350 likes | 636 Views
EMC Next-Generation Backup and Recovery for Oracle. Apollo Aguilan. EMC Backup Recovery Systems. Agenda. Problems many database administrators face Overview of Oracle Recovery Manager (RMAN) functional components Key features of EMC Data Domain deduplication storage systems
E N D
EMC Next-Generation Backup and Recovery for Oracle Apollo Aguilan EMC Backup Recovery Systems
Agenda • Problems many database administrators face • Overview of Oracle Recovery Manager (RMAN) functional components • Key features of EMC Data Domain deduplication storage systems • Impact of Data Domain deduplication in an Oracle environment • Customer Experience • Summary • Questions
Problems Many Database Administrators Face Too many for this talk so we’ll focus on just backup and recovery • Non-stop growth and proliferation • Backup windows remain approximately the same • Recovery is just as important and often overlooked • Limited online retention • Increased frequency of recovery for audits, troubleshooting, updates, etc. • Archive (aka, ILM) becoming increasingly important • Some data must be retained for 5 years, 7 years, forever….
RMAN – What Is It? • For those in the audience that don’t already know: • Oracle’s included backup / recovery utility • Supports full, incremental, and image backups • Retains backup information in the control file and (optionally) a separate database catalog • Many configuration options to tailor to users’ requirements • Automatically eliminates unused data blocks from most backups
Overview of RMAN Functional Components RMAN Client Recovery Catalog Target Database Flash Recovery Area (FRA) RMAN Repository (Control File) Backup Backup Disk Media Management Layer (MML) API Third Party Media Managers Tape
Overview of RMAN Functional Components RMAN Client Recovery Catalog FRA Target Database RMAN Repository (Control File) Backup Backup Disk MML API Third Party Media Managers
Data Domain Deduplication Storage Systems Data Domain
Performance: CPU-Centric versus Spindle-Bound Data Domain 6,000 Fibre Channel SATA Throughput MB/s Most deduplication vendors 50 50 100 150 200 Number of Disk Spindles
Deduplication at Backup / Recovery Speeds 102,400 lookups/sec.: is it already in the multi-TB store? 800 MB/sec, for 8KB segment: >100K lookups per second If random lookups: very slow Each disk: 1 MB/sec E.g. for a 7.2KB RPM Seagate 1TB SATA drive: <120 seeks/second 120/second @ 8KB segment: 0.96 MB/sec/disk Would need 830 disks to go 800 MB/sec (like most dedupers today) This is 4 full height racks of nothing but SATA disks (16 drives / 3U rack) Same performance, in less than half a full height rack
SISL Scaling Architecture Check uniqueness with Summary Vector Check in-memory fingerprint cache Key results • Data Domain answer: SISL™ • Stream-informed Segment Layout includes: • Summary Vector in RAM says if segment is new • Segment Localities minimize seeks if answer is on disk See: http://www.usenix.org/events/fast08/tech/full_papers/zhu/zhu.pdf
Data Domain Basics Easy integration with existing environment Control Tier Target Tier Disaster Recovery Tier Backup and archive applications CIFS, NFS, NDMP, DD Boost Ethernet Virtual Tape Library (VTL) over Fibre Channel EMC Symantec CommVault Tivoli Software BakBone Software Quest Replication DD890 appliance DD890 appliance • 2U • 2 to 10 ports • 10 and 1 GigabitEthernet; 8 Gb/s Fibre Channel • RAID 6 • Up to 285 TB usable capacity with shelves • 2 TB or 1 TB 7.2K rpm SATA hard disk drives in shelf • File system • NVRAM • N+1 fans and redundant, hot-plug power supplies
