Title

AWS re:Invent 2023 - Improve application resilience with AWS Fault Injection Service (ARC317)

Summary

Adrian Hornsby, a principal engineer with the AWS Reliability team, and Iris, a senior product manager, presented on improving application resilience using AWS Fault Injection Service (formerly known as the simulator).
They discussed the high cost of downtime for enterprises and the importance of resilience in maintaining a good reputation and avoiding financial losses.
The concept of resilience was broken down into four pillars: anticipation, monitoring, responding, and learning.
AWS's Resilience Lifecycle Framework was introduced, emphasizing the continuous process of improving system resilience.
The importance of fault isolation boundaries, such as regions and availability zones (AZs), was highlighted, along with the concept of static stability in system design.
AWS Fault Injection Service (FIS) was presented as a tool for resilience testing, allowing controlled experiments to inject faults into systems to uncover hidden issues and improve operational practices.
Iris introduced new features in FIS, including Scenarios for predefined experiment templates and multi-account experiments.
Two new scenarios were announced: AZ Availability Power Interruption and Cross-Region Connectivity, designed to test multi-AZ and multi-region applications.
The session concluded with a call to practice resilience and provided resources for further learning.

The cost of downtime is significant, with enterprises potentially losing hundreds of thousands to millions of dollars per hour of downtime.
Resilience is not just about technology; it involves culture, mechanisms, and tools.
The Resilience Lifecycle Framework is a holistic approach to improving system resilience, which can be entered at any stage.
Fault isolation boundaries are crucial for minimizing the impact of failures and ensuring that systems can handle traffic surges without control plane operations.
AWS FIS is a powerful tool for resilience testing, allowing users to simulate faults and improve their systems' reliability and performance.
The introduction of Scenarios in FIS simplifies the process for customers to start testing their applications' resilience.
The new multi-AZ and multi-region scenarios in FIS enable customers to test complex applications and ensure they can handle real-world failure modes.
Building resilience is an ongoing process that requires regular practice and testing to ensure systems can withstand and recover from failures.