Disaster Recovery as a Service (DRaaS) is very appealing to data center administrators looking to meet ever tightening recovery objectives but still reduce costs. By leveraging DRaaS, organizations can eliminate the cost of a secondary site and buying stand-by equipment for that site. But there are aspects of DRaaS that IT professionals need to be concerned with like seeding, recovery and the time it takes to return to the original primary site after disaster has passed. DRaaS is not the only cloud option, using the cloud as primary storage is an alternative that deserves consideration.
What is DRaaS?
DRaaS is not the outsourcing of the DR process, it is the outsourcing of three critical components of that process; location; compute and storage. The DR process itself is still executed by the organization’s IT staff, the major change that DRaaS brings is that organizations no longer need to have fully functioning secondary sites or establish relationships with a managed hosting providers.
With DRaaS there is still a data movement component. Data has to be backed up or replicated to the cloud provider. But since that connection is through the Internet, it is often slower than a traditional connection. The DRaaS advantage comes when there is a failure. In the event of a disaster, the provider recovers the data that it has under protection and instantiates the organization’s applications in the cloud. If the data was backed up, it is typically recovered from the backup class cloud storage to compute class cloud storage. If the data was replicated, often the provider can recover nearly instantly on the same storage system.
DRaaS is one of the best use cases for the cloud as proven by its popularity, but it is not without its areas of concerns and there are alternatives that may be better for many organizations.
The DRaaS Challenges
The first DRaaS challenge is the data movement challenge. Getting the initial data set from the on-premises data center to the cloud can be an issue. As mentioned above, data has to be moved in every DR use case. Traditional DR used an expensive, point to point connection to route data to the second data center. DRaaS and cloud providers leverage the Internet to make the connection less expensive but that means additional latency.
While all cloud backup and replication utilities have deduplication and compression, these techniques are most valuable only after the initial seed of data has completed.
The second challenge is recovery. Disaster comes in different degrees, ranging from a server failure to full site disaster. The first recovery problem that DRaaS solutions present is assisting in recovery from a minor disaster like a server or storage system failure. In this case the data center is still intact but the organization is facing one or multiple application outages. From a user perspective it is a disaster in every sense of the word. Most DRaaS solutions treat all disasters as major, data loss events and require that the impacted applications are recovered in the cloud. Recovering the application in the cloud is overkill for a minor disaster and unnecessarily introduces failback issues that need to be dealt with.
The second recovery problem, no matter the degree of the disaster is what the DRaaS solution has to do to position data so that the application can be restarted. How data is stored during the protection process directly impacts the time required to restart the application and what the performance of the application might be.
For example, if data is sent to the DR site via a backup application it may be stored on very cost effective, high capacity storage. But that storage is also very slow, meaning it is likely not suitable to host the application. That means the data needs to be transferred to another storage device more suitable for production data. The impact is of course the time it takes to position data to start the application.
The third challenge is the return to the primary data center. At some point the disaster will pass and the primary data center will be suitable to host applications again. Data that was protected in the cloud, and now thanks to DRaaS changed in the cloud, needs to be recovered to the primary data center. To return data to the primary data center will require data be sent to it while the applications continue to run in the cloud. Once data is completely sent to the primary data center, a final sync can occur and allow for a shutdown of the cloud instance. The time and potential for data being out of sync are areas of concern that IT needs to contend with.
Despite these challenges, the DRaaS is a significant step forward in data protection but there is an alternative that organizations should consider, using the cloud as primary storage.
What About Using the Cloud as Primary Storage?
Since data movement is a common denominator to these challenges, an alternative solution is to use the cloud as primary storage. If the data center only used a temporal copy of data and essentially cached data to the cloud, then much of these data movement issues can be resolved. Vendors brought out solutions that cached data in flash on-prem and then as the data aged moved it to the cloud. The problem is the latency difference between on-prem and the cloud was too high and the use cases were limited to file-shares and archives.
A potential solution is to implement another layer between on-prem and the cloud by implementing a metro point of presence (PoP) that is less latent. The three-layered architecture would use an on-premises all-flash cache, the metro-PoP and the cloud as the final storage layer. All new or modified data is stored on the on-prem cache and in the metro-PoP, then eventually replicated to the cloud.
The DR Advantages of Primary Cloud Storage
The metro PoP not only solves much of the latency concerns of cloud storage, it also provides some disaster recovery advantages. First, all data that is on-prem is quickly stored off-site. At a minimum it is at the metro PoP within milliseconds, and more than likely it is stored in the public cloud, and then replicated to a second cloud store minutes later. This architecture provides the organization with a triple layer of protection of active data in multiple locations.
The second advantage is recovery. Data is redundant and protected in the network from the on-prem cache, to the metro PoP for hot and warm data and all data is backed up to the cloud for availability anywhere it’s needed. Having all data available in the network makes recovery straightforward in the event of a minor disaster like an application server failure or a storage system failure where data can be called back from the cloud and metro PoP to resume normal operations. And, in the event of complete data center loss or regional outage, having a secondary edge cache available, either physical or virtual, allows data to be surfaced there as soon as the admin can get the applications up and running. In any of these cases data is stored in an active and accessible state, there is not conversion out of a backup format.
The final advantage is return. When the primary data center is available again, there is no data to restore to the primary site since it didn’t have any unique data in the first place. The cache is simply repopulated based on access. The result is that movement from the cloud is fast.
DRaaS brings many benefits and organizations should explore this new DR option. But for organizations willing to take another step forward, using the cloud as primary storage brings not only the cost savings of cloud storage but also unique DR advantages that surpass even what DRaaS has to offer.
Sponsored by ClearSky Data