Using the cloud as a destination for backups has evolved from a consumer solution into a viable option for many small businesses and larger enterprises. Today production applications and even entire virtualized environments are leveraging the cloud to augment their existing backup processes.
Thanks to technologies like changed block tracking, deduplication and compression the movement of data to the cloud across a relatively low bandwidth internet connection is very viable. Most solutions even leverage a local backup device in some form to assist in completing the initial backup quickly and to provide recovery for the most immediate information.
But what if something goes wrong with the local recovery device or if a version of the data that needs to be recovered is no longer on that device? After all, with a cloud-enabled backup the role of the local device is typically to store the most recent copy of data.
The Reality of Cloud Recovery
In the more extreme, what if the site is damaged during a natural disaster? This is really the ultimate scenario that backup processes are designed for, enabling the business and its data to survive a catastrophic event. In these situations, the bandwidth restrictions of the internet can come back to haunt IT management.
Compression will provide some benefit but deduplicated backups and changed block backups are of no use because, in the end, all the data has to be restored. While a site disaster would be extremely rare, an archived recovery or a backup device failure is not. These more common restore events should be taken into account when considering how to leverage the cloud as part of the backup process.
The most frequent type of cloud recovery is not really from the cloud at all. In this situation the needed data is copied to hard disks, RDX cartridges or tapes and then sent by overnight delivery to the customer. The good news is that the customer gets their data back and most have to endure only one day of downtime. While hurting the company, it typically won’t kill the business. However, this ‘brute force’ process of recovery needs to be replaced with more intelligent methods.
In the modern era of cloud-enabled backups the most common way to get around the recovery challenges caused by limited bandwidth is with a cloud-based recovery process. This method allows a VM or group of VMs to be recovered and restarted on a hypervisor at the provider’s facility. While there may be some networking and routing issues to resolve, cloud recovery allows for a rapid and graceful return to operations.
Cloud-based recovery does require a provider that’s equipped to furnish these services and a software application that supports this “recovery-in-place” process. There is also the challenge of how to move that VM back to the local data center when the disaster has passed and normal local operations are set to resume.
As before, shipping a physical disk could mean downtime while the VM and its data are being transported from the cloud. That said, some providers allow the cloud copy of the VM to run while data is in transit and being seeded at the customer’s site. At that point the customer would still take a short outage to sync the changes that occurred while the disk was in transit.
Companies like PHD Virtual are addressing this missing element of cloud recovery by creating a capability they call “Rollback Recovery”. This process is also called “changed block recovery” because it’s similar to the changed block tracking technology that’s used for backups. With this capability the restore request identifies which changes have occurred between the requested recovery point and the current copy of data. With this information, only the blocks that have changed need to be sent across the internet connection, not the entire VM, significantly reducing the amount of data transmitted.
The Impact of Changed Block Recovery
The first change is likely to be a reduction in the use of local VM recovery-in-place processes, where a VM is run directly from the backup target during the recovery operation. While local recovery-in-place is valuable it does bring with it a unique set of challenges, most notably that the backup device was not designed to be used as primary storage to host a VM. In addition, there is the issue of moving the recovered VM from the secondary storage back to the production store when the cause of failure for the original VM has been resolved.
If the VM to be recovered has not had its data totally lost then a changed block recovery should be very fast. For situations like data corruption or where a new version of an application is causing a problem, a rollback to the prior version of that VM via changed block recovery would be simpler than doing a recovery in place.
However, if the entire disk storage system that held that VM fails then there are no underlying blocks to compare to. In this case it would be faster and simpler to leverage recovery-in-place. Another area where this technology would be an advantage is when creating a “Copy VM” for testing.
It’s clear that both capabilities should be standard issue for the VM backup solutions that IT administrators seriously consider. Missing the ability to do either recovery-in-place or changed block tracking would force them to make compromises in terms of how and when to recover. With products like PHD Virtual Backup no compromises need to be made.
Change block recovery in the cloud
While changed block recovery is certainly an important development, what’s also needed is a cloud-enabled version of this technology. This capability would require that only a minimum amount of data from the cloud be transferred to the customer data center. Both cloud-enabled and local changed block recovery could very well change the way recovery strategies are implemented.
It should be noted that when using cloud storage as the backup target, recovery-in-place will likely become an impractical option, considering the performance impact of the internet on all I/O transactions that would be needed to run that VM. Therefore, changed block recovery provides the best option when the original VM is still in place.
Leveraging the cloud for backup brings many advantages to the virtualized data center. By its very nature the data sent to the cloud is an off-site copy that’s designed to protect against a local or regional disaster. The cloud also allows on-site backup storage to be used only for the latest copy of data, potentially reducing local storage costs and simplifying operations.
Cloud leveraged backup does present a recovery challenge. To overcome that challenge VM backup administrators should look for backup solutions like PHD Virtual Backup that provide the local recovery in place and changed block recovery (both local and from cloud).
PHD Virtual is a client of Storage Switzerland