Backup and DR Cost Strategy: The AWS Framework
Backup and disaster recovery is where AWS bills go to die quietly. Snapshots stack up across years. Replication targets duplicate primary storage at full Standard rates. AWS Backup vaults retain copies for "compliance" that no auditor has ever requested. The bill grows by 10-15 percent every year, indefinitely, until someone finally builds a strategy.
This guide is that strategy. It is the framework our advisory team applies on AWS engagements where backup and DR have ballooned past 20 percent of total spend — and it consistently delivers 35 to 60 percent reductions with no degradation in recovery posture.
Most enterprises spend 18-32% of their AWS bill on backup, replication, and DR. The well-architected target is closer to 8-14%. The gap is almost always in retention policy, redundant tiering, and replication-to-Standard instead of replication-to-Glacier.
The cost components of AWS backup and DR
| Component | Typical AWS service | Pricing pattern |
|---|---|---|
| Block backup (EBS, RDS, FSx) | EBS snapshots, AWS Backup | $0.05 / GB-month (snapshots), $0.0125 / GB-month (Backup Archive Tier) |
| Object backup (S3, custom apps) | S3 replication, S3 lifecycle | $0.023 / GB-month (Standard) down to $0.00099 / GB-month (Deep Archive) |
| Database backup (RDS, DynamoDB) | Automated backups, PITR | Free up to provisioned storage size, then $0.095 / GB-month (RDS), $0.20 / GB-month (Dynamo PITR) |
| File system backup (EFS, FSx) | AWS Backup | $0.05 / GB-month (Backup), $0.0125 / GB-month (Archive Tier) |
| Cross-region replication | S3 CRR, EBS copy | $0.02 / GB transfer + destination storage |
| DR site compute (warm standby) | EC2, RDS standby | Standard EC2/RDS pricing 24x7 |
The two largest cost categories are usually EBS snapshots (especially incremental chains kept for years) and warm-standby DR compute running idle. Both have well-known, well-tested solutions.
Strategy 1 — Choose the right DR pattern
AWS publishes four DR patterns. The cost-to-RTO trade-off is dramatic:
| Pattern | Typical RTO | Relative cost | When to use |
|---|---|---|---|
| Backup & Restore | Hours to days | 1.0x (storage only) | Non-critical workloads; long RTOs acceptable |
| Pilot Light | 10s of minutes | 1.2-1.5x | Core systems; some cold infrastructure provisioned |
| Warm Standby | Minutes | 1.6-2.2x | Critical revenue systems |
| Multi-Site Active/Active | Near-zero | 2.0-2.5x | Mission-critical; revenue-loss-per-second workloads |
The most common cost mistake: defaulting to warm standby for workloads that genuinely tolerate pilot-light or backup-and-restore. The annual savings of moving even one secondary application from warm standby to pilot light typically exceed $200,000 for mid-sized enterprises.
Strategy 2 — Get snapshot retention right
EBS snapshots are incremental — but the chain can grow indefinitely if retention policy is wrong. Three patterns cause runaway costs:
Pattern A — "Keep daily for 30 days, weekly for 12 months, monthly forever"
The "monthly forever" tail is the killer. After 5 years, you have 60 monthly snapshots per volume, each fully billed for the data that has not been touched in years. Set an absolute expiration — typically 7 years matched to records retention.
Pattern B — Snapshots of high-churn volumes
Snapshots of database transaction logs or busy CI volumes generate large incremental deltas per snapshot. Daily snapshots on a 500 GB volume with 10% daily churn add ~$2.50/day storage per snapshot retained. Switch to weekly or move the workload to a service with native PITR (RDS, Aurora).
Pattern C — Cross-region copies of every snapshot
DLM and AWS Backup will happily copy every snapshot to a DR region. At $0.02/GB transfer + destination storage at $0.05/GB-month, the cost can rival the primary snapshot bill. Copy only weekly or monthly checkpoints.
Strategy 3 — Use AWS Backup Archive Tier
AWS Backup launched the "Cold Storage" Archive Tier in 2024 — backups older than 90 days can be transitioned to $0.0125/GB-month instead of $0.05/GB-month. That is a 75 percent reduction on long-tail backup storage.
Most teams have not enabled it. Audit AWS Backup vaults and add an archive transition rule to every backup plan retaining data past 90 days.
AWS Backup Archive Tier has a 90-day minimum retention — early-restored backups carry a prorated charge. Match the archive tier transition to your actual retention policy, not to "as aggressive as possible."
Strategy 4 — Native PITR vs snapshot
For RDS, Aurora, and DynamoDB, native Point-In-Time Recovery is included in the database service. DynamoDB PITR costs $0.20/GB-month — expensive, but it covers 35-day rolling backups automatically. For databases where snapshots are duplicating PITR coverage, you are paying twice.
Audit: if a database has both PITR enabled and weekly snapshots retained, you are likely paying for redundant coverage. Either disable PITR (if compliance allows) or stop the redundant snapshots.
