marchinthesun

marchinthesun / cluster-performance-engine

Public

Cloud-native Kubernetes performance optimizer for high-core bare-metal clusters. A NUMA-aware scheduler for HPC, ML inference, and CI/CD that cuts latency on 128+ core EPYC/Threadripper nodes.

amd-epyc bare-metal dag hardware-topology high-performance-computing
104
0
69% credibility
Found May 02, 2026 at 99 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Go
AI Summary

NexusFlow is a tool that optimizes job performance on large multi-processor Linux servers by intelligently assigning tasks to nearby processors and memory.

How It Works

1
🔍 Hear about NexusFlow

You learn about a helpful tool that makes big server computers run jobs much faster by smartly matching work to the right parts of the machine.

2
📥 Bring it home

You grab the tool with a simple download and one-click setup, and it's ready to explore your server.

3
🗺️ Map your machine

You peek inside to see the neighborhoods of processors and memory, so you know where to send each job for best speed.

4
⚙️ Run smarter jobs

You launch your tasks, telling it how many processors and which neighborhood, and it keeps everything close together for smooth work.

5
📊 Control from dashboard

You open a simple web view to watch charts, run pipelines, and tweak settings easily without typing commands.

🚀 Enjoy the speedup

Your jobs finish quicker with less waiting and power waste, turning slow servers into speedy workhorses.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 99 to 104 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is cluster-performance-engine?

NexusFlow optimizes workloads on high-core bare-metal servers like 128+ core EPYC or Threadripper nodes, acting as a NUMA-aware autopilot layer between jobs and hardware. It discovers topology via sysfs or hwloc, pins CPUs with sched_setaffinity, binds memory via numactl, runs YAML DAG pipelines with Prometheus metrics, and coordinates shared memory or perf counters over Unix sockets. Built in Go with a Python SDK, it boosts throughput for cloud native kubernetes ai day workloads without replacing kube-scheduler or Slurm.

Why is it gaining traction?

It delivers 25-50% gains in LLM inference, kernel builds, and CFD sims by eliminating cross-NUMA hops and cold cache misses—real metrics from node-local pinning and DAG staging. CLI commands like `nexusflow topology hints` export shell vars for Slurm/MPI, while the gRPC daemon manages cgroups and hugepages, and a secure dashboard (TLS, Bearer auth) lets you exec remotely. Stands out for cloud native go github users needing low-overhead perf on cloud native pg github clusters.

Who should use this?

HPC admins tuning multi-socket Linux nodes for simulations, ML engineers running inference on bare-metal EPYC fleets, CI/CD operators with cloud native buildpacks github pipelines hitting tail latency walls. Cloud native certified kubernetes developers or administrators handling cloud native postgresql github on high-core hardware.

Verdict

Worth testing on a beefy node if NUMA stalls kill your cloud native foundation kubernetes perf—benchmarks back the claims. At 19 stars and 0.7% credibility score, it's nascent but docs, CLI, and dashboard are production-ready; pair with cloud native java github stacks for quick wins.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.