
segment_name,inmemory_size,bytes,bytes_not_populated,populate_status,inmemory_priority,inmemory_distribute,inmemory_duplicate,inmemory_compression
from v$im_segments
order by segment_name;
no rows selectedAnd the In-Memory pool is empty:
SQL> select * from v$inmemory_area; POOL ALLOC_BYTES USED_BYTES POPULATE_STATUS CON_ID ___________________ ______________ _____________ __________________ _________ 1MB POOL 738197504 0 DONE 3 64KB POOL 318767104 0 DONE 3 SQL> select pool, alloc_bytes, used_bytes, round(used_bytes*100/alloc_bytes) as percentage_used from v$inmemory_area; POOL ALLOC_BYTES USED_BYTES PERCENTAGE_USED ____________ ______________ _____________ __________________ 1MB POOL 746586112 0 0 64KB POOL 318767104 0 0
Testing with INMEMORY_AUTOMATIC_LEVEL = LOW I cannot use the AUTO MEMCOMPRESS so using below and overwriting PRIORITY even if NONE is the default mode:
SQL> alter table app01.sales inmemory memcompress for query low priority none; Table APP01.SALES altered. SQL> alter table app01.sales2 inmemory memcompress for query low priority none; Table APP01.SALES2 altered.
The tables’ properties have changed but before the first select, nothing is in memory pool:
SQL> select
table_name, inmemory, inmemory_priority, inmemory_distribute, inmemory_compression, inmemory_duplicate
from dba_tables
where owner='APP01'
and table_name in ('CITY', 'SALES', 'SALES2')
order by table_name;
OWNER TABLE_NAME INMEMORY INMEMORY_PRIORITY INMEMORY_DISTRIBUTE INMEMORY_COMPRESSION INMEMORY_DUPLICATE
________ _____________ ___________ ____________________ ______________________ _______________________ _____________________
APP01 CITY DISABLED
APP01 SALES ENABLED NONE AUTO FOR QUERY LOW NO DUPLICATE
APP01 SALES2 ENABLED NONE AUTO FOR QUERY LOW NO DUPLICATE
SQL> select segment_name,inmemory_size,bytes,bytes_not_populated,populate_status,inmemory_priority,inmemory_distribute,inmemory_duplicate,inmemory_compression
from v$im_segments
order by segment_name;
no rows selectedI could have used the typical query that would benefit from In-Memory on SALES table. But at the end a simple Full Table Scan (FTS) on my two test tables is giving the expected result. As usual the /* Yannick * / comment is to find myself in V$SQL:
SQL> alter session set statistics_level=all;
Session altered.
SQL> set timing on
SQL> select /* Yannick */ b.descr,sum(qty) as qty, sum(val) as val
from sales a, city b
where a.city__code=b.code
group by b.descr
order by b.descr;
SQL> select /* Yannick */ /*+ full(s) */ count(*) from app01.sales s;The first run is triggering the load in the In-Memory pool and we see that second run is much faster:
SQL> alter session set statistics_level=all; Session altered. SQL> set timing on SQL> select /* Yannick */ /*+ full(s) */ count(*) from app01.sales s; COUNT(*) ___________ 10000000 Elapsed: 00:00:03.376 SQL> select /* Yannick */ /*+ full(s) */ count(*) from app01.sales s; COUNT(*) ___________ 10000000 Elapsed: 00:00:00.184
The SALES table fit perfectly inside the In-Memory pool:
SQL> select
segment_name,inmemory_size,bytes,bytes_not_populated,populate_status,inmemory_priority,inmemory_distribute,inmemory_duplicate,inmemory_compression
from v$im_segments
order by segment_name;
SEGMENT_NAME INMEMORY_SIZE BYTES BYTES_NOT_POPULATED POPULATE_STATUS INMEMORY_PRIORITY INMEMORY_DISTRIBUTE INMEMORY_DUPLICATE INMEMORY_COMPRESSION
_______________ ________________ ____________ ______________________ __________________ ____________________ ______________________ _____________________ _______________________
SALES 723124224 759537664 0 COMPLETED NONE AUTO NO DUPLICATE FOR QUERY LOW
