This guide is for a CTO, head of engineering or IT lead whose product runs on a public cloud and who has to defend a recovery plan to the business. It sits under cloud platform engineering, and the day-to-day running of the chosen tier is part of managed cloud services.
In this guide
- What are the four disaster recovery tiers?
- What do RTO and RPO actually measure?
- How do you turn business impact into RTO and RPO targets?
- Which tier fits which target?
- When is the cheaper tier the correct choice?
- Why do replication tiers still need backups?
- Choosing and proving a tier
- Where AI fits in disaster recovery planning
- What delivery record exists, and what does not
- Alternatives: other ways to reduce outage impact
- Limits of this guide
- Frequently asked questions
- Next step
What are the four disaster recovery tiers?
The four tiers differ in how much of the system already runs in the recovery region before anything goes wrong. The AWS disaster recovery whitepaper orders them from the lowest cost and complexity to the highest, and each step up shortens recovery.
- Backup and restore: only backups sit in the recovery region. At failover the team rebuilds the infrastructure, deploys the code and restores the data, so infrastructure as code is what keeps the rebuild inside the target.
- Pilot light: data is replicated continuously and the core, such as the database, is always on. Application servers are defined but switched off, so failover means switching them on and scaling out.
- Warm standby: a scaled-down but fully working copy of production runs in the recovery region. It can take traffic at once at reduced capacity, and failover means scaling up.
- Multi-site active-active: the workload serves users from more than one region at all times. There is no failover in the usual sense; losing a region means routing traffic to the regions that remain.
Microsoft and Google use different words for the same ladder. Microsoft's disaster recovery strategy guide calls the middle options cold and warm standby, and Google's planning guide speaks of cold, warm and hot patterns. The names matter less than one question: what is running before the disaster, and what must the team still do after it?
What do RTO and RPO actually measure?
RTO, the recovery time objective, is the longest time a flow may be unavailable after a disaster. RPO, the recovery point objective, is the largest amount of data, measured in time, that the business accepts losing. The Google Cloud planning guide adds the rule that drives every tier choice: the smaller the two values, the more the application costs to run.
The two targets move separately. An order flow may tolerate an hour offline but almost no lost orders, so it needs a small RPO and only a moderate RTO. A reporting service may lose a day of derived data and simply rebuild it, so both targets can be large. A single target for the whole system hides this difference and usually buys too much recovery.
How do you turn business impact into RTO and RPO targets?
Set targets per business flow, not per server. Ask the business owner of each flow two questions and write the answers down with their name next to them.
-
What happens after one hour, one working day and one week without this flow?
- What the answer sets
- RTO: the last duration the business can live with
- Typical evidence for the answer
- Revenue or order flow per day, contract service levels, regulatory deadlines
-
If the last hour or day of data disappeared, could it be re-entered or rebuilt?
- What the answer sets
- RPO: how much data loss is tolerable
- Typical evidence for the answer
- Whether data arrives from outside (orders, payments) or is derived (reports, caches)
-
Does anything else stop when this flow stops?
- What the answer sets
- Recovery order
- Typical evidence for the answer
- Dependency map: database before application, identity before both
The answers then sort each flow into an impact band. These bands are a working convention for this guide, not an industry standard; agree your own wording with the business.
| Impact band | Typical flows | Target pattern | Cheapest tier that usually fits |
|---|---|---|---|
| Can wait a working day, data can be rebuilt | Internal reporting, back-office tools, staging | RTO and RPO measured in hours | Backup and restore |
| Can wait a short time, new data must not be lost | Customer accounts, order history, content management | Small RPO, RTO of a fraction of a working day | Pilot light |
| Customers notice within minutes | Checkout, customer portal, partner API | Small RPO, RTO in minutes | Warm standby |
| Any interruption breaks a contract or a safety duty | Payment switching, real-time dispatch, regulated systems | RTO and RPO close to zero | Active-active |
Which tier fits which target?
