Data Domain Protection For Large Enterprise Databases: Oracle, SAP & IBM
Diana Yang – Advisory Systems Engineer Yatin Patil – Product Management [email protected] [email protected]
© Copyright 2017 Dell Inc. 2
2020+
1PB??
Database sizes keep growing steadily How large is a large database?
2010
10TB
2015
40TB
2017
200TB
How can your backup architecture keep you protected as your data grows?
© Copyright 2017 Dell Inc. 3
The challenges of backing up large databases
Technical Factors Customer Constraints
• Long backup & restore times
• Application impact
• Architectural complexity
• Storage requirements
• Database compression
• Short backup windows, tight SLA
• Application uptime
• Ease of use
• Storage budget pressure
• Database encryption requirements
© Copyright 2017 Dell Inc. 4
Snapshots are not backups Snapshots have their place, but don’t replace a full backup
Production DB server
STD snapshot
Backup software Backup Storage
Mount server Snapshot drawbacks
❌ Consume primary storage space
❌ Can’t protect against disk failures
❌ Low application impact, but other issues – Management overhead – What about LTR? DR?
❌ Backup time includes snapshot time + backup time from mount server • Snapshot can be mounted on an alternate server
• Offloads impact of backup processing on DB server & primary storage
© Copyright 2017 Dell Inc. 5
Large databases challenge traditional backup software
❌ Application impact - all data passes through application server
❌ Missing SLAs (RPO/RTO) – A lot more data to move – Complex data path
❌ Complex and Expensive – Needs more backup or media servers – More network connections – Infrastructure required to meet SLA
continues to grow
❌ Management are shared
Backup Admin
Application Owner
VTL
WAN
DD Replication
Database servers
Backup media servers
VTL
VTL
Backup infrastructure can be the bottleneck
DR datacenter
A better way to protect large databases
© Copyright 2017 Dell Inc. 7
Modern Data Management
The evolution of data protection
Traditional backup
Backup software
Protection storage
Ana ly t i cs & Search
Intelligent copy oversight
Backup servers/agents Self Service (from Data Source)
© Copyright 2017 Dell Inc. 8
App-integrated backup: Data Domain Boost Offloads deduplication processing to the application client
LAN / SAN
DD Boost
Database Server
Data Domain
ID Segment Compress Unique?
Write
unique data segments
Primary storage agonistic
Only changed data sent to Data Domain
Direct data path from database server to Data Domain
50% Faster Backups
99% Less LAN Bandwidth
© Copyright 2017 Dell Inc. 9
Storage-integrated backup: ProtectPoint technology Integrates primary & protection storage - supporting VMAX3, VMAX Flash, and XtremIO
1
Catalog
Agent
Backup
App owner triggered backups
Production
2
3
Primary storage snap Changed blocks to Data Domain
Data Domain creates a full backup in native format
20x faster backup, without application impact
Only changed data sent to Data Domain
Direct Data path from Primary storage
ProtectPoint
App Integration
Protection Storage
CBT, Data mover
10x faster recovery
Data Domain Boost is key for
large database backup
© Copyright 2017 Dell Inc. 11
Fast backup for large databases via DD Boost • Application owner Self Control
• Meet (RPO/RTO)
• Direct model management and reporting
– Oracle Enterprise Manager (OEM) – Data Protection Advisor – Third party tool