SQL> select pool, alloc_bytes, used_bytes, round(used_bytes*100/alloc_bytes) as percentage_used from v$inmemory_area;
POOL ALLOC_BYTES USED_BYTES PERCENTAGE_USED
____________ ______________ _____________ __________________
1MB POOL 746586112 746586112 100
64KB POOL 318767104 4456448 1The expected FTS plan:
SQL> set pages 10000
SQL> select
sql_id,child_number,plan_hash_value,is_bind_sensitive,is_bind_aware,is_shareable,is_obsolete,sql_plan_baseline
from v$sql
where sql_text like 'select /* Yannick%';
SQL_ID CHILD_NUMBER PLAN_HASH_VALUE IS_BIND_SENSITIVE IS_BIND_AWARE IS_SHAREABLE IS_OBSOLETE SQL_PLAN_BASELINE
________________ _______________ __________________ ____________________ ________________ _______________ ______________ ____________________
0hpzr53am37wx 0 1047182207 N N Y N
SQL> select * from table(dbms_xplan.display_cursor(sql_id => '0hpzr53am37wx', cursor_child_no => '0', format => 'all adaptive allstats'));
PLAN_TABLE_OUTPUT
___________________________________________________________________________________________________________________________________
SQL_ID 0hpzr53am37wx, child number 0
-------------------------------------
select /* Yannick */ /*+ full(s) */ count(*) from app01.sales s
Plan hash value: 1047182207
--------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows | Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers | Reads |
--------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 2 | | 25158 (100)| | 2 |00:00:03.27 | 92478 | 92461 |
| 1 | SORT AGGREGATE | | 2 | 1 | | | 2 |00:00:03.27 | 92478 | 92461 |
| 2 | TABLE ACCESS INMEMORY FULL| SALES | 2 | 10M| 25158 (1)| 00:00:01 | 20M|00:00:03.25 | 92478 | 92461
--------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
1 - SEL$1
2 - SEL$1 / "S"@"SEL$1"
Column Projection Information (identified by operation id):
-----------------------------------------------------------
1 - (#keys=0) COUNT(*)[22]
2 - (rowset=5377)
Hint Report (identified by operation id / Query Block Name / Object Alias):
Total hints for statement: 1
---------------------------------------------------------------------------
2 - SEL$1 / "S"@"SEL$1"
- full(s)
Note
-----
- automatic DOP: Computed Degree of Parallelism is 1 because of no expensive parallel operation
37 rows selected.Now if I execute the same FTS on SALES2 table, the first run will not use in memory but the second will. I can tell it from the execution time (as with the plan you cannot say):
SQL> alter session set statistics_level=all; Session altered. SQL> set timing on SQL> select /* Yannick */ /*+ full(s) */ count(*) from app01.sales2 s; COUNT(*) ___________ 10000000 Elapsed: 00:00:02.641 SQL> select /* Yannick */ /*+ full(s) */ count(*) from app01.sales2 s; COUNT(*) ___________ 10000000 Elapsed: 00:00:00.193
If we check the AIM tasks and latest tasks operations we see that SALES got evicted and SALES2 got populated:
SQL> select * from dba_inmemory_aimtasks order by task_id desc fetch first 5 rows only;
TASK_ID CREATION_TIME STATE
__________ ______________________________ ________
2311 22/07/26 16:40:21,000000000 DONE
2310 22/07/26 16:23:40,000000000 DONE
2309 22/07/26 16:21:39,000000000 DONE
2308 22/07/26 16:19:39,000000000 DONE
2307 22/07/26 16:17:39,000000000 DONE
SQL> select * from dba_inmemory_aimtaskdetails where where object_owner='APP01' and task_id=2311;
TASK_ID OBJECT_OWNER OBJECT_NAME SUBOBJECT_NAME ACTION STATE
__________ _______________ ________________________________ __________________________________________ ____________ ________
2311 APP01 SALES2 NO ACTION DONE
2311 APP01 SALES EVICT DONE
2 rows selected.
