Reliability is not just about incident response—it starts with how workloads are placed and provisioned before failures even occur. This session breaks down the journey of translating academic doctoral research into hands-on testing inside Google Cloud infrastructure. We explore how heuristic models and predictive optimization can improve Kubernetes scheduling decisions, mitigate hot-spots, and prevent cascading node degradations. Through concrete benchmarks, test architectures, and post-experiment findings on GKE, this talk delivers actionable insights for SREs looking to optimize workload distribution, lower error budgets consumption, and balance theoretical guarantees with operational reality.
Camilla Martins is a Senior Site Reliability Engineer, Google Developer Expert (GDE), Docker Captain and Computer Science Ph.D. candidate at PPGI/UNIRIO. With extensive experience in cloud infrastructure, Kubernetes, and observability, her research focuses on applying optimization algorithms and heuristics to distributed cloud systems and container scheduling.
