Is it time for cloud disaster recovery?

Questions You Should Be Asking About Cloud Disaster Recovery in 2020

Disaster recovery remains ever present in the minds of IT. As we enter 2020 the challenges facing IT professionals as they establish a disaster recovery plan are more daunting than ever. There is more data to recover, many more applications to recover and the expectations of the organizations are higher than ever. Unfortunately, the funds and staffing to drive the disaster recovery effort are either flat or shrinking.

The cloud seems like an ideal way  for IT to cross the chasm and make everyone happy. As a result, new vendors and legacy backup vendors are adding Disaster-Recovery-as-a-Service (DRaaS) to their offerings. There are even some primary storage vendors claiming you don’t need backup and that they can provide cloud DR without backing up data first.

The problem is not all DRaaS solutions are the same, and some don’t provide sufficient coverage or functionality to meet DR requirements. There are some key questions IT pros should be asking their current and potential future data protection providers about cloud DR. The capability is more than a checkmark item. You need to dig below the single line feature checkmark to really see what the functionality of the solution truly provides.

Push button DR does not mean instant DR

Many data protection vendors claim some sort of recovery in the cloud, which means they use cloud compute to create a software defined version of your data center in a cloud provider’s data center. What is unclear from these vendors is how long, from the moment you initiate a cloud recovery, before your application is available for users to log in, and how much data is lost in the process.

The reason for the vagueness is that if you run an application in the cloud you typically need to convert it to run under the cloud provider’s hypervisor rather than the hypervisor you are running on-premises. What vendors, who have a cloud DR option, often don’t tell you is that it can take hours to convert a VM. In our testing, converting even a small VM took four hours. Most critical applications are going to have a recovery point objective (RPO) that is less than four hours, so instead of counting on their cloud backup solution, IT is forced to look elsewhere. There are workarounds, but each of them quickly becomes expensive. Another challenge with converting VMs is that at some point you have to convert it back when you return on-premises. This is another time consuming process.

IT professionals need to ask their potential cloud DR provider how long it will take from the moment they declare a disaster and start the process till the moment applications are up and running and users are able to log in. Many vendors claim push button recovery but most don’t explain how long the entire process takes.

Can you recover from backup storage?

The cloud storage choices that backup software vendors make are also something to consider. Most, if not all, cloud backup solutions store data on the cloud providers’ primary object store. In the case of Amazon, this tier is Amazon S3. Each cloud provider has its own flavor. Each cloud provider also has a higher performance tier to run production applications, and each provider has an extremely low cost tier for older data. Many backup providers are not effective at managing these various tiers and default to the middle tier which means their customers are paying more for storage than they should. It also means that during a disaster they may not get the performance they need.

For most disaster recovery situations, the organization will want to run their application on the production class storage tier (EBS). The challenge is getting that data to that tier in a timely manner. IT professionals need to make sure that they ask how long it will take to transfer the data from the backup storage tier to the high performance tier. They also need to ask if the vendor provides a way to mitigate the transfer times.

Will it compute?

A final question relates to available compute resources. In a disaster situation will the provider have enough compute to support your entire data center’s computing needs? In a regional disaster the provider may have to support not only your data center’s needs but the needs of hundreds or thousands of customers. Will there be enough compute? If your provider is using one of the big three public clouds then the answer is more than likely, “yes.”

Many cloud DR providers though, create their own cloud or count on the resources of a local managed service provider. In both cases the issue of available computing power becomes a bit more concerning. More than likely they have enough computing power if just your data center has an outage. In the case of a regional disaster, however, where a whole city’s worth of businesses declares a disaster because a hurricane, flood or fire is bearing down on it, there may be a  shortage of compute resources.

Ransomware is also a concern. Even though it doesn’t target a specific city per se, it is often successful in infiltrating an entire agency or even multiple agencies across a state. In most cases, states use a shared DR facility. Making sure these facilities can handle the compute load is critical.

Finding the right cloud disaster recovery solution

With Druva, you don’t have to worry when a disaster strikes. Your organization can bounce back in minutes with cloud-native, one-click automated disaster recovery for on-premises and cloud workloads. Druva recovers VMs in a matter of minutes in your AWS VPC, and VPCs can be cloned across regions/accounts for greater resiliency. VMs are converted to AWS EBS snapshots that are kept-at-the ready in the customer’s AWS VPC for immediate spin-up of AWS EC2 instances at the time of failover. Not to mention, Druva’s cloud-native disaster recovery saves up to 50 percent lower TCO, eliminating the need for on-premises DR hardware.

Learn more here: