Part of the Gremlin Office Hours series: A monthly deep dive with Gremlin experts. To get the most value out of Chaos Engineering and reliability testing, you need a way to observe your service’s behavior. Observability tools offer insight into how your systems are performing, but observability on its own isn’t enough. You need a way to monitor your systems while testing their reliability so you can determine whether your service passed or failed a test.

Unreliable software can have a painful impact on your customers and your business—something we’ve all seen and felt during high-profile outages. And while building on the cloud with AWS unlocks improved scaling and reliability capabilities, the complexity of modern distributed systems can potentially introduce outage-causing reliability risks. How can you be sure your systems are resilient to failure when they’re based on complex architecture, built by hundreds of teams, and are being updated almost constantly?

Part of the Gremlin Office Hours series: A monthly deep dive with Gremlin experts. Serverless applications are ideal for deploying scalable applications without having to manage infrastructure. However, this also makes it difficult to test their reliability. It’s easy to simulate a network outage or latency when you have direct access to the host that your software’s running on. What do you do when you only have control over the code?

For global financial services companies, reliability must be built-in and validated before and after shipping to production. Resilience testing is crucial for verifying the reliability of your applications under real-world conditions. But ad-hoc testing and exploratory experiments aren't sufficient: you need to run automated, standardized tests at global scale.

Part of the Gremlin Office Hours series: A monthly deep dive with Gremlin experts. Cloud platforms make it easier than ever to deploy massively scalable, distributed workloads, but this is a double-edged sword. There are reliability challenges unique to the cloud that didn’t exist before. Failed migrations, recurring incidents, and reliability toil take their toll.