• Backup software model – backup have dependency on backup software
• Backup software management and reporting
– Data Protection Suite Family and DPA
Application Owner
DPA
RAC configuration
DD Boost
DD Boost
DD Boost
Mtree Replication
WAN
DD Boost DD Boost
Data Management
Policy
Reporting
Catalog
Optimal data path with a variety of management options
© Copyright 2017 Dell Inc. 12
DD Boost for Enterprise Applications Native application integration, through which data protection controlled by application owners
DD Boost
via RMAN
via BR*Tools
via backint SAP HANA studio
via db2 commands IBM DB2 studio
via SQL Server Mgmt. studio
via Exchange PowerShell
via CLI commands & Hadoop APIs
© Copyright 2017 Dell Inc. 13
Oracle Real Application Cluster (RAC) architecture
Service Listener
Instance 1 ASM
Oracle Clusterware Operating System
Service Listener
Instance 1 ASM
Oracle Clusterware Operating System
Node 1 Node n
cluster interconnect
Redo / archive logs (all instances)
Database / control files
OCR / Voting disks
Public Network
Shared storage ASM
managed
ASM / RAW devices
© Copyright 2017 Dell Inc. 14
Oracle integration Standalone
RAC configuration
Production Standby
Data Guard
Data Guard
Mtree Replication
WAN
Disaster Recovery Site Local
Data Center
DD Boost DD Boost DD Boost
DD Boost
DD Boost
DD Boost
DD Boost
DD Boost
© Copyright 2017 Dell Inc. 15
Oracle backup performance using DD Boost
• Database backup time using TSM --- 25 hours
• Database backup time using DD Boost – 4 hours
Peak throughput: 30 TB/hr Average throughput: 20 TB/hr
UU634 – Backup Performance
0.00 5.00 10.00 15.00 20.00 25.00 30.00
Original TSM
DDBoost 10Gb (Initial)
DDBoost 10Gb (Subsequent)
Back
up T
ype
Backup (Minutes)
https://www.emc.com/collateral/white-papers/h14026-xtremio-dd-rman-wp.pdf
Four Oracle
Server RAC DD7200 XtremIO
Four Oracle Server RAC
© Copyright 2017 Dell Inc. 16
Best dedupe rate for Oracle Using Data Domain’s Oracle optimization parameter
• Oracle multiplexing reduces effectiveness of other deduplication appliances except Data Domain
• Understands Oracle’s backupset data structure
• Result: 30% increase in deduplication
• Highest dedupe rates of any dedupe appliance.
Block header changes when o Block data changes o Multiplexing is enabled
Backup Stream File1
File3
File2 RM
AN 2 3
Database
Data Set
1 2 3
1 2
1 2 3
3 1
Header Data
Block size – 20 bytes 20 bytes
© Copyright 2017 Dell Inc. 17
Oracle Transparent Database Encryption (TDE)
• Increasing adoption of database encryption
• RMAN encryption not needed -encrypted data is backed up
• Data Domain still dedupes this data because encrypted data doesn’t change till updated
• TDE consumes more backup storage because encrypted data cannot be compressed
Oracle Encrypted Primary storage
O-Encrypted A
O-Encrypted B
O-Encrypted C
O-Encrypted D
Data Domain Backup Storage
O-Encrypted A
O-Encrypted B
O-Encrypted C
O-Encrypted D
O-Encrypted E
O-Encrypted E
Oracle RMAN
O-Encrypted A
O-Encrypted B
O-Encrypted C O-Encrypted D O-Encrypted E
O-Encrypted B
O-Encrypted C O-Encrypted D
O-Encrypted A
© Copyright 2017 Dell Inc. 18
Oracle database encryption (Vormetric)
• Encryption management is a nightmare without centralized key management
• Increasing adoption of enterprise level encryption software like Vormetric
• Backup data are unencrypted on server first
• RMAN encryption is not recommended with Data Domain.