SQL> select * from dba_inmemory_aimtaskdetails where object_owner='APP01' and task_id=2310;
TASK_ID OBJECT_OWNER OBJECT_NAME SUBOBJECT_NAME ACTION STATE
__________ _______________ ________________________________ __________________________________________ ____________ ________
2310 APP01 SALES2 POPULATE DONE
2310 APP01 SALES EVICT DONE
2 rows selected.The In-Memory area is now filled only with SALES2:
SQL> select
segment_name,inmemory_size,bytes,bytes_not_populated,populate_status,inmemory_priority,inmemory_distribute,inmemory_duplicate,inmemory_compression
from v$im_segments
order by segment_name;
SEGMENT_NAME INMEMORY_SIZE BYTES BYTES_NOT_POPULATED POPULATE_STATUS INMEMORY_PRIORITY INMEMORY_DISTRIBUTE INMEMORY_DUPLICATE INMEMORY_COMPRESSION
_______________ ________________ ____________ ______________________ __________________ ____________________ ______________________ _____________________ _______________________
SALES2 743440384 757800960 0 COMPLETED NONE AUTO NO DUPLICATE FOR QUERY LOW
SQL> select pool, alloc_bytes, used_bytes, round(used_bytes*100/alloc_bytes) as percentage_used from v$inmemory_area;
POOL ALLOC_BYTES USED_BYTES PERCENTAGE_USED
____________ ______________ _____________ __________________
1MB POOL 746586112 739246080 99
64KB POOL 318767104 4194304 1The explain plan is an TABLE ACCESS INMEMORY FULL. But doing various testing I have discovered that even if the table is not In-Memory then Oracle will say it did a TABLE ACCESS INMEMORY FULL thinking the table could be In-Memory or trying to put it in memory even if the In-Memory pool does not allow it:
SQL> select
sql_id,child_number,plan_hash_value,is_bind_sensitive,is_bind_aware,is_shareable,is_obsolete,sql_plan_baseline
from v$sql
where sql_text LIKE 'select /* Yannick%';
SQL_ID CHILD_NUMBER PLAN_HASH_VALUE IS_BIND_SENSITIVE IS_BIND_AWARE IS_SHAREABLE IS_OBSOLETE SQL_PLAN_BASELINE
________________ _______________ __________________ ____________________ ________________ _______________ ______________ ____________________
fj02mp831tkqp 0 2862189843 N N Y N
0hpzr53am37wx 0 1047182207 N N Y N
SQL> select * from table(dbms_xplan.display_cursor(sql_id => 'fj02mp831tkqp', cursor_child_no => '0', format => 'all adaptive allstats'));
PLAN_TABLE_OUTPUT
____________________________________________________________________________________________________________________________________
SQL_ID fj02mp831tkqp, child number 0
-------------------------------------
select /* Yannick */ /*+ full(s) */ count(*) from app01.sales2 s
Plan hash value: 2862189843
---------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows | Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers | Reads |
---------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | 25227 (100)| | 1 |00:00:02.55 | 92510 | 92505 |
| 1 | SORT AGGREGATE | | 1 | 1 | | | 1 |00:00:02.55 | 92510 | 92505 |
| 2 | TABLE ACCESS INMEMORY FULL| SALES2 | 1 | 10M| 25227 (1)| 00:00:01 | 10M|00:00:02.54 | 92510 | 92505 |
---------------------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
1 - SEL$1
2 - SEL$1 / "S"@"SEL$1"
Column Projection Information (identified by operation id):
-----------------------------------------------------------
1 - (#keys=0) COUNT(*)[22]
2 - (rowset=5377)
Hint Report (identified by operation id / Query Block Name / Object Alias):
Total hints for statement: 1
---------------------------------------------------------------------------
2 - SEL$1 / "S"@"SEL$1"
- full(s)
Note
-----
- automatic DOP: Computed Degree of Parallelism is 1 because of no expensive parallel operation
37 rows selected.I have also tried to plan with DBMS_AUTOIM.ACTIVITY_REPORT procedure, but it did not display anything:
SQL> set pages 10000 SQL> set long 32000 SQL> select dbms_autoim.activity_report(start_time => sysdate -1, end_time => sysdate, type => 'TEXT', level => 'DETAILED') as aim_report from dual; AIM_REPORT __________________________________________________________________________________ REPORT SUMMARY ------------------------------------------------------------------------------- Start time : 23-JUL-2026 10:36:51 End time : 24-JUL-2026 10:36:51 No. of times auto task scheduled : 3 Statements Analyzed : 0 IM Performance Candidates Identified : 0 Statements Verified : 0 IM Performance Candidates Accepted : -------------------------------------------------------------------------------
Automatic In-Memory Management (AIM) with Automatic Data Optimization (ADO)
You can club ADO policies with AIM to fine tune the behavior of you IM area, the followed algorithm is described in Oracle official documentation:

SQL> alter table app01.sales ilm add policy no inmemory after 7 days of no access; Table APP01.SALES altered.
