news Oracle ZDLRA Zero Data Loss Recovery Appliance December 2014
When Larry Ellison announced the new Oracle Backup Appliance at the Oracle Open World 2013 with the DBLRA (Database Backup Logging Recovery Appliance), he promised numerous features. These features clearly differentiate this solution from others in the market. Just before the introduction into the market the product name was changed from DBLRA to ZDLRA (Zero Data Loss Recovery Appliance). It is obvious that the new product name stresses the minimisation of data loss. Larry's core message was that up to now, solutions have not been developed for databases but rather for files and thus for file systems. But what are the decisive modifications necessary for taking an appliance system designed for file systems and optimising it for Oracle databases? What's more, it is legitimate to ask whether the concepts and features Ellison announced are indeed delivered by the Oracle ZDLRA. The Oracle Zero Data Loss Recovery Appliance is a total package. It consists of hardware and software for the backup of Oracle databases in large environments. With low administrative overhead, thousands of databases can be backed up effectively and reliably. The product is the official cloud solution for database backups. Which Advantages Does the ZDLRA Offer Compared to Traditional Backup Solutions? Replaces the typical backup window Reduces the resource load of the database servers Minimisation of data loss Mirroring of backups over large distances High degree of scalability Autonomous archiving on tapes Central control and monitoring, end-to-end visibility Moreover, the ZDLRA guarantees the option of point-in-time recovery of databases, as is the case with traditional RMAN backups. When is it feasible to use a DBLRA? From an environment as small as 100 databases or for very large databases, the use of the appliance is well worth considering. In your calculation, you should take into account the reduced overhead for administration. In terms of pricing, Oracle is positioned vis-à-vis comparable backup products. Page 2 / 6
How Does the ZDLRA Work? From a technical standpoint, the ZDLRA is an Exadata platform with special pre-installed software. Exadata already has the features built in and can be immediately used for the backup. The core of this system is a RAC database, which serves as central RMAN catalogue. Here the meta data of all backups are saved in ASM disc groups with high redundancy. The backups are saved in the ASM with "normal redundancy". An enlargement of the appliance is possible by increasing database and storage nodes in the same way as it is done with Exadata. For traditional database backups, there is a backup window at night and on the weekend for full backups. This is intended to restrict the periods of increased resource loads on the I/O subsystem or network. With the growth of databases, there are in fact ever increasing conflicts with applications that run in 24/7 mode or are dependent on nightly batch runs. The ZDLRA replaces this window with an "incremental forever strategy". After a complete initial backup of the database, only the redo log files, i.e. change deltas, are backed up. This achieves a significant relief of database server resources. The backup stream distributes itself Page 3 / 6
Oracle ZDLRA automatically and homogeneously throughout hours. The backup of the redo log data on the the time slot. All the costly activities of backups ZDLRA only occurs asynchronously, yet in near are pushed from the database server onto the real-time. appliance. The concept provides for mirroring of the bawith traditional backups, the archived redo log cked-up data on a second appliance. After spedata are backed up externally on the backup cifying the backup policies, the backup occurs medium in particular intervals. The frequency automatically on the closer appliance and the varies from hourly to multiple hours. In case of replication occurs on the distant appliance. In a total failure of the database server, there can case the closer appliance fails, the recovery ocbe a loss of log data in the range of several curs from the distant appliance, without requi- ring intervention in the configuration. This type of mirroring can also be easily realised over large distances, i.e. between Frankfurt/London or San Francisco/New York. Via the definition of security policies, it is possible to divide up the types of backups into groups. For instance, a "Gold Policy" could be created for "mission-critical", a "Silver Policy" for "business-critical" and a "Bronze Policy" for test databases. The Gold Policy would provide for mirroring onto a second appliance as well as an additional backup on tape. Through the expansion of the existing appliance or the addition of more appliances, the ZDLRA offers a high degree of scalability. The backups can expand well into the petabyte range. Backups can optionally be sent from the appliance to a tape library. Here as well, there is a continuous data stream instead of high Page 4 / 6
capacity in the backup window. The operations are performed with the resources of the ZDLRA and not with the resources of the database server. The administration of the ZDLRA as well as the connected database servers occurs centrally with the Oracle Enterprise Manager Cloud Control (OEM). The configuration is specified with the help of backup policies. Monitoring is integrated with preconfigured metrics. In addition, the Enterprise Manager supports an endto-end visibility that cannot be achieved with traditional backups solutions. Key Features in Detail A significant advantage compared to traditional backup is the Incremental Forever Strategy. Here you have a complete backup once, the Initial Backup. After the Initial Backup only changes are saved via Delta Push. Delta Push is used for the following: Incremental Backups Real-time Redo Transport During an incremental backup only changed data is transferred to the ZDLRA. On the appliance this data is checked for corruption and saved as compressed data in the ASM disk group. The backup of the changed data is done by the Delta Store. What is meant by the designation Real-Time Redo Transport? The designation suggests a real-time backup and a 100 percent of data loss prevention. However, does this have a negative impact on the database performance? Oracle has found a good compromise to combine minimal data loss and minimal decline in performance: The transfer of the Redo Records is immediate, real-time, but asynchronous. This is why in normal operation data loss should be maximum one second and there is nearly no impact on database performance. Therefore the user does not wait for ZDLRA confirmation after giving a COMMIT-command like he would have to in the Maximum Protection Mode of Data Guard. As the Redo Record arrives at the ZDLRA it is written into the Redo Staging Area immediately and is persistent then. What follows is the compression and writing into the Delta Store. The technology for the transfer with Real-time Redo Transport is comparable with the Real-tie Apply by Data Guard. The transfer is performed by the Log Buffer of the protected database and happens parallel to writing into the local Redo Log Data. The Delta Store is the central component of the appliance and helps the realization of the Incremental Forever Strategy. Arriving data is changed by incremental backups into a virtual representation of the Level 0 Backup (Initial Backup). This data is administrated by meta Page 5 / 6
data. From virtual point of view, the Level 0 Backup and the Level 1 Backup form a complete Backup at the Level 1 Backup point of time without overwriting the Level 0 Backup. If RESTORE DATABASE is required, the Level 0 Backup and all Incremental Backups are read. The meta data tells the process whether to read a data block from the Initial Backup or from an Incremental Backup. Larry Ellison didn't promise too much: The ZDLRA is a complete hardware and software package for effective backups of Oracle databases in large environments. Its use is recommended for very large/very many databases. It offers a high degree of scalability and relieves the load on system resources. Moreover, it stands for near-zero data loss similar to Data Guard as well as an "incremental forever strategy". Further Information / Community You will find further information here on the Oracle website. Moreover, in the avato intranet you can discuss with our experts and suggest topics for newsletters. Imprint Date: December 2014 Author: Ronny Egner Lutz Fröhlich Contact: marketing@avato-consulting.com www.avato-consulting.com 2014 avato consulting ag Page 6 / 6