"How long would we be down?" It's a simple question to ask a CTO, and an uncomfortable one when the honest reply is "it depends who's on leave." Most of the Australian businesses we work with assume their cloud provider has outages covered. They find the gaps when a customer-facing system is down and nobody can say when it'll be back.
Outages come in many shapes. A region loses connectivity, a managed service degrades, a bad deployment corrupts data, or someone deletes the wrong resource group. Cloud resilience is the ability to keep running, or recover within an agreed time, when any of that happens. You don't get it by moving to the cloud. You design it, write it down and test it.
Here's how we approach it: recovery targets the business signs off, a disaster recovery strategy that matches them, sensible use of Australia's cloud regions, and proof that it works before you need it.
Note: This article is general information, not legal advice. Check the official guidance and get advice for your situation.
Why boards are asking now
Critical operations for many Australian businesses now sit on a handful of cloud platforms, so one provider issue can hit payments, bookings and customer service at the same time.
For APRA-regulated entities the expectation is formal. Prudential Standard CPS 230 Operational Risk Management is designed to strengthen the management of operational risk and includes requirements for business continuity and service provider management. According to APRA's page, the standard commenced on 1 July 2025, and updated versions of CPS 230 and its guidance (CPG 230) commenced on 1 July 2026. If you're regulated, read the standard itself with your risk team. This post won't tell you what it requires of you.
Outside regulated sectors the pressure is commercial. Enterprise customers and partners now ask in security questionnaires how fast you can recover and how much data you could lose. "Our provider is very reliable" doesn't pass.
Recovery targets come before technology
Two numbers drive every decision. AWS defines them in its disaster recovery whitepaper:
- Recovery Time Objective (RTO): the maximum acceptable delay between the interruption of service and restoration of service.
- Recovery Point Objective (RPO): the maximum acceptable amount of time since the last data recovery point. In practice, how much data you can afford to lose.
Set them per system, not per company, through a short business impact analysis. The whitepaper makes a good point about timing: a payroll outage just before pay day hurts far more than one just after. It also says something teams often skip. If a recovery strategy costs more than the failure it protects against, don't build it, unless something else (such as regulation) requires it.
Disaster recovery cloud strategies compared
AWS groups disaster recovery into four strategies. The ideas apply just as well on Azure.
| Strategy | What runs in the recovery region | Relative recovery speed | Relative cost and complexity |
|---|---|---|---|
| Backup and restore | Backups only; infrastructure is redeployed from code | Slowest | Lowest |
| Pilot light | Data replicated, core infrastructure in place, app servers switched off | Faster | Low to moderate |
| Warm standby | A scaled-down but fully working copy of production | Faster still | Moderate to high |
| Multi-site active/active | Full production serving traffic in more than one region | Near zero for most disasters | Highest |
A few lessons from the same guidance that save real pain:
- Backups alone aren't a plan. You also have to redeploy infrastructure, configuration and code, which is why AWS says to always deploy with infrastructure as code.
- Replication isn't backup. Continuous replication copies corruption and malicious deletion as fast as good data. Keep point-in-time backups too.
- Fail over with data plane operations. AWS recommends this because data planes typically have higher availability design goals than control planes. Changing Route 53 weights in a hurry is a control plane call; flipping a pre-built Application Recovery Controller routing control is a data plane one.
- Automate the steps, keep a human on the trigger. Automatic failover on health checks can fire on a false alarm and cost you data. Many teams script every step and keep the decision to start manual.
Multi-region architecture in Australia
You can build a multi-region architecture without your data leaving the country.
- AWS runs Asia Pacific (Sydney),
ap-southeast-2, and Asia Pacific (Melbourne),ap-southeast-4, each listed with three Availability Zones. Melbourne is an opt-in Region, so someone has to enable it in each account first. Check this early: a DR runbook that assumes Melbourne is ready in an account where it was never enabled will fail on the day. - Microsoft Azure pairs Australia East with Australia Southeast, and Australia Central with Australia Central 2, which is a restricted-access region.
Two cautions. First, you may not need a second region. AWS notes that Availability Zones sit a meaningful distance apart, so a multi-AZ design in one region may already cover much of your risk. Second, a region pair isn't a recovery plan. Microsoft says deploying to a paired region doesn't automatically make resources more resilient, and advises against relying on Microsoft-managed failover as your primary disaster recovery approach.
A cloud resilience checklist
- Classify systems by business impact and agree RTO and RPO for each with the business owner.
- Remove single points of failure inside a region first: multiple Availability Zones, managed databases with standby replicas, load-balanced application tiers.
- Back up data, configuration and code, with point-in-time recovery and copies in a separate account or subscription, so one compromised credential can't delete both.
- Define infrastructure in code so the recovery environment can be rebuilt the same way every time.
- Pick a DR strategy per system from the table above, matched to its targets and budget.
- Check quotas and capacity in the recovery region so you can actually scale when it matters.
- Write and rehearse the runbook: who decides to fail over, how traffic moves, and how you fail back.
- Test on a schedule and record actual recovery times against the targets.
Item eight is the one that gets skipped. AWS's guidance on testing disaster recovery says it plainly: the only error recovery that works is the path you test frequently. Untested failovers break in boring ways, such as a missing IAM permission, an AMI that only exists in Sydney, or a service quota in the recovery region that was never raised.
How JRC Software builds resilient cloud systems
We start with a resilience review. In the first week we sit with your business owners to rank systems and agree recovery targets, then check what the current architecture can actually deliver: where backups go, whether anyone has restored one, what's single-AZ, and what would need to be rebuilt by hand.
You get a list of gaps, each with an estimated cost to close, so you can decide deliberately instead of over-engineering everything. Often the right call is not to build a second region for a system that a well-run multi-AZ setup and tested backups would cover.
Then our engineers implement the agreed changes: multi-AZ designs, backup policies with cross-account copies, infrastructure as code for the recovery environment and, where the targets justify it, pilot light or warm standby across Sydney and Melbourne or Azure's Australian regions. We finish with a recovery test, measure the real RTO and RPO, and hand over runbooks your team can follow at 3am. For systems still on legacy platforms, resilience is often part of a wider modernisation decision. This sits within our cloud engineering and modernisation service.
Not sure how long your critical systems would take to recover? Book a resilience review and we'll measure where you stand and show you the most cost-effective way to close the gaps.
Frequently asked questions
What is the difference between RTO and RPO?
RTO is the longest acceptable gap between a service going down and being restored. RPO is the longest acceptable time since the last recovery point, which sets how much data you can afford to lose. Both are business decisions, agreed per system, and together they decide which disaster recovery strategy you need.
Do we need a multi-region architecture?
Not always. A system spread across several Availability Zones in one region already survives the loss of a single data centre. A second region becomes worthwhile when your risk assessment includes losing a whole region, your recovery targets are tight, or regulation requires it. With two AWS regions and several Azure regions in Australia, data can stay onshore either way.
Is the cloud provider responsible for our disaster recovery?
No. Providers run resilient infrastructure, but designing and recovering your workloads is your job. Microsoft, for example, says deploying to a paired region doesn't automatically give you high availability, disaster recovery or failover, and that you should develop your own plans. You choose the strategy, configure backups and replication, and test recovery yourself.
How often should we test disaster recovery?
Test on a fixed schedule for critical systems, and again whenever the system or its infrastructure changes significantly. AWS notes that recovery paths that are rarely exercised tend to fail, so the more critical the system, the more often you rehearse failover and restores. Record actual recovery times and compare them with your RTO and RPO.
Cloud Architecture
Book a resilience review
Tell us where you're stuck and we'll come back with practical next steps.
Get in touch