You can check the ILM tables to see the implemented ADO policies:
SQL> select * from dba_ilmpolicies; POLICY_NAME POLICY_TYPE TABLESPACE ENABLED DELETED ______________ ________________ _____________ __________ __________ P1 DATA MOVEMENT YES NO SQL> select * from dba_ilmdatamovementpolicies; POLICY_NAME ACTION_TYPE SCOPE COMPRESSION_LEVEL TIER_TABLESPACE TIER_STATUS CONDITION_TYPE CONDITION_DAYS CUSTOM_FUNCTION POLICY_SUBTYPE ACTION_CLAUSE TIER_TO ______________ ______________ __________ ____________________ __________________ ______________ ___________________ _________________ __________________ _________________ ________________ __________ P1 EVICT SEGMENT LAST ACCESS TIME 7 INMEMORY no inmemory
You can also more compress objects after a period of no DML, this would more apply with partitioning with a rolling strategy of current partitions (per day, week or month):
SQL> alter table app01.sales ilm add policy modify inmemory memcompress for query high after 2 days of no modification; Table APP01.SALES altered. SQL> select * from dba_ilmpolicies; POLICY_NAME POLICY_TYPE TABLESPACE ENABLED DELETED ______________ ________________ _____________ __________ __________ P1 DATA MOVEMENT YES NO P2 DATA MOVEMENT YES NO SQL> select * from dba_ilmdatamovementpolicies; POLICY_NAME ACTION_TYPE SCOPE COMPRESSION_LEVEL TIER_TABLESPACE TIER_STATUS CONDITION_TYPE CONDITION_DAYS CUSTOM_FUNCTION POLICY_SUBTYPE ACTION_CLAUSE TIER_TO ______________ ______________ __________ _____________________________ __________________ ______________ _________________________ _________________ __________________ _________________ _______________________________________ __________ P1 EVICT SEGMENT LAST ACCESS TIME 7 INMEMORY no inmemory P2 COMPRESSION SEGMENT MEMCOMPRESS FOR QUERY HIGH LAST MODIFICATION TIME 2 INMEMORY inmemory memcompress for query high
Issues encountered
Automatic In-Memory task running forever
The AIM task got stuck in RUNNING state for more than a day:
SQL> select * from dba_inmemory_aimtasks order by task_id desc fetch first 5 rows only;
TASK_ID CREATION_TIME STATE
__________ ______________________________ __________
2311 22/07/26 16:40:21,000000000 RUNNING
2310 22/07/26 16:23:40,000000000 DONE
2309 22/07/26 16:21:39,000000000 DONE
2308 22/07/26 16:19:39,000000000 DONE
2307 22/07/26 16:17:39,000000000 DONEI have tried to stop/start the pluggable database and even the whole instance but it solved nothing. I unlocked the situation with a bunch of INMEMORY_AUTOMATIC_LEVEL value change:
SQL> alter system set inmemory_automatic_level=off scope=memory; System altered. SQL> alter system set inmemory_automatic_level=high scope=memory; System altered. SQL> alter system set inmemory_automatic_level=low scope=memory; System altered.
And finally:
SQL> select * from dba_inmemory_aimtasks order by task_id desc fetch first 5 rows only;
TASK_ID CREATION_TIME STATE
__________ ______________________________ __________
2332 23/07/26 11:05:31,000000000 DONE
2311 22/07/26 16:40:21,000000000 RUNNING
2310 22/07/26 16:23:40,000000000 DONE
2309 22/07/26 16:21:39,000000000 DONE
2308 22/07/26 16:19:39,000000000 DONEReferences
- New Database In-Memory Features in Oracle AI Database 26ai
- Automatic In-Memory
- Automatic Data Optimization for Database In-Memory
- When and How to use Oracle Database In-Memory Advisor





