<Insert Picture Here>
Maximize Database Availability and Performance
Matthew Baier Oracle Product Marketing Joseph Meeks Oracle HA Product Management
Agenda
Data Protection and High Availability Improve Performance using Idle Redundancy Address Advanced Replication Requirements
Problem: Data Loss and Downtime
Actual Service Request from a Retail Brokerage Firm
ORA-01251 - Corrupted file header. This could be caused due to missed read or write or hardware problem or process external to oracle overwriting the information in file header.
Data Protection and High Availability
For All Outages, Regardless of Cause
Site outage due to large-scale disasters
Fire, floods, hurricanes, earthquakes . . .
Local outages that occur more frequently
Faulty system components Data corruptions Backup/recovery of bad data Wrong batch job Operator errors Planned maintenance Faulty HW/SW installations, upgrades
4
Oracle Data Guard 11g
Included in Enterprise Edition Database License
Data Guard
Primary Site
SYNC / ASYNC
Standby Site
Primary Database
Physical or Logical Standby Database
Data Guard Broker Oracle Enterprise Manager
Data Guard
Database-Aware Physical Replication
Primary Database Network I/O
Log Buffer Online Logs
fil
Standby Database Oracle Apply & Validation
End-to-end validation Automatic block repair Real-time reporting
Archive Logs Flashback Logs Control Files
Data Files
SYSTEM USER TEMP UNDO
Storage agnostic
Storage Remote Mirroring
Primary Volumes
Log Buffer Online Logs
fil
Target Volumes
Network I/O
Oracle Apply & Validation
7X more network volume 27X more network I/Os
Archive Logs Flashback Logs Control Files
Data Files
SYSTEM USER TEMP UNDO
Zero Oracle awareness
Data Guard Provides Best Protection
Better Isolation, Continuous Validation
FEATURE Transmit directly from memory Standby uses different code path Multiple Oracle corruption checks Detect silent corruptions Standby is always on
BENEFIT Isolate I/O corruptions Isolate firmware/software errors Data is logically and physically consistent Data protection unique in the industry Known state ready for production
What Did the Brokerage Firm Do?
They had a Data Guard standby database
Corruptions isolated to the primary database Zero data loss failover over to standby database Resumed operation in minutes Investigated & resolved h/w problems with no impact
Polling Question
Has a storage failure ever resulted in unplanned downtime for your critical databases? Yes No
10
Data Protection AND High Availability
Data Guard Automatic Failover
Events that trigger an automatic database failover:
Data Guard Observer
SYNC / ASYNC
Database down Designated health-check conditions Request from application
Automatic redirection of client connections to new primary
Site A
Primary
Site B
Standby
Integrated role-based services Automatically start/stop services Fast Application Notification (FAN)
11
Data Protection AND High Availability
Automatic Failover
Observer detects failure
Data Data Guard Guard Observer Observer
Failure state is confirmed with the standby database
SYNC / ASYNC
Site A Site A
Primary Primary
Site B Site B
Standby Standby
12
Data Protection AND High Availability
Automatic Failover
Observer automatically executes failover
Data Data Guard Guard Observer Observer
Standby becomes primary
SYNC / ASYNC
Site A Site A
Primary
Site B Site B
Standby Primary
13
Data Protection AND High Availability
Automatic Failover
Observer automatically reinstates the failed primary as a standby database
Data Data Guard Guard Observer Observer
SYNC ASYNC SYNC / /ASYNC
Built-in controls prevent any possibility of a split-brain condition
Site A Site A
Primary Standby
Site B Site B
Primary Standby Primary
14
Amazon
Using Data Guard Automatic Failover since 2007
End-to End Failover Time
60
Resolve Respond Identify
50
Minutes 40 30 20 10
350 Data Guard configurations 150 are configured for automatic failover
Before Data Guard
Data Guard Automatic Failover
15
Unplanned AND Planned Downtime
Data center moves Technology refresh 32bit to 64bit Windows to Linux AIX 64bit to Solaris Sparc NEW Single node to Oracle RAC Migrating to ASM Testing new features Migrating to Exadata storage System maintenance for single node (non-RAC) databases Database rolling upgrades Other database changes Index and storage changes Implementing Advanced NEW Compression Migrating to Secure Files NEW Migrating to Exadata - when changing database extent size
16
Superior Protection AND Lower Cost
Data Guard runs on any storage
Different storage subsystems Different vendors Mixed primary/standby configurations
17
Comprehensive AND Easier to Manage
Enterprise Manager HA Console
18
Polling Question
Which is the greater obstacle to implementing Disaster Recovery for your mission critical Oracle databases? A) Complexity B) Cost C) Other
19
Agenda
