How to Scale Cloud Infrastructure Without Downtime?

Using CloudWatch alarms set to trigger at 80% of resource usage to enable auto-scaling is a good starting point. Just keep in mind that CPU usage isn’t always the first bottleneck; things like connection pools and memory usage can cause issues before CPU gets stretched. Also, remember there’s usually a lag of a few minutes after the alarm goes off for the new instances to be ready, so set your threshold accordingly. And, don’t forget: your instances need to be stateless to handle traffic spikes effectively.

It's really tough to give precise advice without knowing more about your specific architecture, but here are a couple of tips that could help:

1. Make sure you're monitoring critical resources like databases, queues, and shared file systems. Have a plan in place for scaling these resources or offloading heavy transactions.
2. Follow the 12 Factor App principles for your services. This approach will help ensure that your application can scale horizontally more effectively.

There’s a lot to consider, but implementing these strategies can definitely set you on the right path.

A clean rollback strategy has saved our skin several times. We use feature flags, which allow us to revert changes without the need for complete redeploys. Also, validating changes earlier in the deployment process has made rollback scenarios much less stressful for the team.

One major point to remember is that while your app may scale, your database might not follow suit. Configuring read replicas and enabling connection pooling helps significantly. Trust me, it’s better than just waiting and failing gradually.

Check out the AWS Cloud Architect Associate exam prep materials. They cover a lot of scaling strategies based on various business scenarios, which could be really useful for your situation. It would also help if you could share more details about the specific challenges leading to downtime when scaling up, as that would give a clearer picture.