AWS: Graviton4 Gets a 12% Price Reduction
Amazon quietly reduced Graviton4 instance pricing across all regions. Combined with the existing 20–30% price-performance advantage over x86 equivalents, this makes the migration case compelling for any compute-heavy workload still running on Intel or AMD. If you haven't run a Graviton spike on your primary services, schedule one.
Google Cloud: Cross-Cloud Networking Preview
GCP announced a public preview of Cross-Cloud Network, enabling private connectivity between GCP and other cloud providers without traversing the public internet. Early adopter reports show sub-5ms latency between GCP us-central1 and AWS us-east-1. Worth evaluating for multi-cloud data pipelines.
The RDS Over-Provisioning Problem
A detailed analysis published this month showed the median production RDS instance running at under 15% average CPU. The root cause: fear of database downtime drives over-provisioning at launch, and nobody revisits it. Aurora Serverless v2 and scheduled scaling are worth re-evaluating if your RDS line item is disproportionate to your traffic.
Kubernetes 1.33: Pod Scheduling Readiness Graduates to Stable
K8s 1.33 graduated Pod Scheduling Readiness, allowing external controllers to gate scheduling explicitly. This is useful for ensuring dependent resources — FPGA allocation, license servers, external secrets — are available before a Pod begins consuming node capacity.
Tool of the Month: OpenCost
OpenCost (CNCF sandbox) now supports multi-cloud cost attribution at the namespace level. If you're running a shared cluster across multiple product teams, namespace-level showback has become meaningfully easier without a commercial FinOps platform. The Prometheus integration is solid.