Skip to content

Blog

Technical articles, experiments, and project notes from the Gthulhu community.

  • When Does CPU Scheduling Help AI Inference?

    What our vLLM and DGX Spark experiments actually show about CPU contention, static pinning, sched_ext, and when host scheduling becomes part of the AI inference critical path.

    Read article

  • Hardening Gthulhu Network Security with Istio Ambient Mode

    Reduce the blast radius of privileged scheduler components with Kubernetes NetworkPolicy, Istio Ambient, Waypoint, Gateway API, and L7 authorization.

    Read article

  • Improving Network Performance with Custom eBPF-based Schedulers

    An early Gthulhu case study on using eBPF-based CPU scheduling to improve networking workloads. Published on the free5GC community blog.

    Read on free5GC

  • Implementing GTP-driven Automatic Scheduling Optimization with eBPF-based Scheduler

    Explore how GTP-aware signals can drive automatic runtime scheduling decisions for 5G packet-processing workloads. Published on the free5GC community blog.

    Read on free5GC

What we write about

Gthulhu sits between Kubernetes resource allocation and Linux task execution, so most articles focus on practical questions around:

  • Linux sched_ext and eBPF scheduling;
  • Kubernetes workload and topology awareness;
  • AI / GPU inference host-side scheduling;
  • networking and packet-processing workloads;
  • runtime observability and scheduling policy;
  • security and production hardening.

New experiments and project notes will be added here rather than making the top-level Blog navigation point directly to a single external article.