So if we look at the data protecdtion strategies offered by Oracle. First and foremost – for Backup & Recovery of an Oracle database running on Exadata we will use Recovery Manager or RMAN, and RMAN can be combined with a media management software such as Oracle Secure Backup to provide our Physical proection. Additionally, for protection from logical failures, Oracle has a suite of Flashback Technologies.With all the different recovery options available, Data Recovery Advisor will help identify the problem and help plan how to recover from a particular situation.In the event of Disaster Recovery, Oracle Data Guard is the recommended protection strategy.We will see how we can use these different products to architect a backup & recovery solution.
Before starting to code your backup scripts though, the recovery requirements must be gathered. Begin by asking these simple questions….How Quickly do you need to recover the database / application. This is your Recovery Time Objective or RTO…….
We know the rates that can be achieved, but lets look considerations for performance v cost.On the left we have our Exadata rack, which offers high backup & recovery rates and we can use all the smart features of the Exadata solution. Additional performance can be achieved if an Exadata Storage Expansion Rack, which in addition to storing database backups can also be used to as a Historical Archive or even an expansion of the application data.Providing High Performance with added Flexibility, we have the ZFS Storage Appliance which offers snapshot and clone solutions in addition to being a backup destination for both database and non-database files, and the StorageTek Tape Library, which offers Offsite backups and Vaulting in addition to being a backup destination similar to the ZFS.The StorageTek Tape Library backup rate is limited by the number of tape drives available. For me – I had 8 tape drives, but the storageteksolition can scale to hundreds of tape drives if required.
Lets focus in on RMAN nowFullBackups offer Faster Restores over Incremental Backups which are great when the amount of data that is changed between backups is less than 20% and when the backup windows are smaller.For Full Backups, there are two types – Image Copies which reside on disk storage and are approximately the same size as the database or Backup Sets that can reside on disk or tape and which are smaller as they only backup the data and not the entire data file and the backup can also be compressed.For Incremental Backups, there are two types as well, incrementally updated backups which reside on disk and offer typically a 1 -> 7 day rolling recovery window, and a combination of Full & Incremental BackupSets that can reside on disk or tape and in addition to being compressed typically offer longer retention windows required to meet legal requirements.
The FRA or Fast Recovery Area is deployed in an Exadata environment in the RECO disk group, and resides on the Exadata storage and will contain the archived redo logs as well as the flashback logs if flashback database is enabled.Optionally, the FRA may also contain the database backups depending upon the recovery solution that is being architectured.The FRA is configured using just two parameters and the database self manages this space. The database will purge files that are no longer required, based upon the RPO, when the space allocated comes under pressure.
We are now going to introduce the concept of tiered storage. Now this probably not new to you, but there are some additional tiers compared to what you are use to.Tier 1 – Our Extreme performance tier using Exadata Storage in the database machine. In additional to potentially housing the controlfiles and redo log files, the FRA will hold the archived redo logs and flashback log files and possibly database backups.Tier 2 – Still using Exadata storage but this time using the High Capacity drives. This is what we introduced previously with the BKUP disk group.The ZFS/SA makes a great 3rd tier. The ZFS/SA can be configured to house the database backups if you decide not to use an Exadata Expansion rack. The ZFS/SA also allows for the creation of dev and test environments via the use of snapshots and clones, and with the ZFS, you can still access all your data including HCC data in it’s original storage.Tier 4 & Tier 5 are our tape offerings. Tier 4 storage is using the tape library for a combination of Weekly Full backups and Daily Incremental backupsTier 5 storage is when the requirements call for a once a week backup or to meet some legal requirement. Both Tier 4 & Tier 5 offer the ability for the backup to be taken offsite for security or legal reasons.Customers often find that a combination of the tiers are required in order to meet the RTO, RPO and data retention targets, and the tiers can be combined offering a D2D2T solution.So lets look at some customer configurations and how they are using these tiers.
If I have a tape library with 8 tape drives that can backup at 8TB/hr, and my database is 80TB in size, then the backup will run in 10 hours.If my database doubles in size, but I still have 8 tape drives then the backup will run in 20 hours.So if I consolidate two databases, one sizes 80TB and the other sized 40TB, and I have my 8 tape drives that can backup at 8TB/hr, my options areI backup the first database starting at midnight and running until 10am, and then at 10am I backup the second database and this would run until 3pmAlternatively, I could start both backups at midnight, and allocate 5 tape drives to the 80TB database, meaning the backup would now run in 16 hours, finishing at about 4pm, and the 40TB database gets 3 tape drives, meaning the backup would run in 13.3 hours finishing around 1:20pm.Alternatively, I could start both backups at midnight, and let the 8 tape drives service the two databases (total 120TB), and while we would not know when each database would finish, both database will have finished by 3pm. In order for both databases to finish by 10am, we would need to add 4 more tape drives (total 12 tape drives or 12TB/hr), and the backup of 120TB would then finish by 10am.
Jeff – are we good with 1.1PB of usable space in RAID-Z?