Returning workloads on-premise,
as a managed project.
Organizations adopted the public cloud for elasticity and now carry costs that grow faster than the business. For steady-state workloads, repatriation is frequently the lower-cost position — provided those workloads can be relocated without interruption.
Three reasons the arithmetic has changed
Predictable workloads pay a premium
Elastic pricing is worth paying for where demand varies. A database that has operated at a constant size for three years is paying for an option it does not exercise.
Egress charges compound
The cost of departure increases with every terabyte stored. Deferring the decision increases its price.
Data sovereignty is now a requirement
Regulators, customers and boards now ask where data physically resides, and which jurisdiction could compel its disclosure.
How a repatriation runs
The same engine relocates the workload and then protects it. A single procurement covers both; there is no second exercise for disaster recovery once the migration completes.
Land the storage
S3 Ocean is installed on your own premises, speaking the same S3 API your applications already use. No application changes.
Replicate rather than copy
S3 Bridge and VM Bridge replicate continuously while production continues. Up to 500 machines in parallel.
Cut over in seconds
Disk formats convert automatically, VLANs and subnets are preserved, and the reverse path remains available until you close it.
Protection switches on
Vault DR and Vault Backup take over the moment the workload lands, under the four-hour recovery commitment.
What the audit measures
Completed during the audit from your own invoices. Where the payback does not hold, the document states so; that finding carries the same value as the alternative.
Storage, per TB per month
Current cloud cost against modelled on-premise cost at your actual capacity.
Egress and request charges
Current spend, against none once the data has landed.
Compute, steady state
Current cloud cost against modelled on-premise cost for workloads that do not flex.
Disaster recovery
Current spend, against protection included in the platform.
One-off cost to move
Modelled, including the egress charge to get the data out.
Payback
Expressed in months, with the assumptions written down.
When repatriation is not advisable
- Genuinely variable demand, where burst capacity is what the hyperscalers provide well.
- Workloads deeply integrated with managed services that would require rebuilding.
- Organizations without the capacity to operate hardware, and without a partner to operate it for them.
Most estates are mixed. The useful question is not whether to repatriate, but which portion of the estate should return first.
Establish what repatriation would cost
Two weeks, at no charge, and the model remains yours whatever you decide.