Tips for VMware Disaster Recovery
By Samuel Calder, Zerto Pre-Sales Engineer
What is the best way to replicate data in VMware? It is always best to take the direct route and go directly to the source. Avoid detours and get your data from point A to B by performing hypervisor based replication.
There are many attractive reasons for going from a physical environment to a virtual environment. A few advantages are:
— Resource optimization
— Business continuity
— Centralized control and management
If we take a look at traditional methods for replication, then we can see the steps required and road blocks that we need to overcome to replicate effectively in a virtual environment.
Array based replication is performed at the storage layer, which limits granularity and replicates the whole LUN
- Requires frequent coordination between Storage, Network and VMware teams
- Presents architecture design challenges to meet the storage requirements between both sites
Snapshot-based replication often requires a significant local storage overhead and impact to the production environment
- VMs may freeze for long periods leaving applications inaccessible to users of IT
- VM snapshots for replication can conflict with backup products that also require snapshots
Guest-based replication goes inside the guest operating system of each VM and requires agents, which use OS and VM resources
- Performance of the protected VMs is impacted because of the increase in disk I/O
- Very difficult to manage and scale due to agents per guest
The methods above all require multiple components that need to be coordinated to get the job done. Each component adds a level of complexity that takes away from at least one of the advantages of virtualizing; resource optimization, business continuity, and/or centralized control and management.
Now let’s take a look at replicating at the source via the hypervisor.
- No impact to production VMs and can be managed by the virtual team rather than multiple teams needing to coordinate efforts
- Snapshots are not used and block level replication is performed in near-real time, creating significantly lower RPO’s down to the single digit seconds
- Underlying storage becomes a non-factor and agnostic, so you have the flexibility to make the decisions that the business demands
Why not use snapshots? Many snapshot based technologies require database administration, such as quiescing the database prior to the snapshot being taken to avoid issues that may arise with snapshots such as database inconsistency and failures. By performing replication at the hypervisor, there is no need to perform database administration to meet the demands of a snapshot. There are no snapshots. Also, snapshots only replicate to a point in time that is some fifteen minutes ago, at best.
With hypervisor-based replication your environment retains crash consistency; continuous replication via the hyper-visor layer maintains write-order fidelity, ensuring the disks in the target site are always crash consistent. Only confirmed writes are replicated, which means data is only replicated when the OS has confirmed the change has been written to the local disk. This also means that you are able to go back to a point in time only seconds ago, minimizing the amount of data loss.
By performing replication at the hypervisor layer, your replication strategy allows you to take full advantage of the benefits of virtualization without sacrificing on your business continuity process.