The VMware Exit Is a Protection Upgrade

The decision to leave VMware presents an opportunity to reassess data protection strategies within the data center. Most teams plan to transfer their backup design as is and rebuild it on the new hypervisor, but a better approach would be to leverage the exit to enhance resiliency and accelerate recovery, while retaining the existing backup application.

Currently, backup designs in most data centers rely on vSphere mechanics, including features such as changed block tracking, vStorage APIs, and VMware snapshot behavior, as well as storage integration and proxies at the cluster level. Transitioning to a different hypervisor will necessitate a reevaluation of these assumptions. Major backup vendors have made significant advancements in supporting KVM-based platforms in recent years.

The critical question arises upstream of the backup application. Traditionally, production workloads ran independently, while backup functions were handled by separate tiers. These processes included resilience, rapid rollback, replication, site failover, and immutable copies, each managed by distinct products with their own licenses, consoles, and renewal schedules.

This division of labor had its merits, as server virtualization abstracted individual servers, allowing virtual CPUs, memory, and disks to be presented to guests while storage resilience and recovery were handled by storage arrays and backup products separately. However, the emergence of hyper-converged infrastructure (HCI) blurred these lines, integrating storage into servers and maintaining the same separation of responsibilities.

This arrangement often involved running controller VMs on each node. When a backup tier managed everything from failed drives to site loss disruptions, the consequences of a single site failure became evident. Such an incident impacted multiple components simultaneously, complicating recovery efforts. The exit provides a rare chance to reconsider these assumptions.

Datacenter virtualization abstracts the entire environment, encompassing compute, storage, networking, and data protection under a unified code base and metadata system. Consequently, the platform gains a comprehensive understanding of workloads rather than fragmented views. This shared perspective enables production to assume responsibilities that were previously delegated to backup systems.

With fewer events reaching the backup tier, hardware failures become straightforward maintenance tasks, and software issues are resolved through rapid snapshot rollbacks that barely affect application owners. The role of backup expands in tandem with this shift. Initially, migration work will focus on transitioning day-to-day recovery responsibilities to the production platform.

A backup application capable of supporting both the legacy hypervisor and the new one will utilize existing backups as the migration pathway. Veeam Backup & Replication exemplifies this approach, now supporting VergeOS as a platform. This enables organizations to restore their vSphere backups onto the new environment, test them, and schedule the cut-over as needed.

Once the migration is complete, the backup application retains its strengths, such as long-term retention, compliance archives, air-gapped and offsite copies, and granular recovery options for application owners, all while remaining outside the production platform's failure domain. Evaluating the platform side of this transition underscores the importance of evidence-based decision-making.

Concentrating responsibility on a single vendor's engineering is a trade-off that should be scrutinized against real-world scenarios. Request vendors to demonstrate their solutions by subjecting them to failure simulations, including drive removals, simultaneous drive failures exceeding protection limits, VM deletions, and site migrations.

A vendor capable of executing these tests proves their reliability, while those relying on theoretical diagrams lack practical evidence. Following this evaluation, conducting a test migration through the backup application and verifying the restored VMs' functionality on the new platform is crucial. The core decision presented by this exit revolves around the division of protection responsibilities between production and backup systems.

This choice will determine the environment's operational costs over the next five years and its resilience during critical failure scenarios. Reconstructing the outdated division of labor is feasible and merits thorough examination, comparable to the scrutiny applied to licensing agreements. VergeIO, for instance, has developed VergeOS based on the datacenter virtualization model, integrating resilience, snapshots, instant recovery, and site failover within the same codebase as compute, storage, and networking.

VergeOS support in Veeam Backup & Replication facilitates migration and the subsequent backup tier. A recent session led by Aaron Richman and Dave Vincent demonstrated the effectiveness of these approaches through comprehensive failure scenarios and was made available for viewing on demand.