Documente Academic
Documente Profesional
Documente Cultură
Databases
Exadata Architecture
Complete Database platform using standard servers for Compute and Storage
Scale-Out Database Servers
8x 2-socket, or 2x 8-socket Xeon database servers
Oracle Database, ASM, RAC; Linux or Solaris
Standard Ethernet to data center
Scale-Out Intelligent Storage Servers
2-socket storage servers, Exadata Storage Software
Up to 500 terabytes disk per rack
56 PCI Flash memory cards per rack
InfiniBand Network
Unified internal connectivity ( 40 Gb/sec )
Time
(Days)
7
Custom Configuration
Performance Achievement
Performance Achievement
100%
Assemble
dozens of
components
Measure,
diagnose,
tune and
reconfigure
Time
(Months)
Multivendor
finger
pointing
X3
2009
2008
Warehouse
Scale-Up
Massive Flash
All I/Os to Flash
Database In-Memory
Exadata Hardware
10
11
Exadata X3-2
14
Exadata X3-8
Storage Expansion
Lower
Cost
Faster than
2010 Quarter Rack,
and 2008 Half Rack
15
Eighth Rack
Quarter Rack
Half Rack
Full Rack
X2-2
Qtr to Half Upgrade
in 2011
V2
Initial Quarter Rack
deployed in 2010
18
19
X3-2
Half Rack
X3-2
Quarter
X3-2
Eighth
Disk Data
Bandwidth1,3
25 GB/s
12.5 GB/s
5.4 GB/s
2.7 GB/s
18 GB/s
9 GB/s
4 GB/s
2 GB/s
Flash Cache
Data Bandwidth1,3
100 GB/s
50 GB/s
21.5 GB/s
10.7 GB/s
93 GB/s
46.5 GB/s
20 GB/s
10 GB/s
50,000
25,000
10,800
5,400
28,000
14,000
6,000
3,000
8K Reads
1,500,000
750,000
375,000
187,000
8K Writes
1,000,000
500,000
250,000
125,000
16 TB/hr
8 TB/hr
4 TB/hr
2 TB/hr
Disk IOPS
Flash IOPS2,3
Data Load Rate4
1 - Bandwidth is peak physical scan bandwidth achieved running SQL, assuming no compression. Effective data bandwidth will be much higher when
compression is factored in.
2 - IOPS Based on read IO requests of size 8K running SQL. Note that the IO size greatly effects flash IOPS. Others quote IOPS based on 2K, 4K or
smaller IOs that are not relevant for databases. Exadata Flash read IOPS are so high they are typically limited by database server CPU, not IO.
3- Actual Performance varies by application.
4 Exadata load rates are typically limited by database server CPU, not IO. Rates vary based on load method, indexes, data types, compression, and
partitioning
20
X3-8 or X3-2
Full
X3-2
Half
X3-2
Quarter
X3-2
Eighth
100 TB
50 TB
21.6 TB
10.8 TB
504 TB
252 TB
108 TB
54 TB
22.4 TB
11.2 TB
4.8 TB
2.4 TB
45 TB
22.5 TB
9.5 TB
4.5 TB
224 TB
112 TB
48 TB
23 TB
Usable Triple
Mirrored Capacity1,3
30 TB
15 TB
6.5 TB
3.25 TB
150 TB
75 TB
32 TB
16 TB
1- Raw Disk Capacity defined using standard disk drive terminology of 1 TB = 1000 * 1000 * 1000 * 1000 bytes.
2- Capacity calculated using normal space terminology of 1 TB = 1024 * 1024 * 1024 * 1024 bytes.
3 - Actual space available for a database after mirroring (ASM normal redundancy) and allowing one disk (Quarter
and Half) or two disks (Full Rack) of free space to automatically remirror after disk failures.
4 - Actual space available for the database computed after triple mirroring (ASM high redundancy).
