Vous êtes sur la page 1sur 24

Oracle Exadata:

An Overview

Gaurav Chandra Pandey


AIOUG Bangalore Chapter
5th Oec 2015
Agenda
About Exadata
Exadata Machine XS configuration
Overview Exadata Storage Server Software
Key Features Overview
Exadata Architecture Overview
Managing Exadata using EM12c overview
Traditional Deployment Exadata Deployment
Database Servers
D
D
iiii

Storage Arrays
What is Exadata?

Hardware - Scale-out, Database optimized compute, networking,


and storage for fastest performance and lowest costs

Software - Unique full-stack algorithms for fastest and most


efficient OLTP, Analytics, Consolidation, and ln-Memory DB

lntegration - Full-stack optimization, automation, and support to


reduce operations costs
Two Different Exadata XS Data base Server (Sixth Generation)
(Two Sockets) (Eight Sockets)
Exadata XS-2 Database Server Exadata XS-8 Data base Server
Fastest lntel Haswell Xeon Processor 8 Eighteen-Core lntel Xeon E7-8895 v3
- 2 Eighteen-Core lntel" Xeon ES-2699 v3 Processors (2.6GHz)
Processors (2.3GHz) 2 TB (64 x 32GB) - Expandable to 6TB (192 X
- 256 GB DDR4 (8 x 32GB)- Expandable to 32GB)
768GB (24 X 32GB) 4 x Dual-port 10GbE PCle ports
On-board 4 x 1GbE/10GbE Base-T Ethernet, 2 x 4 x Dual-port lnfiniBand QDR (40 Gb/s) Express
10GbE SFP+ PCle Ethernet card modules
Dual-port lnfiniBand QDR (40 Gb/s) HCA 8 x 2.5" 600GB Disk Orives
4 x 2.5" 600GB 10K RPM Disk Orives Disk Controller HBA with 1 GB Cache
Disk Controller HBA with 1GB Cache Redundant Power Supplies, Fans, ILOM
Redundant Power Supplies, Fan, ILOM
Exadata XS Storage Server Configuration
High Capacity Extreme Flash
2 Eight-Core lntel Xeon ES-2630 v3 2.4 GHz 2 Eight-Core lntel Xeon ES-2630 v3 2.4 GHz
- 96 GB DRAM (4 x 8 GB + 4 x 16 GB) - 64 GB DRAM (8 x 8 GB)
Fastest PC!e Flash Fastest PCle Flash Drives
- 4 x 1.6 TB Flash Accelerator F160 PCle Card - 8 x 1.6 TB 2.5-inch Flash Accelerator F160 PCle
(6.4TB) Orives (12.8TB)
12 x 3.5" 8TB SAS drives - 96 TB Raw Capacity - Hot Swappable

Dual-port lnfiniBand QDR (40Gb/s) HCA- all Dual-port lnfiniBand QDR (40Gb/s) HCA- all ports
ports active active

Disk Controller HBA with 1 GB Cache and Super Redundant Power Supplies, Fans, ILOM
Capacitors
Redundant Power Supplies, Fans, ILOM
Exadata Configurations Full Half Quarter OneEighth
Oatabase Servers 4 2 2
Oracle Exadata Database Machine Full Rack Storage Servers 14 7 3 3
lnfiniband 2 2 2 2
Oracle Exadata Database Machine Half Rack Flash 89.6TB 44.8TB 19.2TB 9.6TB
Extreme Flash 179.2TB 89.6TB 38.4TB 19.2TB
Oracle Exadata Database Machine Quarter Rack Storage Capaclty or
HC Disks 1344TB 672TB 288TB 144TB
Oracle Exadata Database Machine Eighth Rack

Half Rack
XS-2 Elastic Configurations
Optimize Exadata for any Workload
Configuration Examples
DB ln-Memory Extreme Flash OLTP Data Warehousing
Full Rack Machine Machine Machine

Oatabase Server
t:H=fm (iMl!WiZAf"
ffl
et!-
aai
Extreme Flash Storage :

trilee
15 DB Servers 11 DB Servers 8 DB Servers
Startwith Add Servers s Storage Servers 11 Storage Servers 14 Storage Servers
2 Database Servers Any Kind 57608 Cores 39608 Cores SU Cores
3Storage Servers Any Quantity 13.3 TBRAM STBRAM 90 T8 flash Coche
480TB Disk 140TB Flash 1,344 TB Disk
New Oracle Exadata Cloud Service