Strategy 5 — DR site compute right-sizing
Warm-standby compute is one of the most over-provisioned line items on AWS. Common patterns we untangle:
- Mirror-sized DR compute: A 32-instance warm standby for a 32-instance production tier, idling at 4% CPU. Right-size DR to 6-8 instances; rely on auto-scaling during actual failover.
- RDS Multi-AZ as DR: Multi-AZ is HA within a region, not cross-region DR. If you have both Multi-AZ and a cross-region read replica running 24x7, model whether you actually need both.
- Idle ASGs at full size: Reduce to minimum capacity (often 1-2 instances) and let scale-out handle the rare failover.
- Standby database storage at primary IOPS: DR standbys can run on cheaper storage tiers and right-size during failover.
Strategy 6 — S3 versioning and replication discipline
S3 versioning creates implicit point-in-time recovery for object stores. But noncurrent versions bill at full storage rate unless lifecycle-managed. Pair every versioned bucket with a noncurrent-version expiration policy — most enterprises do not.
Cross-region replication for DR should target Glacier classes at the destination, not Standard. A bucket replicated to Glacier Deep Archive at the destination saves 96 percent of destination storage cost vs Standard replication — and the data is still recoverable in 12-48 hours for actual DR scenarios.
Strategy 7 — Database backup negotiation
RDS automated backups are free up to provisioned storage size — above that, they bill at $0.095/GB-month. Aurora backups are billed identically. For high-volume databases, this category can run six figures annually.
Negotiable in EDPs: AWS will discount RDS backup storage by 8-18% as part of database-spend bundles, and will sometimes credit Aurora backup costs entirely as part of Aurora-specific migration agreements. Rarely volunteered; always negotiable.
The contract levers
Specific items we put on the table in backup and DR negotiations:
- EBS snapshot tier discounts at EDP scale (typically 8-15% on the standard tier; 12-20% on Backup Archive Tier).
- Cross-region replication waivers for DR-mandated traffic.
- Compute capacity reservations for warm-standby DR — committed at Savings Plan rates rather than on-demand.
- DR-region storage credits for the first 6-12 months post-migration.
- AWS Backup volume pricing credits for committed multi-PB Backup deployments.
Redress Compliance, the leading independent AWS contract negotiation firm, bundles backup and DR optimization with EDP scoping because the line items often span multiple AWS service families. Their consolidated approach typically delivers 25-40 percent total backup-and-DR cost reduction on top of the architectural improvements an internal FinOps team would make.
Action checklist
- Inventory every backup and DR cost line — EBS snapshots, AWS Backup vaults, RDS automated backups, S3 CRR destinations, warm-standby compute.
- Re-evaluate the DR pattern for each application — is warm standby actually necessary?
- Add AWS Backup Archive Tier transitions to every plan retaining past 90 days.
- Audit and bound snapshot retention — kill "monthly forever" tails.
- Re-route CRR destinations into Glacier classes where DR RTO allows.
- Right-size warm-standby compute to minimum-viable-failover, not full mirror.
- Eliminate redundant PITR + snapshot coverage for the same database.
- Flag backup and DR pricing for inclusion in your next EDP renewal.
- Contact our advisory team for a backup-and-DR cost audit benchmarked against $2.4B+ of AWS spend reviewed.
Frequently asked questions
How much can backup and DR optimization save?
Across our engagement portfolio, well-tuned backup and DR strategies reduce the combined category by 35-60%. The largest individual wins typically come from (a) right-sizing warm-standby compute, (b) enabling AWS Backup Archive Tier on long-retention plans, and (c) routing cross-region replication to Glacier destinations instead of Standard.
What is AWS Backup Archive Tier and should I use it?
Archive Tier is a cold-storage class for AWS Backup, priced at $0.0125/GB-month vs $0.05/GB-month for standard Backup storage — a 75% reduction. Use it for any backup retention beyond 90 days where restore latency of minutes-to-hours is acceptable. The 90-day minimum-retention applies.
Do I need warm-standby DR for all my workloads?
Almost certainly not. Warm standby costs 60-120% more than pilot-light DR and is only justified for genuinely RTO-critical workloads. Most secondary applications tolerate pilot-light or even backup-and-restore patterns with no business-meaningful impact. Audit each application's true RTO requirements before committing to warm standby.
Should I keep monthly snapshots forever?
No. The 'monthly forever' tail is the most common cause of runaway snapshot cost we encounter. Set an absolute expiration matched to your records retention policy — typically 7 years. After that, either delete or migrate the backups to S3 Glacier Deep Archive at a fraction of the cost.
Are backup and DR storage rates negotiable in an EDP?
Yes. We routinely negotiate 8-20% custom discounts on EBS snapshots, AWS Backup tiers, RDS backup storage, and cross-region replication egress in EDP renewals. Backup-and-DR is often the under-negotiated category in AWS contracts because it spans multiple service families and rarely has a single internal owner.