Content by johnson shi (3)

Johnson Shi explains how AKS GPU memory profiling helps you move from “GPU memory pressure” alerts to the exact functions responsible for allocations, using eBPF-based tracing and Grafana flame graphs so you can right-size GPU memory and validate fixes with repeatable tests.

Johnson Shi provides an operational guide to running a geo-replicated Azure Container Registry (ACR) for high availability, explaining how global endpoints, regional endpoints, and dedicated data endpoints behave during incidents, throttling, and DNS changes, with concrete Azure CLI steps for setup, routing control, and troubleshooting.

Johnson Shi explains how Azure Container Registry (ACR) keeps multi-tenant performance predictable using a regional “stamp” architecture, including how registries are transparently moved between stamps when a stamp gets hot and when ACR provides additional isolation for exceptional workloads.