21
X3-2
Half Rack
X3-2
Quarter
X3-2
Single Cell
Disk Data
Bandwidth1,3
32 GB/s
16 GB/s
7.2 GB/s
1.8 GB/s
23 GB/s
11.5 GB/s
5 GB/s
1.3 GB/s
Flash Cache
Data Bandwidth1,3
130 GB/s
65 GB/s
29 GB/s
7.25 GB/s
121 GB/s
60.5 GB/s
27 GB/s
6.75 GB/s
64,000
32,000
14,400
3,600
36,000
18,000
8,000
2,000
8K Reads
1,900,000
950,000
425,000
107,000
8K Writes
1,350,000
675,000
300,000
75,000
Disk IOPS
Flash IOPS2,3
1 - Bandwidth is peak physical scan bandwidth achieved running SQL, assuming no compression. Effective data bandwidth will be much higher when
compression is factored in.
2 - IOPS Based on read IO requests of size 8K running SQL. Note that the IO size greatly effects flash IOPS. Others quote IOPS based on 2K, 4K or
smaller IOs that are not relevant for databases. Exadata Flash read IOPS are so high they are typically limited by database server CPU, not IO. This is
especially true for expansion racks.
3- Actual Performance varies by application.
22
X3-2
Full
X3-2
Half
X3-2
Quarter
X3-2
Single Cell
128 TB
64 TB
28 TB
7.2 TB
648 TB
324 TB
144 TB
36 TB
28.8 TB
14.4 TB
6.4 TB
1.6 TB
58 TB
29 TB
13 TB
3.25 TB
288 TB
144 TB
64 TB
16 TB
Usable Triple
Mirrored Capacity1,3
38 TB
19 TB
8.5TB
2.1 TB
192 TB
96 TB
43 TB
10.75 TB
1- Raw Disk Capacity defined using standard disk drive terminology of 1 TB = 1000 * 1000 * 1000 * 1000 bytes.
2- Capacity calculated using normal space terminology of 1 TB = 1024 * 1024 * 1024 * 1024 bytes.
3 - Actual space available for a database after mirroring (ASM normal redundancy) and allowing one disk (Quarter
and Half) or two disks (Full Rack) of free space to automatically remirror after disk failures.
4 - Actual space available for the database computed after triple mirroring (ASM high redundancy).
23
X3-2 Full
X3-2 Half
X3-2 Quarter
X3-2 Eighth
160
128
64
32
16
4096
2048
1024
512
512)
InfiniBand switches
Ethernet switch
14
14
168
168
84
36
18
22.4 TB
22.4 TB
11.2 TB
4.8 TB
2.4 TB
High Perf
100 TB
100 TB
50 TB
21.6 TB
10.8 TB
High Cap
504 TB
504 TB
252 TB
108 TB
54 TB
High Perf
45 TB
45 TB
22.5 TB
9.5 TB
4.5 TB
High Cap
224 TB
224 TB
112 TB
48 TB
23 TB
High Perf
30 TB
30 TB
15 TB
6.5 TB
3.25 TB
High Cap
150 TB
150 TB
75 TB
32 TB
16 TB
Database Servers
24
X3-2 Half
X3-2 Quarter
Single Cell
InfiniBand switches
Ethernet switch
18
216
108
48
12
28.8 TB
14.4 TB
6.4 TB
1.6 TB
High Perf
128 TB
64 TB
28 TB
7.2 TB
High Cap
648 TB
324 TB
144 TB
36 TB
Usable mirrored
capacity
High Perf
58 TB
29 TB
13 TB
3.25 TB
High Cap
288 TB
144 TB
64 TB
16 TB
High Perf
38 TB
19 TB
8.5TB
2.1 TB
High Cap
192 TB
96 TB
43 TB
10.75 TB
25
X2
X3
2008
2009
2010
2012
336
504
504
3X
5.3
5.3
22.4
4X
64
64
96
128
2X
576
1152
2048
8X
24
184
400
400
50X
Flash (TB)
CPU (Cores)
Connectivity (Gb/s)
26
V1
88
Exadata Innovations
27
Exadata Innovations
Hybrid Columnar Compression
Intelligent storage
Scale-out InfiniBand storage
Smart Scan query offload
uncompressed
Data remains
compressed
for scans
and in Flash
Benefits Cascade
to Copies
28
compress
primary DB
standby
test
dev
backup
What were
my sales
yesterday?
