Gajus - Fotolia
VMware's acquisition of Datrium in 2020 strengthened its disaster recovery-as-a-service products with cloud-native DR that enables IT administrators to replicate on-premises vSphere VMs to AWS.
Datrium started as an integrated hardware and software storage appliance provider. It combined licensed software and shared storage shelves to enable solid-state drives (SSDs) to provide storage performance and data persistence in ESXi hosts. Datrium also developed replication between clusters for on-premises disaster recovery.
The company then focused on cloud-based object storage replication for cloud-native DR in VMware Cloud on AWS. This cloud-native DR uses VMware Cloud on AWS as the restore location, which means a system can replicate and recover on-premises vSphere VMs to cloud-based vSphere instances.
Datrium repackaged its software for replicating vSphere VMs into a virtual appliance following limited market adoption of its hardware and software storage offering. As a result, Datrium's cloud-native DR became compatible with any vSphere system. This cloud-native DR capability led to VMware acquiring Datrium in July 2020, and VMware released Datrium's disaster recovery as a service (DRaaS) offerings for VMware Cloud on AWS.
Datrium helps bolster VMware DRaaS functionality
VMware DRaaS paired with Datrium technology enables a system to replicate on-premises vSphere VMs to AWS Simple Storage Service, which is a cloud-based object storage service. For existing clusters, the system can recover VMs by spinning them up as native vSphere VMs on a VMware Cloud on AWS cluster within a few minutes. But the recovery process might take longer if admins require a new VMware Cloud on AWS cluster or more nodes.
The vSphere VMs then run from a temporary Datrium storage cluster. Admins usually migrate these VMs to the VMware Cloud on AWS storage and shut down the temporary cluster. The Datrium storage cluster uses a combination of storage types -- SSDs and RAM -- to provide performance to run VMs that remain on object storage before a migration to VMware Cloud on AWS. This setup helps achieve transactional performance.
When admins return the VMs to on-premises vSphere, they only have to replicate the changed data blocks, rather than the entire data set -- provided the pre-failover VMs are still available. A virtual appliance handles deduplication, compression and replication processes for on-premises vSphere setups.
The virtual appliance stores replicated data on cost-effective object storage. This removes the need to copy VMs or convert them from one format to another before the VM powers on, which is central to meet a rapid recovery time object (RTO) and a simple failback.
The business value of VMware DRaaS
The reality of DR is it's an unwelcome expense until needed. Then, DR becomes the most valuable asset for admins and organizations to have. The trick is to minimize the monthly cost of a DR product without compromising the asset's value at recovery.
The most significant benefit of VMware DRaaS is that admins can maintain relatively low RTO without high cost. Usually, up-to-date VM replicas must permanently reside on expensive transactional storage to achieve low RTO. But VMware's integration of Datrium technology into its DRaaS tooling means replicas reside on object storage, which incurs lower costs per month.
The higher cost and higher performance transactional storage is only used from failover until migration to VMware Cloud on AWS. An organization only pays for higher storage performance when they use that performance. After migration, VMware Cloud on AWS provides the storage performance through vSAN as part of its service.
Dig into VMware's DR cloud offerings
There are two fundamental approaches used in DR cloud products for vSphere: Replicate VMs to an active vSphere cluster for rapid recovery or replicate VMs to object storage for lower monthly costs.
VMware's Site Recovery Manager replicates on-premises VMs in vSAN to VMware Cloud on AWS. Because the VMs already reside in vSAN and replicate to an active vSphere cluster, admins can achieve fast RTO -- they just register and power on the VMs.
The drawback is how the replicated VMs reside in vSAN at all times, which is relatively expensive transactional storage. The replication of VMs to object storage provides a lower storage cost. However, it's not suitable to run VMs directly on object storage because of its performance characteristics.
Most DR products that replicate to object storage require conversion to different storage when admins must recover VMs, which results in a failover process that's similar to a restore from backup. For some DR products, the restore process involves VM recovery on a vSAN data store as usual.
Other DR products restore to an Amazon Elastic Compute Cloud instance, rather than a vSphere VM, which removes the need for a VMware Cloud on AWS cluster. Still, this path often comes with a confusing failback trail to on-premises vSphere.
Keep in mind that when a DR event occurs, admins must recover a large number of VMs. Ensure the DR product executes parallel operations, rather than serial, so the recovery time isn't directly related to the number of VMs.