EMC Data Domain RMAN disk-only backup and recovery strategies • Simplifies the backup process • RMAN disk-only backup and recovery strategies are more straightforward • Leverages Oracle Enterprise Manager • GUI for managing backups, restore points, creating backup reports, etc • Supports other Oracle backup capabilities • Incremental merge • Flash Recovery DRDatabases ProductionDatabase Dev/TestDatabases DR Restores can be performed manually or automated via scripts RMAN backups are performed natively via NFS/CIFS Operational Restores can be leveraged for Dev/Test environment refresh Data Domain Deduplication Storage Data Domain Deduplication Storage RMAN backups are replicated automatically using minimal bandwidth
“This is the format for a quote slide.” Attribute
Thurs Incremental A C K Second Friday Full Backup Friday Full Backup Mon Incremental A B H B C D E F L G H A B C D A E F G Weds Incremental Tues Incremental E C G B I J A B C D E F G Data Deduplication: Technology Overview Store more backups in a smaller footprint Backup Estimated Data Logical Reduction Physical FRIDAY FULL 1 TB 2–4x 250 GB Monday Incremental 100 GB 7–10x 10 GB Tuesday Incremental 100 GB 7–10x 10 GB Wednesday Incremental 100 GB 7–10x 10 GB Thursday Incremental 100 GB 7–10x 10 GB Second FRIDAY FULL 1 TB 50–60x 18 GB TOTAL 2.4 TB 7.8x 308 GB H I J K L
Generate Checksum Verify Data Re-Checksum and Compare Verify the file system metadata integrity File System Deduplication Verify user data integrity Local Compression RAID Verify stripe integrity Data Integrity: Data Invulnerability Architecture • End-to-end data verification • Checksum • Deduplication, write to disk • Verify • Self-healing file system • Cleaning • Expired data • Defrag • Verify • Other • RAID 6 • NVRAM • Snapshots End-to-end data verification
Weekly Full Backup – With Deduplication Weekly: Full image backups RMAN> ALLOCATE CHANNEL CH1 DEVICE TYPE DISK FORMAT ‘/dd/backup/ora.weekly/%U’;RMAN> ALLOCATE CHANNEL CH2 DEVICE TYPE DISK FORMAT ‘/dd/backup/ora.weekly/%U’;RMAN> BACKUP AS COPY TAG ‘MAY9’ DATABASE INCLUDE CURRENT CONTROLFILE;RMAN> BACKUP ARCHIVELOGS TAG ‘MAY9’ ALL NOT BACKED UP DELETE ALL INPUT; Before: Without deduplication 1 TB Disk 1 TB Disk 1 TB Disk Target DB 1 TB
Weekly Full Backup – With Deduplication Weekly: Full image backups RMAN> ALLOCATE CHANNEL CH1 DEVICE TYPE DISK FORMAT ‘/dd/backup/ora.weekly/%U’;RMAN> ALLOCATE CHANNEL CH2 DEVICE TYPE DISK FORMAT ‘/dd/backup/ora.weekly/%U’;RMAN> BACKUP AS COPY TAG ‘MAY9’ DATABASE INCLUDE CURRENT CONTROLFILE;RMAN> BACKUP ARCHIVELOGS TAG ‘MAY9’ ALL NOT BACKED UP DELETE ALL INPUT; After: With deduplication Full 500 GB Target DB 500 GB 1 TB
Weekly Full Backup – With Deduplication Weekly: Full image backups RMAN> ALLOCATE CHANNEL CH1 DEVICE TYPE DISK FORMAT ‘/dd/backup/ora.weekly/%U’;RMAN> ALLOCATE CHANNEL CH2 DEVICE TYPE DISK FORMAT ‘/dd/backup/ora.weekly/%U’;RMAN> BACKUP AS COPY TAG ‘MAY9’ DATABASE INCLUDE CURRENT CONTROLFILE;RMAN> BACKUP ARCHIVELOGS TAG ‘MAY9’ ALL NOT BACKED UP DELETE ALL INPUT; After: With deduplication Full Deduplication applied to fulls requiring much less disk 500 GB Target DB 500 GB 1 TB
Daily Incremental Backup – With Deduplication Daily: Incremental backup with update RMAN> ALLOCATE CHANNEL CH1 DEVICE TYPE DISK FORMAT ‘/dd/backup/ora.weekly/%U’; RMAN> ALLOCATE CHANNEL CH2 DEVICE TYPE DISK FORMAT ‘/dd/backup/ora.weekly/%U’; Run { RECOVER COPY … BACKUP INCREMENTAL … } Incrementals after deduplication 500 GB Weekly Fulls Target DB 1 TB
Daily Incremental Backup – With Deduplication Daily: Incremental backup with update RMAN> ALLOCATE CHANNEL CH1 DEVICE TYPE DISK FORMAT ‘/dd/backup/ora.weekly/%U’; RMAN> ALLOCATE CHANNEL CH2 DEVICE TYPE DISK FORMAT ‘/dd/backup/ora.weekly/%U’; Run { RECOVER COPY … BACKUP INCREMENTAL … } Incrementals after deduplication 500 GB Weekly Fulls Target DB • Daily Incrementals 1 TB
Benefits of Using Inline Deduplication • Daily fulls instead of incrementals • Longer online retention of full backup images • No requirement for other backup software plug-ins • Non disruptive introduction into RMAN process • Simplified replication using EMC Data Domain Replicator Software • Ideal location for Flashback Recovery Area • Incrementally updated backups
Incremental Merge EMC Data Domain combines rapid cloning and deduplication • Nightly take cumulative incremental; apply to snapshot of weekly full First image backup2X storage savings • Results in up-to-date nightly full images available for rapid recovery DB Full Snap Snap Snap Snap Snap Incremental Snap Mon Tue Wed Thu Fri Sat Sun
Alternatives Oracle Backup Options • RMAN to 3rd Party Backup Media Management Layer (MML)/ Serial Backup Tape (SBT) • Makes random access disk look like streaming tape for VTL support • Limits the options available to the DBA for backing up, cloning, merging, etc. • RMAN communicates through MML to agent on 3rd party backup application media server • Can front a Data Domain system as either disk-based backup or VTL • All RMAN I/O is via MML which presents only a DEVICE TYPE TAPE Data Domain Systems can work with an enterprise backup application but does not require it.
Data Domain Boost + RMAN = Next-Generation Backup and Recovery for Oracle • Oracle Database Administrator self-reliance • Superior Backup and Recovery Performance • Replication Policy Management • Takes advantage of SBT integration using standard RMAN scripting tools for simple setup Data Domain
Distributed Segment Processing Enabled Distributes parts of deduplication workflow to SBT plug-in on Oracle database server Anchors and fingerprints segments Checks with Data Domain system to find which segments are new/unique Compresses and sends new segments only Avoids sending duplicate data to Data Domain system Provides load balancing and link failover Supported with single controller Data Domain systems, Global Deduplication Array and DD Archiver DD Boost for RMAN Distributed Segment Processing
Start Replication 1 4 3 Done Update Catalog DD Boost for RMAN – Optimized Replication • Managed, bandwidthoptimized replication • Oracle server directs replication of individual backups using RMAN • Data Domain system replicates only unique compressed segments • Reduces workload on the Oracle server • Multiple topologies – one-to-one, bi-directional, one-to-many, many-to-one, hierarchical • Encrypted optimized replication for security over WAN transparent to Oracle servers Oracle Database Catalog DR site Data Domain system Local Data Domain system Copy 1 Copy 1 Copy 2 Copy 2 2 Secure Data Transfer
Advanced Load Balancing and Link Failover Advanced Load Balancing & Link Failover DD Boost level aggregation of multiple 1GbE or 10GbE links on Data Domain system Backup/restore job load automatically distributed on multiple ports on DD system Dynamic load balancing DD Boost library negotiates with the Data Domain system for an interface Load distribution based on number of jobs on the interfaces Transparent failover of jobs In-flight jobs on failed ports on DD system are transparently moved over to healthy links Can be used in conjunction with network level / switch assisted aggregation
Summary • How deduplication benefits Oracle/RMAN backups • More online retention • Less data to replicate • Allows the benefits of incremental update/merge backups using less disk space • Simplification • Minimal changes to the environment • DBA’s familiarity with NAS storage • Savings • Less disk-based storage for longer retention • Reduce the cost of tape consumptions • Administrative cost • Less floor space
More Information… • For more information on EMC Data Domain deduplication storage systems and Oracle, contact your EMC sales representative or visit http://www.datadomain.com/solutions/oracle.html