Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Add title by clicking
HCC – A turbo seldom recognized as such
Implementation Operation
Conceptual design Consulting
1
Nuremberg, 11/18/2015, C. Trieb, P. Böhme
Profile and Aspirations
Personal - Experienced - Efficient
2
60 employees at headquarters in Friedrichsdorf and Munich
Turnover of €11 millions
Industries: retail, logistics, public sector, pharma, banks
Flat hierarchies, open communication, flexibility Comprehensive consulting with a focus on practice Individually optimized solutions
We are a fair, professional partner on par with you
Strong partner network
Product Portfolio
Oracle in Use
3
ORACLE Real Application Clusters
ORACLE Text
ORACLE Data Guard
ORACLE Partitioning ORACLE Enterprise Manager Cloud Control
Oracle Exadata
Oracle ZFS Appliance ZS3-2
Agenda
Overview
6
Paragon‘s architecture Use of HCC HCC in a nutshell HCC Fields of application Summary
Paragon‘s Architecture
Engineered Architecture
7
ADS Gateway Server (Beta)
Storage Server HCC
DB 2 Nodes RAC
Back-up
Standby Server RAC
RZ2 Platinum Support
Gateway Server RZ1
Nagios Server
Cold Data
Data Guard
Paragon Data Exadata
ZS 3
Storage
ZS 3
Storage
HCC
OEM 12c Server
RMAN Repository
ZS 3
Storage
HCC
Use of HCC
Reasons for and Ways of Compression
8
Saving disk space Improving performance through faster I/O Reducing costs Ways of compression:
Index Key Compression BASIC OLTP (Advanced Compression Option) HCC Query Low HCC Query High HCC Archive Low HCC Archive High
HCC in a Nutshell
Hybrid Columnar Compression – When and what for?
9
BASIC -Only for bulk load operations
OLTP (Advanced Compression Option) -To be paid extra Warehouse Compression (optimized for performance): HCC Query Low HCC Query High Archive Compression (optimized for storage savings): HCC Archive Low HCC Archive High Combination of column- and row-based data storage Data sets are organized into logical compression groups and sorted into columns Repeated data is grouped in the same blocks
HCC in a Nutshell
Hybrid Columnar Compression – When and what for
10
Data with similar characteristics Bulk load operations
Parallel DML CTAS SQL*LDR Insert with Append Hint
Many read operations Archived data
HCC in a Nutshell
Compression with HCC
11
HCC Works only with Exadata or ORACLE Storage Not possible with the data type “Long“! Great benefits / high compression ratio only with bulk load operations Lower compression ratio with INSERT / UPDATE Tables are organized into compression units, a compression unit is larger than
an ORACLE block Within compression units, data is organized by column instead of by row =>
similar values together, enhancing compression
Müller Müller Schultz Meier
Müller
Müller
Müller
Schultz
Schultz
Schultz
Meier
Meier
HCC in a Nutshell
Hybrid Columnar Compression Syntax
12
HCC Query Low •Create Table ... column store compress for query low ...; •Alter Table ... column store compress for query low ...;
HCC Query High •Create Table ... column store compress for query high ...; • Alter Table ... column store compress for query high ...;
HCC Archive Low •Create Table ... column store compress for archive low ...; •Alter Table ... column store compress for archive low ...;
HCC Archive High •Create Table ... column store compress for archive high ...; •Alter Table ... column store compress for archive high ...;
HCC in a Nutshell
Hybrid Columnar Compression Syntax
13
In that case the “Alter Table” command only applies to data sets that have been newly inserted afterwards. Therefore the existing data needs to be compressed explicitly: •Alter Table ... move compress for ...;
Undo: •Alter table ... nocompress;
HCC in a Nutshell
How do I know if a table was compressed?
14
Check which type of compression there is: select TABLE_NAME, COMPRESSION, COMPRESS_FOR from user_tables; (or also dba_tables) TABLE_NAME: Description of the object COMPRESSION: DISABLED / ENABLED COMPRESS_FOR: Type of compression, e.g.: BASIC / ADVANCED / QUERY HIGH / ARCHIVE HIGH
HCC in a Nutshell
Compression Overview
15
28
18
14
9 9 8
0
5
10
15
20
25
30
Unkomprimiert OLTP HCC Query Low HCC Query High HCC Archive Low HCC Archive High
GB
100%
64%
50%
32% 32% 29%
0%
20%
40%
60%
80%
100%
120%
Unkomprimiert OLTP HCC Query Low HCC Query High HCC Archive Low HCC Archive High
%
16 16 16 16
HCC in a Nutshell
HCC – Results of a Partitioned Table
16
0
50
100
150
200
250
300
DA
F_A
BH
OLF
AC
H_M
AX
VA
LUES
DA
F_A
BH
OLF
AC
H_8
00
1_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
00
2_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
00
3_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
00
4_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
00
5_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
00
6_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
00
7_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
00
8_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
00
9_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
01
0_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
01
1_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
01
2_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
01
3_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
01
4_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
01
5_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
01
6_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
01
7_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
01
8_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
01
9_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
02
0_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
02
1_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
02
2_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
02
3_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
02
4_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
02
5_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
02
6_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
02
7_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
02
8_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
02
9_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
03
0_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
03
1_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
03
2_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
03
3_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
03
4_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
03
5_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
03
6_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
03
7_
13
_1
3_
13
DA
F_A
BH
OLF
AC
H_8
09
1_
13
_1
3_
13
Unkomprimiert
Komprimiert
Result/ Savings Compression: All: 52% Selective: 70%
17
HCC in a Nutshell
HCC – The Success Story
Bibliography/ catalogue system: 1 master for processing 10 slaves for read applications 3500 users at the same time Very good response times of the application expected Oracle Text full-text search + Oracle snapshots/ materialized views
BUB Master
C1 C2 C3 C4 C5
C6 C7 C8 C9 C0 High effort for hardware and administration!
Consolidation BUB on Exadata? Oracle Text almost only block IO performance? Oracle HCC on Exadata fewer reads PoC
BUB Master
Exadata with 2 BUB
RAC instances
Application without HCC: critical performance with HCC: Performance gain of 30% Less hardware but standby and HCC? Oracle ZS3-2!
HCC Field of Application
What kind of hardware is needed for HCC?
18
Works only on Engineered Systems und ORACLE Storage But without additional licensing fees there
ZFS Storage ZS3-2 Oracle FS1 Flash Storage System Exadata Database Machine
HCC Field of Application
What can HCC achieve?
19
Saving disk space Improving performance => Reducing costs
Pufferlesezugriffe
Unkomprimiert OLTP Query Low Query High Archive Low Archive High
HCC in a Nutshell
Hybrid Columnar Compression Syntax
21
Thank you!
Christian Trieb, [email protected] and Patrick Böhme [email protected]