Single Cloud Service for ALL Database Workloads OLTP

Deploy Mission Critica! Enterprise Databases


with a few clicks Disaster Recovery

100% Database Compatibility enables easy


migration and hybrid deployments Data Warehousing

Oracle Experts Manage lnfrastructure


DB Consolidation
Pay per Use
Complete lsolation of tenants with no
overprovisioning ln-Memory Analytics

App Certification
a
1.. l l
1
-

11-- --

Unique Software Optimizes Database Processing


Query Offloading High Availability
-Data intensive query operations offloaded -Fastest recovery of failed database,
to storage CPUs server, storage or switch
-Storage lndex data skipping -Fastest backup. Incremental offl oad
-exachk t o p -to-bottom validation of
Hybrid Columnar Compression
hardware, software, settings
-Hybrid Columnar for l0x DB size
reduction and faster analytics
Database optimized network
Smart Flash Cache -SQL optimized lnfiniBand protocol for
-Smart caching of database data high throughput low latency SQL
-2.66 Million Database 1Os/sec
-Smart Flash log speeds transactions
Offloading
Exadata storage servers also run more complex
operations in storage (v$sqlfn_metadata)
- Join filtering
- Incremental backup filtering
- 1/0 prioritization
- Storage lndexing

Exadata lntelligent Storage - Database level security


Grid - Offloaded scans on encrypted data
- Data Mining Model Scoring
lOx reduction in data sent to DB servers is common
Traditional Query Processing
o
SELECT 0
customer_name Rows Returned
With traditional storage, all

t
FROM call5 WHERE database intelligence resides
amount > 00;
in the database hosts

l 0
DB Host reduces terabyte of
data to 1000 customer
Very large percentage of data
returned from storage is
Table Extents names that are returned to
discarded by database servers
ldentified client Discarded data consumes
l t

valuable resources, and
@) o impacts the performance of
1/0s lssued 1/0s Executed: other workloads
lterabyte of data
returned to hosts
Storage Processing with Smart Sean
Reduce data transfer
o Only the relevant columns
SELECT customer name
FROM calls WHERE - customer_name and required rows
amount > 00;
0
Rows Returned - where amount>200
are are returned to hosts

1
Extents and metadata
0
Consolidated
CPU consumed by predicate evaluation is
offloaded

Moving sean processing off the database


sent to storage Result Set Built host frees host CPU cycles and eliminates

1
From Ali Cells massive amounts of unproductive

t
messaging
- Returns the needle, not the entire hay stack
smart Sean identifies
rows and columns
e
2MB of data
within terabyte table
returned to server
that match request
Smart Seans

Exadata cells implement smart scans to greatly


reduce the data that needs to be processed by
the database
- Only return relevant rows and columns
- Offload predicate evaluation
Data reduction is usually large
- Column and row reduction often decrease data to be
returned to the data base by lOx
Join filtering
Storage lndex -Anti lndex
Table Region
lndex
11111111
a11
Exadata maintains summary information about table data in
memory
- Store min/max values of columns
- Typically one index entry for MB of disk Min 8=1
Max 8=5
11
Eliminates 10 when MIN/MAX can never match the "where"
clause of the query

Completely automatic and transparent 11


Example: select count(*) from table where B=l; 11 } Min 8=3
Max 8=8
11
Exadata Hybrid Columnar Compression

- Data is organized and compressed by column


- Dramatically better compression

Speed Optimized Query Mode for Data


Warehousing
- -1ox compression typical
- Runs faster because of Exadata offload!

Space Optimized Archiva! Mode for infrequently


accessed data
Faster and s,mpler
Backup, DR, (aching, Rcor, Clone - 15X to SOX compression typical
Exadata 1/0 Resource Management
Mixed Workloads and Multi-Database Environment

Ensure different databases are allocated the


correct relative amount of 1/0 bandwidth
Database A Database B
- Database A: 33% 1/0 resources
- Database B: 67% 1/0 resources

Ensure different users and tasks within a


database are allocated the correct relative
amount of 1/0 bandwidth
- Database A:
Reporting: 60% of 1/0 resources
ETL: 40% of 1/0 resources
- Database B:
lnteractive: 30% of 1/0 resources
Batch: 70% of 1/0 resources
Exadata Storage Server Architecture Overview

Enri-se
r.11an.ager

iDB Prorocol over


lnfnBand wrh
Parh Failover

CeU
Control
Ct..l
CellCLI Overview
Command line utility fer managing cell resources
CellCLI runs en the Exadata storage cell cc.1,!,,ec,,:,_--,;. 1 :.cll
c:"ncu i:,-te,o,..... t ; t ;,..,..1,,.....,.,r, ,..,,., ll'r no,,: in,;,o ,K;i'f in.a

Run locally from a shell prompt C"'J>YT'<l"" ,:_ ,, ;-,11, Oreel.. 'J. qcp,. l""',r,
..d
C:.J.l UC....ci..:.f .:...u.o,c. l .01!:)
Run remotely via ssh or dcli
Run automatically by EM agent with Exadata EM plugin
Can run non-interactively

DCLI Overview
The DCLI script runs commands en multiple cells in parallel threads.
- File copy and command execution occur en a set of cells in parallel.