High Availability and Data Protection Improve Performance using Idle Redundancy Address Advanced Replication Requirements
20
Problem: Expensive, Idle Redundancy
Traditional High Availability and Disaster Recovery
Redundant systems & storage Production Server IdleFailover Server
HP ServiceGuard IBM HACMP
Does it all work? Find out when you try to failover Idle Disaster Recovery Systems
Veritas Volume Manager EMC SRDF BMC SQL Backtrack
Proprietary Storage Array
Matching Proprietary Storage Array
21
Active Oracle Data Guard 11g
An Option for Enterprise Edition
Fast Read-write Real-time Incremental Workload Reporting Backups
Data Guard
SYNC / ASYNC
Primary Database
Active Standby Database
Primary Site
Standby Site
22
Active Data Guard 11g
Offload Read-Only Workloads to Standby Database
Read-write Workload Real-time Reporting Fast Incremental Backups
Data Guard
SYNC / ASYNC
Primary Database
Active Standby Database
Primary Site
Standby Site
23
Improve Performance
For all Workloads
3500
Transactions / sec
3000 2500
2,610
Double read-write throughput 70% increase in readonly throughput
2000 1500 1000 500
1,530
standby
630 290
primary
R/O R/W
Eliminate contention between read-write and read-only workload
Primary Only
Primary and Standby
24
Improve Availability
Detect and Repair Corrupt Blocks Active Data Guard Automatic Block Repair Automatic detection and repair of corrupt blocks Transparent to the user and application
Read/Write Workload Real-time Reporting
Continuous redo shipping, validation & apply Primary Database Active Standby Database
25
Improve Availability
Detect and Repair Corrupt Blocks Active Data Guard Automatic Block Repair Automatic detection and repair of corrupt blocks Transparent to the user and application
Read/Write Workload Real-time Reporting
Continuous redo shipping, validation & apply Primary Database Active Standby Database
26
Discover Financial Services
Active Data Guard User Active Data Guard Oracle RAC clusters Remote data centers Data Guard Maximum Availability
Recoup DR Investment Impact of ad-hoc queries is isolated to active standby Longer run windows for reports
27
Polling Question
Do you have read-only workloads that could benefit from being offloaded to a standby? Yes No
28
Agenda
High Availability and Data Protection Improve Performance using Idle Redundancy Address Advanced Replication Requirements
29
Problem: Many Replication Requirements
Disaster recovery Data integration Many-to-one replication Active-Active multimaster replication Replication of data subsets Transformations Cross-platform migration
30
Oracle Golden Gate
The Oracle Solution for Information Integration Best-in-class data replication Broad support for heterogeneous platforms Flexible solution for minimal/zero downtime upgrades and migrations
31
Real-time information
Real-time Access
When To Use
Active Data Guard or GoldenGate
Active Data Guard
Disaster recovery, data protection, HA
GoldenGate
Information distribution, data integration
High performance, drop-in solution, support all data types & applications
One-way, complete, physical replication, sync or async Replica open read-only Database rolling upgrade/maintenance
Active-Active HA (avoid/resolve conflicts)
SQL replication multimaster, transformations, subsetting, multiple topologies, asynchronous Replica open read-write Great flexibility for planned maintenance and cross-platform migrations
32
Complementary Solutions
HA/DR + Production Offload + Heterogeneous Replication
Read-write Transactions Read-only Workloads
Production Database
Active Data Guard
Data Guard Standby
33
Complementary Solutions
HA/DR + Production Offload + Heterogeneous Replication
Read-write Transactions Read-only Workloads
Production Database
Active Data Guard
Data Guard Standby
(GoldenGate Source*)
Capture
GoldenGate Replication
Source Trail Target Trail Delivery
GoldenGate Targets
* GoldenGate archive log mode (ALO)
34
Two Ways to Buy Active Data Guard
Either: Standalone an option for Oracle Database 11g Enterprise Edition
Or:
Included with Oracle Golden Gate License
35
Resources
Hands-on Lab, Online via Amazon Cloud Services Learn more about Database High Availability: oracle.com/ha Active Data Guard Hands-On Lab (HOL)
oracle.com/technetwork/database/features/availability/data-guard-hol-176005.html
Same experience as on-site HOL at Oracle OpenWorld 2010
Active Data Guard technical information
oracle.com/goto/dataguard
36
Why Oracle Active Data Guard
From:
Idle redundancy Performance bottlenecks Standby rarely used Low ROI
To:
Fully utilized redundancy
Offload workload to standby database
Always On High ROI
37
Thank You for Attending
Maximize Database Availability and Performance
Save 15% on Oracle Enterprise Edition Options at the Oracle Store
oracle.com/store
Enter promo code 11G at checkout Offer valid through November 15, 2010
38