| Tier | Running before the disaster | What the team does at failover | Choose it when |
|---|---|---|---|
| Backup and restore | Backups and infrastructure code only | Rebuild from code, restore data, switch DNS | The flow can wait hours and the rebuild has been rehearsed |
| Pilot light | Replicated data stores; application defined but off | Switch on and scale application servers, promote the database | Data loss must be small but some downtime is acceptable |
| Warm standby | Full stack at reduced capacity | Scale up and move traffic | Customers would notice an outage within minutes |
| Active-active | Full stack serving traffic in two or more regions | Route traffic away from the failed region | Near-zero downtime is a hard requirement and the team can run multi-region writes |
The cost of a standby tier is mostly the capacity that waits idle. Standby capacity that runs all year belongs in the commitment-level decision, and pilot light keeps that waiting capacity smallest of the replication tiers.
When is the cheaper tier the correct choice?
Backup and restore is the correct tier more often than vendors suggest. The AWS whitepaper notes that for the loss of one data center, a well-architected workload spread across availability zones may need only backup and restore; the higher tiers answer the loss of a whole region or a regulatory requirement. Keep the cheaper tier when all of these hold:
- the flow owner has signed an RTO of hours, not minutes;
- the data can be re-entered or rebuilt, or point-in-time backups meet the RPO;
- the workload already runs across more than one availability zone;
- the full rebuild from infrastructure code has been timed in a drill and fits inside the RTO.
Move up a tier when one of these fails and the gap cannot be closed more cheaply, for example by taking backups more often or by splitting one critical flow out of a larger system. Moving one flow up is usually cheaper than moving the whole product.
Why do replication tiers still need backups?
Replication copies bad changes as quickly as good ones. A deleted table, a corrupting deployment or encrypted files reach the standby region within seconds. The AWS whitepaper states that even active-active needs point-in-time backups, and that recovery from data corruption always returns to a point before the problem was found. Every tier therefore keeps versioned, point-in-time backups, ideally in a separate account, and the RPO for corruption is set by the backup schedule, not by replication.
Choosing and proving a tier
-
List the business flows
Name each flow, its owner and the systems it depends on, including identity and DNS.
-
Ask the impact questions
Record each owner's answers and turn them into an RTO and an RPO per flow.
-
Pick the cheapest tier per flow
Use the decision matrix; group flows that share a database into one tier.
-
Write the recovery runbook
Recovery order, who declares a disaster, failover and failback steps, and contacts.
-
Rebuild from code
Keep infrastructure, configuration and machine images in code so any tier can be recreated.
-
Drill and time it
Restore backups on a schedule and run a failover drill; record the measured time and data loss.
-
Review after change
Re-run the drill when the architecture, data volume or flow owner changes.
The drill is where a tier becomes real. The Microsoft guide says production-level drills are the only way to verify that a plan meets its targets, and recommends practising in non-production first. Detection matters too: a recovery clock starts only when someone notices, so the telemetry design is part of the recovery time.
Where AI fits in disaster recovery planning
AI helps with the paperwork around recovery. An AI assistant can read infrastructure code and runbooks, draft the dependency map, flag a resource that exists in production but not in the recovery definition, and summarise drill logs into a first report. Netbase works with the major commercial and open-source models and chooses them per project.
Engineers still own the tier decision, run every drill and sign the measured times. An AI summary of a drill that nobody ran proves nothing.
- In delivered work
Product recommendation engine
Built for 4over4's online printing store from browsing and purchase history; a delivered AI feature named with a client, not a recovery project.
- Available capability
AI agents in delivery
AI-assisted review of infrastructure code and runbooks is a capability Netbase offers, not yet tied to a published recovery case.
What delivery record exists, and what does not
Netbase works as offshore development partner on a multi-tenant cloud ERP delivered as SaaS on AWS for a US client, not named, and builds and runs its own SaaS products, such as Cloodo Workspace. Netbase names AWS, Google Cloud, DigitalOcean and Cloudflare as the providers it uses, and lists disaster recovery among its security practices, next to role-based access control, MFA for admin dashboards and vulnerability scanning.
