The AWS us-east-1 Wobble This Morning: 3 Resilience Patterns Every India SaaS Should Have

At 9:14 am IST this morning, us-east-1 started returning elevated error rates on a handful of control-plane calls. Not a full outage — a wobble. But three of our clients pinged the same Slack channel within four minutes asking why their dashboards felt slow. If one AWS region having a bad twenty minutes can rattle your product, you do not have a resilience strategy — you have a single point of failure with good uptime so far. This post gives you the three patterns we actually ship, with real ₹ costs and an RPO/RTO matrix you can copy. ## What should an Indian SaaS do about AWS region failures? Pick one of three resilience tiers and build to it on purpose. Tier 1 is multi-AZ in a single region (most of you are here by accident). Tier 2 is warm standby in a second region with Route 53 health-check failover. Tier 3 is active-active across regions. Most Indian B2B SaaS at ₹1-20 crore ARR should be at Tier 2 for the database and Tier 1-plus for everything else. Cost: ₹2-6 lakh to set up, plus 30-80% on your infra bill. ## Why this matters now (July 2025) This morning's us-east-1 hiccup was minor. The pattern it exposes is not. AWS publishes every incident on its [Health Dashboard](https://health.aws.amazon.com/health/status), and if you read a year of them, one fact jumps out: us-east-1 is special. It is the oldest region, the default in most SDKs, and the home of global control planes for IAM, CloudFront, Route 53, and DynamoDB global tables. A regional event there has knock-on effects in regions you thought were independent. For Indian SaaS, the stakes changed in 2024-25. Your buyers now ask for an uptime SLA in the MSA. Enterprise procurement teams send a security questionnaire with a "describe your disaster-recovery plan" field. "We run on AWS" is no longer an answer — they want your RPO and RTO numbers in writing. If you cannot fill that field, you lose the deal to a competitor who can. There is also a quieter cost. When us-east-1 wobbles, your on-call engineer drops what they are doing and stares at dashboards for an hour. Across a year, those false-alarm hours add up to real money and real burnout. A proper failover design lets the system ride out a 20-minute regional blip without a human touching anything. ## The 3 patterns, ranked by leverage ## Pattern 1: Route 53 health-check failover (the DIY walkthrough) Route 53 failover routing sends users to your primary region while it is healthy, and automatically to a standby region when a health check fails. This is the highest-leverage pattern because it is cheap, it is DNS, and it works even when the primary region's API is on fire. Here is the build. ## Pattern 2: the cross-region database replica Failover routing without a database in the second region just points users at an app that has no data. The database is the hard part of disaster recovery, and the part most teams skip because it costs money. For PostgreSQL on Aurora, a cross-region read replica gives you a warm copy you can promote to primary. Typical replication lag between ap-south-1 (Mumbai) and ap-southeast-1 (Singapore) sits in the low single-digit seconds — that is your real-world RPO. The cost is honest and worth stating plainly. A cross-region Aurora replica roughly doubles your database compute spend and adds cross-region data-transfer charges. For a mid-size SaaS that might mean ₹35,000-70,000/month extra. That is the price of not losing a day of customer data when a region goes down. We tell clients to size this against one number: what is a full day of your transactional data worth to the customer who loses it? ## Pattern 3: the RPO/RTO matrix (costs nothing, decides everything) RPO (Recovery Point Objective) is how much data you can afford to lose, measured in time. RTO (Recovery Time Objective) is how long you can be down. The mistake is treating these as one number for the whole system. Different data classes deserve different targets, and once you write them down, every infrastructure decision becomes obvious. Here is the matrix we hand clients.

Data class	RPO target	RTO target	What that buys
Payments / financial ledger	0 (no loss)	< 5 min	Synchronous replication or multi-region writes. Most expensive tier.
Core transactional data (orders, users)	< 1 min	< 15 min	Cross-region read replica, promote on failover.
Activity logs / analytics events	< 30 min	< 2 hr	Snapshot copies + S3 cross-region replication.
Cache / derived data	N/A (rebuildable)	< 1 hr	Rebuild from source. Do not pay to replicate it.

## When NOT to do this Do not jump to active-active multi-region if you are a pre-revenue or sub-₹1-crore-ARR product. The engineering and run cost will starve the work that actually grows you, and a 20-minute regional blip a few times a year will not kill a young product. Get to clean multi-AZ, write the RPO/RTO matrix, and add Route 53 failover for the front door. That is enough for most early-stage SaaS. Do not build cross-region database replication if your data is genuinely rebuildable from a source of truth elsewhere. We have watched teams pay to replicate a derived analytics store that they could rebuild from raw events in 40 minutes. Replicate what you cannot recreate; rebuild the rest. And do not trust any failover you have not broken on purpose in the last quarter — the graveyard of DR plans is full of configs that looked right and failed live. ## Real example: a Bengaluru fintech SaaS A Bengaluru-based fintech SaaS — 60 enterprise customers, lending-ops dashboards — came to us after a single-region scare cost them a deal. Their largest prospect's security team asked for an RTO under 30 minutes in writing, and they had no answer. They were single-region in ap-south-1, multi-AZ, no DR story. We did three things over five weeks. We wrote the RPO/RTO matrix with their CTO in one afternoon. We added a cross-region Aurora replica in ap-southeast-1 with a documented promotion runbook. We put Route 53 health-check failover on the API and dashboard. Setup came to ₹4.1 lakh; their infra bill rose about 44%. The payoff was concrete: they filled in the procurement questionnaire with real numbers and closed the deal — and rode out the next us-east-1 blip without paging anyone. We took the same approach building the resilience layer for Radiant Finance's lending platform, where downtime maps directly to lost loan disbursals. As Hrishikesh, our CTO, puts it: the goal is not zero downtime, it is downtime you chose and budgeted for. A region having a bad morning should be a non-event your system absorbs, not a fire drill for your engineers. For the broader picture, see our take on the infrastructure and automation patterns that keep Indian SaaS stable under load. ## Frequently asked questions ### How fast does Route 53 failover actually switch traffic? With a 3-check threshold at 10-second intervals and a 60-second record TTL, expect end-to-end failover in roughly 90-180 seconds. The variable is DNS-resolver caching on the client side, which is why a short TTL matters. Test it live to get your real number. ### Is multi-AZ the same as multi-region? No. Multi-AZ protects you from one data centre failing inside a region — it is the baseline and you should already have it. Multi-region protects you from an entire region having a bad day, which is what us-east-1 events are. They solve different problems; you need both for a serious DR posture. ### What RPO and RTO should an early-stage Indian SaaS target? For core transactional data, aim for an RPO under one minute and an RTO under 15 minutes once you cross roughly ₹1 crore ARR or sign your first enterprise customer. Below that, a few-minute RPO via snapshots and a one-hour RTO is a reasonable, affordable starting point. ### Why does us-east-1 affect regions I do not even use? Several AWS global control planes — IAM, CloudFront, ACM in some flows, and DynamoDB global tables — are anchored in us-east-1. A regional event there can degrade those global services, which then ripples into your stack even if all your compute runs in ap-south-1. Audit for these hidden dependencies first. ### How much does proper multi-region DR cost to run? The setup is ₹2-6 lakh for a Tier 2 warm-standby design. The ongoing cost is the bigger commitment: a cross-region database replica plus data transfer typically adds 30-80% to your infra bill, depending on data volume. The RPO/RTO matrix tells you exactly how much of that you actually need. ### Can you build this on our existing AWS account? Yes. We work inside your account, document every change as infrastructure-as-code, and hand you the runbook so your team can run the failover without us. The first call is technical, with the engineer who would lead the work — not a sales pitch.