When did you last test your DR failover? The most common answer we hear: "never" or "when we first set it up."
That matters, because Azure Site Recovery isn't cheap. And if you've never validated it, you're spending thousands on a safety net that might not catch you.
What ASR Actually Costs
The headline price is ~£20 per protected instance per month. For 40 VMs, that's £800/month just for the licence. But that's only the start.
Every protected VM has replica managed disks in your target region, same size and same tier as the source. Premium SSD source means Premium SSD replica.
Then there's the cost most people miss: inter-region replication bandwidth. ASR continuously replicates data changes between regions. Every write to a protected disk generates cross-region traffic, and Azure charges for egress between regions. For busy workloads with significant daily change rates, this adds hundreds per month. It never shows up on the ASR line item. It appears as networking cost, so nobody connects it to DR.
Add vault storage on top, and the total DR cost for 40 VMs commonly lands between £1,500 and £3,000 per month.
We regularly see DR-related costs accounting for 5-10% of total Azure spend.
The "Protect Everything" Mistake
The most expensive pattern: every VM gets ASR, no exceptions. It feels responsible. It's also wasteful.
Not every VM needs ASR. The question is: can this workload be recovered faster by redeploying from code than by failing over from a replica?
- Stateless VMs behind a load balancer. Redeploy from Terraform in minutes. ASR is slower than the alternative.
- Infrastructure deployed from IaC. The VM isn't valuable, the data and config are. Code-based recovery is faster and cheaper.
- Dev/test environments. Nobody is invoking DR for a dev box. Yet we see it constantly.
- Circular replication. VMs in the recovery region replicated back to primary. Doubles cost for zero benefit.
Premium Replicas for Non-Premium Recovery
ASR mirrors source disk tiers by default. But replica disks sit idle 99.9% of the time. They only matter during failover.
Switching replicas from Premium SSD (£100/month for 1TB) to Standard SSD (£60/month) saves hundreds per month across an estate. The performance trade-off during failover is acceptable for most workloads.
RTO and RPO: The Conversation IT Needs to Have with the Business
Before deciding what gets ASR, you need two numbers per workload, and they shouldn't come from IT alone:
- RTO (Recovery Time Objective): how long can this workload be down before the business is materially impacted? Minutes? Hours? A day?
- RPO (Recovery Point Objective): how much data loss is acceptable? Zero? An hour's worth? Yesterday's backup?
ASR gives you low RTO (minutes) and low RPO (near-zero data loss). Azure Backup gives you higher RTO (hours) and higher RPO (last backup point). The cost difference is significant.
The problem: IT teams often set RTO/RPO based on what feels safe, not what the business actually needs. A dev environment doesn't need a 5-minute RTO. A reporting database that refreshes daily doesn't need near-zero RPO. But unless someone asks the business owner "what actually happens if this is down for 4 hours?", everything gets the premium treatment.
Get the business to sign off on RTO/RPO per workload. You'll find most things need far less protection than they're getting.
ASR vs Azure Backup
- ASR: Continuous replication, minutes to recover, near-zero data loss. For genuinely critical, stateful systems where RTO and RPO are measured in minutes.
- Backup: Periodic snapshots, hours to recover, cheaper. For operational recovery like accidental deletion, corruption, or rollback where RTO of a few hours is acceptable.
If "back within a few hours" meets the business requirement, Backup at a fraction of the cost is the right answer.
Tier Your Workloads Honestly
Most organisations find only 30-50% of their VMs genuinely need ASR:
Tier 1, ASR: Databases, transaction systems, workloads where data loss is unacceptable. Pay for it, but check replica disk tiers.
Tier 2, Backup only: Important but tolerant of hours-long recovery. Documented procedures plus appropriate retention.
Tier 3, Rebuild from code: Stateless infrastructure, anything deployed from IaC. Your Terraform state file is more valuable than an ASR replica.
Tier 4, Nothing: Temporary workloads, proof-of-concept, batch jobs that restart.
Test It or It Doesn't Count
Untested DR might not work. Configurations drift, NSGs in the target region don't match, DNS doesn't update, dependencies don't resolve.
A quarterly test failover costs a few hours of VM compute. Trivial compared to discovering your DR is broken during an actual outage. If you're paying for ASR, make sure it delivers.
Not sure what your DR is actually costing you? Request a FinOps assessment. We break down ASR spend and identify over-protected workloads.