Exadata
Storage
Grid
Optimizer Chooses
Partitions & Indexes
to Access
partitions / indexes
10 TB scanned
1 GB returned
to servers
30
operations in storage
Join filtering
Exadata Intelligent
Storage Grid
is common
31
Database Servers
Perform complex database processing such as
Data Intensive
Processing
32
Index
A B C D
1
3
Min B = 1
Max B =5
5
5
Min B = 3
Max B =8 Completely automatic and transparent
3
Select * from Table where B<2 - Only first set of rows can match
33
Order_Date
Partitioning Column
Ship_Date
2007
2007
2008
2008
2009
2009
Item
Order#
Takes advantage of ordering created by partitioning or sorted loading
34
Dimension
Name
A
Perform IO and
apply bloom filter
Accord 1
3
Civic
Prius
M C D
Camry
Fact
3
5
5
Skip IO
Due to Storage Index
5
5
35
Query
Benefits Multiply
37
2 or 4 TB
DRAM
Hottest Data
22 TB PCI
FLASH
Active Data
500 TB
DISK
100 GB/sec SQL data scan rate for reporting and warehouses
Cold Data
38
1.5 Million SQL random read I/Os per second for OLTP
New
Transaction
Response Times
Automatic and
Transparent
40
Query Throughput
100
4 to 13
Storage Array
41
Exadata
500 TB of disk per rack for historical data, low activity data,
Cost of Disk
I/Os of Flash
Speed of DRAM
42
Sub-second analytics
Real-time decision making, instant reports
Active Data
43
Database B
44
Exadata Cell
InfiniBand Switch/Network
Exadata Cell
Exadata Cell
Database A:
Reporting: 60% of I/O resources
ETL: 40% of I/O resources
Database B:
Interactive: 30% of I/O resources
Batch: 70% of I/O resources
to hardware
Over 5x faster
encrypted database
Queries decrypt data at
hundreds of Gigabytes/second
46
Exadata Management
47
Comprehensive CLI
Local Exadata Storage cell management
Distributed shell utility to execute CLI across multiple cells
48
49
FRU replaced by
Field Engineer
Oracle Field
Engineer
FRU
dispatched
by Support
Engineer
Fault occurs
SR routed to
Support
Engineer
Customer
SR creation email
notification to customer
ASR
Manager
Products auto-diagnosis
facility sends SNMP trap to ASR Manager
50
ASR
Service
Service Request
created
Comprehensive
Fault Coverage
CPU
Disk controllers
Disks
Flash Cards
Flash modules
InfiniBand
Cards
Memory
System Board
Power supplies
Fans
24/7 support
and updates
Available for certified configurations on Exadata
1 Covered
51
system must be within an Oracle two-hour service area to receive two-hour response as a standard service.
Real
Applicatio
n Clusters
WAN
ASM
RMAN / Fast
Recovery
Area
Oracle
Secure
Backup
Standby
Dev/Test
53
Business Objectives
Reduce cost and complexity
Faster
Applications
Storage
Savings
Maximum
Availability
Cost
Savings
10X speedup
1 Petabyte
99.95%
75% Cost
Reduction
32 billion
queries/day
Replaced $$ SAN
Exadata V2
Primary/Standby
Exadata X2-2
Primary /Standby
Exadata X2-8
Primary/Standby
Data Guard
Data Guard
Data Guard
Consolidation
Solution
2010
54
Automated Targeting
Systems
2010 - 2011
Automated Export
Systems
2011 - 2012
Passenger
Systems
Enterprise Data
Warehouse
Business Objectives
Storage
Savings
Reduced
Admin
10X
900 TB
20%
27 min to 3 min
(avg for 50k rpts)
Pre-Exadata
Data Warehouse
Speed up BI
Lean, green data center
Faster
Reports
Hitachi
USP-V
5 Racks
EMC DMX-4
5 Racks
Data Center
Cost Savings
80% Less Power
30 m2 Less Space
1,000 TB to 100 TB
Exadata V2
Data Warehouse
2010
2 Exadata X2-2
Prod
2011
Backup
Original V2
Prod/DR/Dev
Restore
Solution
2010: Replace 11 racks with
1 full-rack Exadata V2 for DW
2011: Add 2 full-rack Exadata
X2-2s for DB consolidation
55
250 TB
Raw Data
25 TB
Compressed
2 Prod Databases
600 TB Raw / 60 TB
Compressed
16-node RAC Cluster
4 Prod, 2 Test
Databases
400 TB Raw/
40 TB Compressed
2 RAC Clusters
Extreme
Capacity
Storage
Savings
Capture Lost
Revenue
10 TB/day
900 TB
$ Millions
1,000 TB to 100 TB
Business Objectives
Ensure billing accuracy
Enhance customer service
Predict data traffic trends
Exadata V1
Data Warehouse
(2009)
Exadata V2
Data Warehouse
Exadata V2+X2-2
Data Warehouse
Exadata X2-2
Qtr Rack (2010)
Billing Analysis
System
Spam Filtering
2010
2011
Solution
2009: Create new DW on Exadata
V1 half-rack
2010: Move to full-rack Exadata V2
2011: Add full-rack Exadata X2-2
56
Billing Analysis
System
Benefits
Oracle Exadata helped us meet our challenges by
enabling us to consolidate to a single footprint.