Command output is collected and displayed after file copy and command execution is finished en all cells.

Setup:
- Copy DCLI from cell (/opt/oracle/cell/cellsrv/bin/dcli) to host from which management is done.
- Create files which contain a list of cells to which commands are issued, e.g. mycells
- Run "dcli-k-g mycells" to create ssh key equivalence en cells
CellCLI Overview
Command line utility fer managing cell resources
CellCLI runs on the Exadata storage cell c,;,1:od:nl..r..@!::.l):.. ..1, .:.e.u
C'tlCU i:-1-.- :t J <' l t P""11 1c,-.1,,,. .,,. l"n C-..<: l Q,".IJH Y.m' <'01.J

Run locally from a shell prompt l;l"'pYTJ"1"'1 o'1.., ;-,


1, Or$C.I'" >J. 1,;p,"''" rer,a
e.u "i.!t...tl..c:v u.. 1.oc1
Run remotely via ssh or dcli
Run automatically by EM agent with Exadata EM plugin
Can run non-interactively

DCLI Overview
The DCLI script runs commands on multiple cells in parallel threads.
- File copy and command execution occur on a set of cells in parallel.

Command output is collected and displayed after file copy and command execution is finished on ali cells.

Setup:
- Copy DCLI from cell (/opt/oracle/cell/cellsrv/bin/dcli) to host from which management is done.
- Create files which contain a list of cells to which commands are issued, e.g. mycells
- Run "dcli -k-g mycells" to create ssh key equivalence on cells
Storage Cell Management

Storage Cell monitoring and
,.,,.._
--- --->, .'--- .
-=.
administration support

-- __ . J . ':
-......,-
....e,
- Cell Home page and performance

--
f
,,
__...,._ ,1..,..9.._. pages

-- -------
---
__,;U

..-......
1,...,._.
- Actions supported: Start/stop Cell,
___
,.._,.,.. __ _.......,._
=
::::_ .-
-:-,:___ ';;:
== ===(

---- " ..-....._,..... ---. -..


._._.T___

-
__
X___
verify connectivity, setup SSH


Automatic discovery of Exadata cells

-.....__
Management by Cell Group

-. ... , ---- - --
!

...
- AII cells used by a database

--
.,., .. .. automatically placed in a group

1 - Cell Group level administration

----
operations (batch job monitoring)

-

Monitoring: ILOM, Cisco Switch, KVM
Common metrics monitored
Power supply failure
Fan failure
Temperature out of range

.,._J h"'-i dot,..,i, ,,rnJ ublt-.mi, ::;.J )b Atli vily

flr"''

Loc.ol torQl'!ltor'ld,dd.edt.,19et, r--1 . ""''""' .. y 10 0 10 > 1 1.> O 1 r- o 1 ForJOhtest.t. <Jet-, wtu,ti..ffl,t'1 d...,,..

I!?" lfl'tlet s..,. tXbtYll!Jllll .,, o


CtA,111 (1f (11.,.t. o ,,.,u,n e,.:..,., FT<ltlli:mseoAlol'I, )( 1


@ l>.',v.,, tl'W"i!l,et'II

,_
N';,11;t


ArtkVI R,eo.-.IE;<OCUUOf,11 .,, O
1 11PY_,__...,( lil o 1Nl..1H'lt 6,d"Yt?MI,.,..
5ch.iledEl/00.lt.DnS O
0.-l!l'II Wi)llr I Cil l,d- I>d4v,a;;;,hco.ll'f Rf:Jcec,;1t10ns o
Coming Sessions ...... .

Session 2 Session 3
Configuring Exadata and administer ASM on Exadata Oracle Exadata Key Features
Exadata Architecture Smart Sean & Offloading
Configure Cell, Celldisk and Griddisk Hybrid Columnar Compression

Configure Flash Exadata Smart Flash

Administer ASM on Exadata IORM

Vous aimerez peut-être aussi