As a leading cloud provider, our goal is to inform customers with real information. Setting expectations and being transparent has been a pillar of our success. With Zerto being the fastest replication product in the market today for disaster recovery, we thought it would be interesting to compare its recovery in AWS, Azure, Google Cloud, and our own cloud: PvDC.
To get straight to the point – Google Cloud is going to be eliminated right off the bat simply because Zerto does not support it today.
Now down to three: AWS, Azure and PvDC
THE SET UP
AWS and Azure are instance based, so you have to assign every workload to a matching instance size per VPG (Protection Group). This has to be pre-planned for every workload and any changes to those workloads requires a re-evaluation of the VPG instance size. Any changes may cause a complete rebuild and full sync for those affected VPGs.
On networking, the best setup for either is utilizing BGP VPNs. These enable seamless changes at the source sites which pass those changes to the target DR sites. BGP VPNs are great on AWS; however VPCs cannot cross regions. This means a complete separate VPN and Zerto setup at every region you may want protected or targeted. You can however cross regions in Azure with BGP VPNs. Zerto ‘pairing’ is possible between Azure regions but not possible between any AWS region.
Also note: For AWS and Azure , Public IPs are required for every Zerto ZVM in order to "call home" - which....is an added cost.
Net'3 PvDC has none of the limits above. There’s no re-configuration of VM instances and no re-sizing VPGs for Zerto to connect to PvDC Cloud. With PvDC, you receive a bucket of resources and connect your data in the way that is easiest and makes the most sense.
Failover/Failback performance measurements we found:
Why does AWS take so long?
Because AWS is not using real time replication from Zerto, a core strong suit of Zerto. Instead, it takes snapshots like a backup. In order for the failover to work properly, AWS temporarily uses 2 VMs per source machine. Due to this 2 VM split storage process, the recovery takes longer. That’s an additional 2-3 steps in the recovery process in order for machines to come up. Since recovery is using 2 machines, the cost for each machine doubles during the recovery window.