One of the new data backup techniques fundamentally changing the way people recover from a storage system failure is recovery-in-place. This backup software feature allows an application to access its data directly from a backup storage device, saving the time required to replace a faulty storage device and then restore data across the network.
What is recovery-in-place?
Recovery-in-place is most common in virtualized environments, but some backup software offerings deliver the feature to bare-metal servers. Data backup techniques like recovery-in-place have the ability to present a backup job as a volume that can be accessed by the application. The time required to make the volume ready for access varies depending on how the application stores the data and efficiently prepares it. If data is stored in a live, native format, there is almost no time involved in volume prep. However, a native format has less capacity efficiency and may be more difficult to transfer to a secondary storage device like tape.
Fully leveraging recovery-in-place requires planning, as the backup storage device will be asked to "act" as primary storage for a period of time. This means backup storage has to perform well and have enough redundancy to sustain operations. However, most backup storage devices are optimized to store data cheaply, and do not necessarily provide high performance.
When considering data backup techniques, there are three main considerations:
- Deduplication. While deduplication makes disk backup cost-effective, it may cause a performance problem. The key issue is that the mounted volume has to be deduplicated on the fly as data is being read and written to it. Some disk appliances will overcome this problem by creating a native area or non-deduplicated landing zone. Others will provide extra disk spindles or processors to make the deduplication process happen faster and more seamlessly.
- Frequency of backups. Recovery-in-place can potentially provide an application with access to its data within 30 minutes, but if the most recent backup occurred last night, the user or application is working with day-old data. This may make data protection methods like recovery-in-place unacceptable. If recovery-in-place is the primary method of protection, the backup administrator needs to ensure that the frequency of backup will meet the demands of the user and applications.
- The expectation of the users based on what normal performance looks like. For example, if customers use all-flash arrays, running their application on a deduplicated, compressed, hard drive-based backup appliance may be of little value to them.
The solution is either to have a disk backup appliance with a non-deduplicated area or to use a secondary disk array as a staging area. The most recent backup can go to this array while older ones can go to the original disk backup appliance.
Best use cases
Recovery-in-place is not a replacement for replication or other forms of highly available data protection. IT professionals should assume at least 30 minutes of downtime during preparation of the volume. They should also prepare for less-than-optimal performance, as it is unlikely the disk backup appliance will perform as well as the primary storage system. Finally, they need to factor in the frequency of backups with these kinds of data backup techniques. If backups occur more than every 30 minutes, recovery-in-place can be a very viable and cost-effective approach.
Despite these considerations, most organizations would be very happy with a 30- to 60-minute recovery for their most mission-critical applications. But there are some organizations that will want or even require a much faster recovery time. As a rule of thumb, anything less than 30 minutes will require replication. That way, data protection events can occur more frequently and data is stored in a native, instantly usable state.
Analyst George Crump discusses the use of two more data backup techniques -- snapshots and replication -- and their effectiveness in storage environments.
The reality is that while data backup techniques like recovery-in-place -- sometimes called instant recovery -- are faster than most typical recovery methods, they are seldom instant. IT professionals counting on this method of recovery should assume at least a 30-minute window for recovery. Recovery-in-place also requires rethinking the disk backup appliance and necessitates the use of a secondary disk array instead of a backup appliance. But for most workloads and most data centers, recovery-in-place can be the cornerstone of their data protection strategy.
Use recovery-in-place to back up VMs
Service-level objectives important in recovery-in-place setting
New data protection techniques include instant recovery