oracle 12c IN-Memory Performance study on Dell Infrastructure

oracle 12c IN-Memory Performance study on Dell Infrastructure


Applies to:
Oracle database – 12.1.0.2

Case Study:

Oracle 12c In-memory option is introduced in 12.1.0.2 patch set. This feature stores copies of tables and partitions, and other database objects in columnar format which is IM Column store. IM Column store is an optional area of the SGA and it is not replacement of buffer cache. Instead, both memory areas can store same data in different formats.

IM COLUMN Store performance Benefits:

If the database objects in IM Column store, database can more perform scans, joins and aggregates much faster than on disk. When In memory will give more performance

  • Querying small number of columns from large number of columns
  • Queries that scan large number of rows and applying filters
  • Queries that apply aggregate data
  • Queries that join a small table to large table(fact table)

Study 1:

Oracle 12c In Memory option tested on Dell infrastructure.

Operating system: RHEL6.5

Database: 12.1.0.2

Memory: 512G

SGA Size: 450G

In Memory size: 300G

Database Size: 1TB

I am enabling the inmemory option using inmemory_size parameter.

Sql> alter system set inmemory_size=300g scope=spfile;

It will effect only after instance restart.

We can find the size of inmemory using

SQL> show parameter inmemory;

NAME TYPE VALUE
inmemory_clause_default string
inmemory_force stringstring DEFAULT
inmemory_max_populate_servers integer 30
inmemory_query string ENABLE
inmemory_size big integer 300G
inmemory_trickle_repopulate_servers_percent integer 10
optimizer_inmemory_aware Boolean TRUE

Table1: SQL Parameter inmemory


After that create tablespace using In Memory attribute

Sql> create tablespace quest datafile '+DATA’ size 10g default inmemory;

We can change any time inmemory parameters using alter statement

Sql> alter tablespace quest DEFAULT INMEMORY MEMCOMPRESS FOR CAPACITY HIGH;


After you can create any tables under this tablespace it will automatically store into IM Column store.

Using BMF7.0 we can load the data into particular tablespace. If we enable inmemory parameter at tablespace level before loading it will take less time compare to without inmemory attribute.

Please refer this link

http://www.toadworld.com/platforms/oracle/b/weblog/archive/2014/09/12/benchmarking-oracle-12c-in-memory-column-store.aspx

After loading data, we can use the different compression techniques and priority levels to populate the data into IM Column store. Objects are populated into the IM column store either in a prioritized list immediately after the database is opened or after they are scanned (queried) for the first time.

Here are my runtime results against Dell Infra structure.

We can compress and populate the data into IM Column store using different methods.

The table h_partsupp which has 59.59 GB original data. We can use different compression techniques to populate the table objects into IM Column store and see the compression ratio.

Compression Techniques ratio Compressed Data
Memcompress for query hight 1.29 31.08
Memcompress for query hightMemcompress for query hightMemcompress for query hightMemcompress for query hightMemcompress for query hightMemcompress for query hightMemcompress for query hightMemcompress for query hightMemcompress for query hightMemcompress for query Low 1.67 35.64
Memcompress for Capacity low 3.04 19.63
Memcompress for capacity high 4.88 12.21
No MemCompress 1.17 50.86

Table 2: Compression Techniques Ratio

After that we can raise a query against the table and see the results of performance of inmemory and without In Memory.

Sql> select max (PS_SUPPLYCOST) from h_partsupp;

Compression Technique Sec
Memcompress for query high 15.97
Memcompress for query Low 18
Memcompress for Capacity low 18.13
Memcompress for capacity high 99.86
No Inmemory 2961

Table 3: Compression Techniques Time

As per results, In Memory is giving more performance (185X) compare to without In Memory.

Study 2 :

The table h_part which has 29.27Gb original data. We can use different compression techniques to populate the table objects into IM Column store and see the compression ratio.

Compression Techniques ratio compressed Data
Memcompress for query high 3.32 8.43
Memcompress for query Low 2.77 10.09
Memcompress for Capacity low 5.77 4.85
Memcompress for capacity high 7.96 3.51
NO MemCpmpress 1.24 22.49

Table 4: Compression Techniques ratio

After that we can raise a query against the table and see the results of performance of inmemory and without In Memory.

Sql> select P_NAME from quest.h_part where P_TYPE='SMALL BRUSHED NICKEL';

Compression Techniques sec
Memcompress for query high 6.04
Memcompress for query Low 6.26
Memcompress for Capacity low 8.86
Memcompress for Capacity high 19
No inmemory 30
NO MemCompress 9.10
Table 5:Compression Techniques time



Article ID: SLN312264

Last Date Modified: 08/14/2018 01:31 AM


Rate this article

Accurate
Useful
Easy to understand
Was this article helpful?
Yes No
Send us feedback
Comments cannot contain these special characters: <>()\
Sorry, our feedback system is currently down. Please try again later.

Thank you for your feedback.