Title

AWS re:Invent 2023 - What’s new with AWS cost optimization (COP204)

Summary

The session covered the visibility journey in cloud optimization, emphasizing the importance of understanding what resources are running and who owns them before optimization can be effective.
Four key mechanisms for cost optimization were discussed: discounts (savings plans, reserved instances), deleting unused resources, right-sizing, and hibernating non-production workloads.
The speaker demonstrated how applying multiple optimization methods together can lead to significant cost savings.
New features in Compute Optimizer were introduced, including GPU right-sizing, SQL license optimization, and customization features such as tuning options for CPU utilization and instance family selection.
The session also highlighted best practices for purchasing discounts incrementally to avoid management difficulties when large savings plans expire.
AWS Cost Optimization Hub was introduced, which consolidates optimization opportunities across accounts and regions into a single dashboard, providing a clear view of potential savings and efficiency benchmarks.

The visibility journey underscores the need for a mature approach to cloud management, where security, governance, and optimization are progressively addressed.
The combination of discounts, deletion, right-sizing, and hibernation can be tailored to different types of workloads (production vs. non-production) to maximize cost savings.
Compute Optimizer's new features reflect AWS's commitment to continuous innovation, providing more granular control over optimization recommendations and supporting the latest EC2 instance types.
Incremental purchasing of discounts is recommended to maintain flexibility and avoid cost spikes when savings plans expire, indicating a strategic approach to long-term cost management.
The Cost Optimization Hub is a significant development, offering a unified view of optimization opportunities and potential savings, which can help organizations make informed decisions and track their optimization progress over time.