• Data Domain have encryption in flight and encryption at rest
Vormetric Encrypted
Primary storage
V-Encrypted A
V-Encrypted B
V-Encrypted C
V-Encrypted D
Data Domain Backup Storage
D-Encrypted A
D-Encrypted B
D-Encrypted C
D-Encrypted D
V-Encrypted E
D-Encrypted E
Oracle RMAN
D-Encrypted A
D-Encrypted B D-Encrypted C D-Encrypted D D-Encrypted E
A B C D E
A B C D
D-Encrypted A
D-Encrypted B D-Encrypted C D-Encrypted D
© Copyright 2017 Dell Inc. 19
Tuning for device type SBT
RMAN Parameters Section Size Break large file to multiple channel
Number of Channels Performance increase with more channels
Backup type Full and/or cumulative incremental
Format %d_set%s_piece%p_%T_%U
Block Change Tracking enable
Compression No
Encryption No (Oracle database can be encrypted)
© Copyright 2017 Dell Inc. 20
DB2 shared-nothing architecture for scalability DB2 Data Partition Function (DPF)
Shared nothing DB2 database via Hash Partitions • Database is divided into multiple partitions • Partitions run on different servers • Each partition has balanced resources • Parallel processing occurs on all partitions and is coordinated by the DBMS • Single database system image to user, DBA and application
data + log Partition 1
Engine
data + log Partition 2
Engine
data + log Partition 3
Engine
data + log Partition N
Engine
FCM network
Tables
Database
select . . . from table
© Copyright 2017 Dell Inc. 21
DD Boost for IBM DB2 IBM DB2 DPF integration
Mtree Replication
WAN
Disaster Recovery Site Local
Data Center
DD Boost DD Boost Application
Owner DB2 DPF Node 1
DD Boost
DB2 DPF Node 0
DD Boost
DB2 DPF Node n
DD Boost
• Direct App Owner control of backup using IBM DB2 Studio
• Faster backup and recovery using native DB2 utilities
• Supports IP or Fibre Channel
© Copyright 2017 Dell Inc. 22
Best dedupe rate for DB2
• Multiplexing - DB2 multiplexes by reading data from
several tablespaces simultaneously
• DD Boost DB2 integration - No performance impact from DB2
multiplexing - Great dedupe rate without DB2 backup
parameter “dedupe_device” - Dedupe_device reduces the backup
performance when there are only few large tablespaces
Tablespace C
Tablespace B
Tablespace A
db2m
db2m
db2m
db2agent
db2med
parallelism sessions
backup buffers
db2med
db2med
DB2s backup process model
© Copyright 2017 Dell Inc. 23
DB2 native encryption • DB2 has two different processes to
encrypt data and backup data: • DB2 native encryption can encrypt the
database and send encrypted or unencrypted backup data to backup storage
• Data Domain supports DB2 database encryption and unencrypted DB2 backup
• Data Domain offers encryption in flight and encryption at rest
• Disable DB2 backup encryption - Change DB2 database config parameters
“ENCRLIB” and “ENCROPTS” to null - Make sure DB2 backup script don’t encrypt
DB2 Encrypted Primary storage
DB2-Encrypted A
DB2-Encrypted B
DB2-Encrypted C
DB2-Encrypted D
Data Domain Backup Storage
D-Encrypted A
D-Encrypted B
D-Encrypted C
D-Encrypted D
DB2-Encrypted E
D-Encrypted E
D-Encrypted A
D-Encrypted B
D-Encrypted C
D-Encrypted D D-Encrypted E
A B
C D E
B
C
D
A
D-Encrypted A
D-Encrypted B
D-Encrypted C
D-Encrypted D
© Copyright 2017 Dell Inc. 24
DB2 Tuning
DB2 Backup Parameters Dedupe_device Optional, not required like other dedupe appliance
Number of sessions Performance increase with more sessions
Backup type Full and/or cumulative incremental
DBPARTITIONNUMS ALL
Archive log destination DD Boost or DD NFS
Compression No
Encryption No (DB2 database can be encrypted)
© Copyright 2017 Dell Inc. 25
DD Boost for SAP for Oracle
• Direct App Owner Control of Backup using SAP BR*Tools
• SAP BR*Tools can use Backint or oracle RMAN to backup SAP Oracle
• Supports IP or Fibre Channel
• Support SAP with Oracle: BR*Tools: 7.10, 7.20, 7.40 (64 bit)
SAP/ Oracle
DD Boost Mtree
Replication
WAN
Disaster Recovery Site Local
Data Center
DD Boost DD Boost
SAP Certified April 2014
© Copyright 2017 Dell Inc. 26
SAP HANA Large Database Architecture Scale-up or scale-out