Delivery runs in a six-step lifecycle with weekly reviews and KPI dashboards, and most Netbase projects are delivered on fixed-price contracts. No published Netbase record names the tier of a client's recovery set-up, and no recovery time, uptime or cost result is claimed.
Alternatives: other ways to reduce outage impact
High availability inside one region, across availability zones, handles the most common failures and often comes before any cross-region tier; it does not protect against a regional outage or bad data. Graceful degradation keeps part of a product working, for example browsing without checkout, and can lower the RTO a flow really needs. A provider's managed recovery service can replace parts of a hand-built pilot light, but the runbook, the drills and the dependency order stay with the team. To check whether recovery is already tested, the maturity scorecard asks for a restore-test log. When the tier is chosen as part of a move to the cloud, plan it within cloud migration and modernization, and the automation behind rebuilds belongs to DevOps and platform engineering.
Limits of this guide
- The impact bands are a convention for discussion; contracts and regulation can set stricter targets.
- The tiers describe region-level recovery on a public cloud; on-premises and hybrid estates add other constraints.
- A tier on paper is not a capability: only a timed drill shows the real recovery time and data loss.
Plan the next step with a Netbase consultant
Frequently asked questions
Often it is. If the product runs across availability zones, the business accepts hours of downtime for a regional outage, and a timed rebuild from infrastructure code fits inside that window, backup and restore is the correct tier.
Pilot light keeps data replicated and the application switched off, so it cannot serve users until servers start. Warm standby keeps a smaller working copy running, so it can take traffic at once and only needs to scale up.
No. Active-active replicates corruption and deletion to every region as fast as good data, so point-in-time backups and tested restores are still required.
The business owner of each flow sets the targets, because they carry the cost of downtime and data loss. Engineering then proposes the cheapest tier that meets them and proves it in a drill.
Next step
If you want a second reader for your recovery targets, book a solution review and bring the list of flows and their owners.
Related services and solutions
AI-ready cloud platform engineering for SaaS and commerce workloads Netbase provides cloud platform engineering for SaaS and commerce teams that need a secure, repeatable home for their software and its AI features, on AWS, Google Cloud, DigitalOcean or Cloudflare. We design the architecture, write the environments as code, connect managed and AI services and hand over a platform your team can run, built the way we run ours.
Learn More
Managed cloud services with AI-assisted monitoring that keep your platform healthy Netbase provides managed cloud services for companies with a live platform to keep it secure, current and affordable after launch: AI-assisted monitoring and alert triage, patching, backups and regular cost reviews on AWS, Google Cloud, DigitalOcean or Cloudflare. Our team supports you Monday to Saturday, with Sunday off. SLAs are agreed on request; we publish no standard SLA figures.
Learn More
AI-assisted cloud migration with a cut-over plan and a way back Netbase provides cloud migration services for companies running business applications on ageing servers or hosting to move them to AWS, Google Cloud, DigitalOcean or Cloudflare without losing data or control. AI speeds up the code and dependency assessment; every migration still has an inventory, a rehearsed cut-over, cost controls from day one and a tested rollback point.
Learn More
AI-assisted DevOps and platform engineering for routine releases Netbase provides DevOps and platform engineering for SaaS and commerce teams to make releases routine instead of risky. We set up CI/CD pipelines, repeatable environments and observability on AWS, Google Cloud, DigitalOcean or Cloudflare, with AI assistance for tests, reviews and incident triage, so every change ships the same way and problems surface before customers notice.
Learn More
Work with Netbase remotely
Every Netbase engagement is delivered remotely, so clients in any country and any industry can work with us; Netbase has worked with clients in 80 countries. Contracts and invoices are issued by Netbase JSC, and payment is made by bank transfer to Netbase JSC's company account or PayPal Business. Talk to the CEO, Huy Nguyen (David), on WhatsApp or by email.
Discuss a project
Netbase JSC helps organizations design, build, modernize, and operate digital products and AI-enabled business systems.project@netbasejsc.com
+84 937 869 689
91 Nguyen Chi Thanh, Dong Da, Hanoi, Vietnam
Get in touch
Tell us what you want to build, modernize, or operate.