Regardless of how big or small an organization is, chances are that they rely heavily on a database to power some of their most important mission-critical applications. For this reason, organizations go to great lengths to ensure databases are protected and data has been backed up multiple times to various locations. While the focus seems to always be on backups, a strong data protection strategy must also take into consideration recovery–more specifically, the Recovery Time Objectives (RTO). After all, it’s the recovery that gets us back online, granting end-users and customers access to what they need.
The database of choice for many enterprises is Microsoft SQL Server. Traditional recovery methods that move data from a data protection solution back into production SQL Servers are ideal for entire database restores, but they lack the ability to restore individual tables or records. Should we really have to wait for storage to be provisioned and a complete database to be copied across the network just to restore a few rows or records that were accidentally deleted? When dealing with large databases, traditional recovery methods can drastically affect whether or not an RTO is met. Even more, when dealing with highly-transactional databases, companies run the risk of losing some transactions just to restore others.
Rubrik’s Live Mount for SQL Server solves these object-level efficiency challenges by eliminating the need to provision storage resources up front. This is achieved by mounting a point-in-time copy of the database directly from the Rubrik cluster, which allows Rubrik to copy the needed data using standard ad-hoc queries without taking the production database offline.
Rubrik’s SQL Server Support
Before we move into the architecture of Live Mount, let’s first take a look at process and options around protecting SQL Server within Rubrik. As with physical Windows and Linux workloads, SQL Server support utilizes a connector installed via an MSI, the Rubrik Backup Service (RBS). Once installed, the lightweight service will discover and inventory all of the SQL Server instances and databases residing on the host, sending that data to Rubrik. From there, databases and instances are protected just like any other object within Rubrik. Users simply add the desired database or instance to an SLA Domain. Rubrik then ingests backups at the database/instance level and ensures data protection policies and RPOs are satisfied. On top of that, SQL Server backups can take advantage of many Rubrik features such as incremental-forever backups, fast and efficient point-in-time restores, transaction log replaying, and CloudOut archive capabilities.
When it comes to recovery, Rubrik provides three different methods to cover a variety of disaster scenarios. The traditional Restore option facilitates the copy of data back to the original production SQL Server instance, completely recovering and overwriting the original production database. Similarly, the Export option will copy the database to a SQL Server instance, but the user can choose which SQL Server host to use.
Both the Restore and Export options require the storage resources provisioned up front, as well as the need to wait for the entire database to be copied across the network. While this may be sufficient for smaller databases, as we approach the hundreds of gigabytes or even terabyte size, challenges around maintaining RTOs emerge. In addition, Restore and Export work on the database level, so to restore data in one table, the user must restore the entire database. In the case of highly-transactional databases, this could result in data loss from some tables to support recovery in others.
Rubrik’s third recovery option, Live Mount, is designed around solving these operational challenges.
Live Mount for SQL Server
The Live Mount recovery method for SQL Server provides a means for customers to restore databases, both complete and at a granular object level, without having to provision storage or wait for data to traverse the wire. This is achieved by utilizing Rubrik’s resources as the underlying storage that hosts our database file structure. The point-in-time backups, along with any transaction logs, are replayed and presented to the SQL Server of choice via an SMB 3.0 share. The RBS then creates a new database on the SQL Server, mapping the underlying file structure to the mdf/ldf files located on the Rubrik cluster.
We only need to select a point-in-time backup to mount, provide information around a new database name, and choose a SQL Server target. Rubrik will then replay any transaction logs between backups, create the SMB share, and mount the database – all automatically. Due to the nature of not needing to copy large amounts of data across the wire, this drastically reduces customers’ RTOs down to minutes or even seconds. Live Mount opens up a lot of use cases, including object-level recovery.
Restore Only What We Need!
Utilizing Live Mount, Rubrik can gain access to point-in-time copies of a user’s databases within minutes. Since a database is presented in its native SQL format, we can run standard ad-hoc T-SQL queries to copy data out of the mounted database and back into the production database.
This approach allows users to easily recover only the desired portions or objects within their databases. No need to wait for a complete database copy to traverse the network, provision its supporting storage, or overwrite a complete database just to access a small number of records. Additionally, Live Mount is not restricted to just the data living locally on-premises. Any database backups that have been archived, either locally or to the cloud, can also be utilized in Live Mount operations.
When it comes to efficiency, there is no better recovery method than Live Mount. To learn more about how to perform a SQL Server Live Mount, check out this overview on granular object-level recovery with SQL Server. To dive deeper into all of Rubrik’s SQL Server functionality, read the Protecting Microsoft SQL Server with Rubrik white paper.