Versant Varehouse

Varehouse is an add-on software module that enables the Versant Object Database to use the disk mirroring features of EMC Symmetrix or other enterprise storage systems to take an online backup of very large data volumes without impacting availability (High Availability Backup, HAbackup)

Overview

Increasingly, big enterprises are deploying Versant Object Database in mission critical applications. Typically, such customers use mirrored disk devices for high performance and availability. Varehouse provides

• fast database backup/restore without affecting the server availability
• creation of duplicate production databases as a standby for decision support systems (DSS)

Varehouse, Versant's High Availability Backup solution, is a generic way of achieving continuous online backup of the Versant Object Database. It exploits the capabilities of special storage devices. It allows the user to execute certain operations such as, splitting a mirrored device, after bringing the database to a consistent state. Thus, a copy of the database can be made at a consistent point without jeopardizing server availability, data integrity and performance. Subsequently, this copy can be used for various purposes including, continuous online backup and stand-by database for DSS applications.

Versant's incremental backup (Vbackup) strategy consists of three different backup levels viz. level 0, 1 and 2. As database size increases, it becomes more important to set the level in a way, which minimizes backup size and time. Level 0 performs a full backup. Level 1 backs up all changes made since the last level 0 backup. Level 2 backs all changes made since the last level 0 or level 1 backup, whichever was most recent. Roll-forward (RF) archiving is another feature that preserves logical log records generated by a database during normal operation in a log archive. These records can be replayed on the database during recovery. Thus, RF makes it possible to recover a database to its state just prior to the crash.

The Varehouse backup solution is an extension that exploits the device mirroring functionality of the modern disk arrays.

Key Features

Ability to perform Split

A device pair consists of a mirror device and a standard device. The standard device (on which production database resides) can have any mirror structure (unprotected, RAID, RAID with SRDF). After configuration and initialization of a mirrored disk array, the mirror devices contain no data. The device pair should be synchronized so that both the standard and mirror devices contain the same data. Typically the device vendors provide the ability to synchronize and split mirrored pair.

Figure 1 illustrates a Versant object database [say, production_db] in operation. The system, physical log and logical log volumes are shown to be residing on mirrored disk array.

A split operation is carried out by using the –split command of habackup, which ensures that the appropriate timestamp fields on the database are updated and at the same time it is in a consistent state before the split operation is issued. The (device) vendor-specific command to be executed for a split operation is indicated by the -cmd option to habackup. Habackup executes this command or even a script and waits for it to finish. It is assumed that the vendor specific split command or script would return an error if it is unsuccessful. Based on the error returned, habackup will either complete or abort the split. The database image on the mirror device, obtained after the split operation, is a consistent snapshot of the database, which can be archived to a backup media. The "-split" command of habackup is synonymous to level 0 backup of vbackup.

Level 1 and Level 2 backups

Habackup also offers the ability to perform incremental backups. After performing split it is possible to take level 1 or level 2 backups of the production database.

Roll-forward Archiving

Roll-forward archiving can be enabled with -roll option during split or level 1/2 backups. Similar to vbackup utility, -log option can be used to start the archive process and -off to stop it.

Restore

In the event of a crash, the standard device can be restored by forming device pairs (operation against of split) using the commands provided by device vendors. After the device level restore takes place, the standard device contains the snapshot of the database at the time of split. The level 1/2 backups and RF archives of the production database after the split, are useful at this point to bring the database to the state just prior to crash.