- Ed Link, Vice-President IT, Garmin
Business Objectives
4x Growth
Data Center
Cost Savings
Up to 11X
Pre-Exadata
12 Prod Servers
99.95%
Uptime
Exadata V2
DB Consolidation
Exadata V2
Dev/Test/Local Standby
Data Guard
Solution
2012: Consolidate 12 servers
onto two Half-rack Exadata
V2 systems
57
High-Capacity
Exadata Storage
Servers
5 Prod
Databases
Archived
Data
Tape backup
Benefits
One number for me to call; no more forwarding
phone calls and email between vendors when we
are digging into an issue.
- DBA Team Leader
Business Objectives
Uniform
HA/DR
Simplified
Support
Up to 20X
Pre-Exadata
Exadata X2-2
DB Consolidation
Data Guard
2011
Solution
2011: Consolidate14 servers
onto two Qtr-rack Exadata
X2-2
58
Data Guard
14 Legacy Servers
PeopleSoft HR
3 O/S Versions
Ordering
18 Oracle Databases
Timekeeping
5 Oracle DB Versions
Decision Support
12 Application Teams
Data Center
Cost Savings
Exadata X2-2
Dev/Test
Business Objectives
Exadata V2
Data Warehouse
Prod Non-Prod
Faster
Applications
Storage
Savings
2 - 30X
3x 10x
Compression
Exadata X2-2
Data Warehouse
Prod
DR
Dev
UAT
Simplified
Support
Exadata X2-8
Data Warehouse
Prod
Non-Prod
Exadata X2-2
OLTP/Analytics
Prod
Non-Prod
Lower costs
Boost performance
Solution
Data Guard
59
Trade/Mkt/POS Data
Shipments Data
Finl/Planning Data
Trade Fund
Management
Objectives
Reduce cost and complexity
Faster
Applications
Reduced
TCO
Maximum
Availability
5X+ Speedup
50%
99.95%
Pre-Exadata
(Hosted)
Exadata X2-2
DB Consolidation
Data Guard
2012
Data Guard
Solution
X2-2 for production E-Bus Suite
X2-2 for Standby DR and test/QA
Database Appliance for testing
60
Production
Disaster Protection
3 O/S Versions
E-Business Suite
Test and QA
Data Center
Cost Savings
Oracle Database
Appliance
Dev/Test
Benefits
Faster
Response
5x 60x
Business Objectives
Company-wide collaboration for
> 100K users
100%
Uptime
Exadata V1
Storage Servers
Exadata X2-2
Production
Exadata X2-2
Standby
2011
Data Guard
Guarantee uptime
Solution
96 V1 storage servers
9 full-rack X2-2
9 full-rack X2-2
Post-Sun acquisition,
CPU and disk
oversubscribed
Triple mirroring
48 TB flash
61
12:1
advantage
36 total racks
52
GB/sec
63
1 Rack Exadata
52 GB/sec
64
8 Rack Exadata
$625,000
9:1
advantage*
$5,630,000
* Hardware list price comparisons
65
IBM
P780
DS8800
with Flash
67
68