Disaster Recovery is based on the fact that all involved components (SAN, HBAs, FC-Switches, RAID - disk resources, cluster nodes etc.) are layed out redundantly. For a 24/7 availability this redundancy may go as far as a Backup-Data-Centre separated by open ground. We at OSL have addressed ourselves to the task to bring efficiency and ease of use to the process of recovering after disaster situations.
All major RAID-system manufacturer offer different Disaster Recovery (DR) solutions. In the majority of cases it is assumed that one Productive-RAID-System (Primary-Data-Centre) is at least partly mirrored to a Backup-Data-Centre via suitable technology (FibreChannel, Infiniband, ...). If the Primary-Data-Centre fails, the secondary data centre (Backup-Data-Centre) takes over. In principle this setup offers some reliability. Unfortunately it also complicates the overall handling, since an assignment application ↔ LUN may not always be easily reastablished. Same problems are known from backup and restore. Unfortunately, there is frequently no fast restart possible. In particular crucial are databases, where besides database files, archive logs, online logs and control files (e.g. oracle) are required up to breakdown time for a restart without loosing any DB transaction. The second obstacle comes through the above mentioned mirroring to the Backup-Data-Centre, since database failures (drop table) are also mirrored.
A holistic concept aimed at large databases/SAP environments with OSL Storage Cluster
OSL Storage Cluster offers a solution which combines Disaster Recovery, Backup-to-Disk, Backup-to-Tape and Instant Recovery for (not necessarily large) Database/SAP-System. A well-balanced combination of synchronous and asynchronous mirroring, ensure a maximum protection against typical breakdown and failure situations, scores high with fast database restart, guarantees an outstanding performance and allows cost reduction due to minimal hardware requirements. One major advantage over our competitors is due to our application-oriented approach. Once an application is defined with all necessary resources (virtual interfaces, filesystems, start-,stop-,break-procedures, etc) n OSL Storage Cluster, this application may be started and maintained by all cluster nodes.
Application Mirror for Disaster Recovery
A mirrored application may be created with a single command. (XDM). Due to the concept of "Storage Universes", mirrored applications have the same name as their source, they only differ in their Universe number. A reasonable assignment of Storage Universes may be used in order to tell permanently mirrored applications from Backup-to-Disk mirrors apart. Backup-Applications may be resynced and disconnected time consistently. This way a immediately can-be-started image is on disk, on which we may fall back on Disaster Recovery. Databases (e.g. oracle) are set to Backup-Mode, an Online-Backup becomes available. Together with permanently mirrored Redologs, these Backup-to-Disk may be easily rolled forward, without involving any tapes. To make a long story short: Instant Recovery. Furthermore this Backup-Application can easily be written to tape (e.g. Networker). Since the Backup-Application is first written to Disk and later written to tape by a dedicated Backup-Node, this online backup shows no impact on your productive application. One more convincing argument: There is only one Backup-Node per cluster necessary. License fees for Backup-Clients may be saved. Nice side effect: Streaming to tape is done over a dedicated SAN paths at significantly increased speed. If in an environment like this the Primary-Data-Centre breaks down, using Backup-Applications on Backup-Data-Centre resources and you are up and running in no time.