Scale out configurations: • Multi-node • Scalable & highly available • Easy non-disruptive expandability • Compellent fibre channel SAN for
enterprise class features • Built to be mission-critical • Now certified and available
Single-server configurations: • Self-contained XS/S+/Medium configs • SSD and SAS internal storage • For small workloads or development • Available since May 2012
256GB expandable to 512GB
- 2-4 x Intel E7 - 785GB PCIe flash - 1-2TB storage volume
128GB
- 2 x Intel E7 - 785GB PCIe flash - 1TB storage volume
512GB
- 4 x Intel E7 - 785GB PCIe flash - 2TB storage volume
1TB
- 2 R910 active. 1 Stby - 4 x Intel E7 - 1 Compellent SC8K
2TB
- 4 R910 active. 1 Stby - 4 x Intel E7 - 1 Compellent SC8K
4TB
- 2 R910 active. 1 Stby - 4 x Intel E7 - 1 Compellent SC8K
© Copyright 2017 Dell Inc. 27
DD Boost for SAP HANA Fully integrated into SAP HANA studio and SAPs Backint API
Master Name Server
Index Server
Savepoint
Savepoint
Name Server
Index Server
Savepoint
Name Server
Index Server
Savepoint
Synchronized backup savepoint
Backup
Backup/Restore using HANA’s backup API - Backint • Full support of Backint – SAP API for backup agents • Performance: backups transferred via named pipes
Log Backups • Save redo log entries contained in the log area • Log Backups are carried out automatically
SAP-Certified Backint integration
© Copyright 2017 Dell Inc. 28
SAP HANA Backup and Recovery Improved performance: Backint now uses parallel streams for data backups
To configure the number of parallel streams, use the parallel_data_backup_backint_channels ini file parameter (default: 1, max: 32). Same number of streams used is the same during recovery (independent of the current setting of the parameter)
Parallel streaming: • Individual backups distributed across available streams • Used for both full & delta backups • Must be 128GB or larger to distribute
ProtectPoint Technology
© Copyright 2017 Dell Inc. 30
Backup with ProtectPoint technology Storage-integrated data protection: support VMAX3, VMAX Flash, and XtremIO
Application Server
1
2
3
Production
Backup
Application Owner
Catalog
Agent
• App owner triggers backup at an application consistent checkpoint
– RMAN proxy option – DB2 snapshot option
• Primary storage sends changed blocks directly to Data Domain
• Data Domain uses the changed blocks to create full backups in native format
• 20X faster backup No Application Impact
© Copyright 2017 Dell Inc. 31
Backup with ProtectPoint Storage-integrated data protection: support VMAX3, VMAX Flash, and XtremIO
+Data
+Redo
+FRA
Production Devices
VMAX3 Native Devices
snaps
+Data
+Redo
+FRA
Encapsulated backup vdisks
snaps
+Data
+Redo
+FRA
Encapsulated restore vdisks
Management host Production Recovery host (optional)
1a
1b
2
2
4a 4c
4b
Data Domain static images
ProtectPoint workflow With SnapVX
2 3
© Copyright 2017 Dell Inc. 32
Want to win a Levitating Death Star Speaker?
• Follow @DellEMCProtect while at Dell EMC World
• 2 Winners will be chosen daily from
Monday May 8 to Thursday May 11 • All winners will be notified through
Twitter Direct Message
NO PURCHASE NECESSARY. Ends 05/11/2017. To enter and for Official Rules, visit http://thecoreblog.emc.com/dell-emc-world-follow-win-sweepstakes-2017/
© Copyright 2017 Dell Inc. 33
Learn More: Join The Conversation @DellEMCProtect
Dell EMC Storage and Data Protection
Dell EMC Data Protection Community
Data Protection on EMC.com
Mozy.com
Spanning.com
© Copyright 2017 Dell Inc. 34
You may also be interested in these sessions …
Session Breakout Session Title First
Session Second Session
dps.04 Dell EMC Cloud Data Protection: Extend To The Cloud For Long Term Retention With Data Domain Mon 12:00 Thu 01:00
dps.06 Dell EMC Data Domain: What's New For 2017 Mon 08:30 Wed 01:30
dps.07 Dell EMC Data Protection Suite: What's New For 2017 Tue 03:00 Wed 12:00
dps.12 Data Domain Protection For Emerging Database Platforms: Mongo, Hadoop, MySQL & Others Tue 12:00 Thu 10:00
dps.13 Data Domain Protection For Microsoft Applications: SQL, SharePoint & Exchange Wed 03:00 Thu 11:30
bof.14 Bird’s Of A Feather: Data Domain Ask The Experts Tue 01:30