This article can also be found in the Premium Editorial Download "Storage magazine: Hot, warm, cold: Pick the right disaster recovery site."
Download it now to read this article plus other related content.
For some ROBOs, centralized backups may not be an option at all unless network changes are made and additional bandwidth is provisioned. Features like source-side
The biggest challenge of centralized backup without a local copy is recovery. While backups are incremental with changes trickling across the network over time, recovery is more taxing since a much larger amount of data may have to be restored in a short span of time. Even with a feature like source-side deduplication that helps to quicken recovery, a predefined recovery plan that outlines the recovery strategy for various scenarios is a must.
“Due diligence of analyzing the impact of remote backups on the network and a clearly defined recovery strategy are essential to preventing unpleasant surprises,” StorageIO Group’s Schulz said.
Backing up to a centralized data center with a local copy at the ROBO. The practicality of a centralized backup of a ROBO without a local copy for quick recovery declines with the amount of data to be protected. “As you approach and exceed 10 TB of data, a cached copy of the latest backup set becomes increasingly relevant,” said Steve Wojtowecz, vice president of storage software development for IBM Tivoli. In a centralized backup with a local copy scenario, a backup application usually backs up data to a local disk target where it’s then replicated asynchronously to the central data center backup infrastructure. All major backup application vendors have supplemented their apps with replication options: CommVault Simpana with DASH Copy; EMC Avamar by means of a storage node in the ROBO to replicate local backups to a centralized Avamar Data Store; IBM with Tivoli Storage Manager FastBack; Symantec with both NetBackup and Backup Exec appliances. In all these solutions, replication is a component of a larger backup suite and architecture, and works well in places that have standardized on a backup suite. They’re less suitable in companies with a more heterogeneous backup landscape.
In heterogeneous backup environments, replication is usually deferred to backup targets that are capable of performing replication independently of the backup software in use. One way of accomplishing this is via general-purpose disk-based or software-based replication with the caveats of lacking backup focus and features, and adding complexity. That’s when dedicated backup targets with replication capabilities come into play. They’re backup storage appliances padded with features relevant to backup and recovery, such as replication, deduplication, compression and centralized management. Leading the pack and best known is the EMC Data Domain family, with appliances ranging from 1 terabyte to hundreds of terabytes of capacity. Hewlett-Packard with its StoreOnce Backup System and Quantum’s DXi-Series of backup appliances also compete in this space. Because these intelligent backup targets are decoupled and independent from backup applications, they’re required in both the ROBOs and the central data center. While the data center appliance that aggregates backups from the ROBO appliances is usually a large redundant system with sufficient capacity, appliances in ROBOs are generally much smaller, single systems. Unlike replication options that leverage an existing backup suite, these backup appliances are usually less cost effective, especially since they’re required in all locations.
Using a cloud backup service without retaining a local copy. The recent rise of cloud computing and cloud storage gives companies yet another data protection option for their ROBOs. Backing up directly to the cloud is very similar to backing up to a centralized data center, especially from the ROBO’s perspective.
This was first published in January 2012