Skip to content

README.md

Latest commit

 

History

History
122 lines (88 loc) · 5.42 KB

README.md

File metadata and controls

122 lines (88 loc) · 5.42 KB

Demo Starter Kit

Everything you need to demo GCO (Global Capacity Orchestrator on AWS).

🎬 Live demo recording (click to expand)

GCO Live Demo

Automated demo showing costs, capacity-aware routing, 4 schedulers running simultaneously (Volcano, Kueue, YuniKorn, Slurm), high-performance storage (FSx, Valkey, EFS), and live LLM inference — all on one platform. Re-record with bash demo/record_demo.sh.

📦 Deploy recording (click to expand)

GCO Deploy

Fresh gco stacks deploy-all -y from a clean account (re-record)

🗑️ Destroy recording (click to expand)

GCO Destroy

Full teardown with gco stacks destroy-all -y (re-record)

Table of Contents

What's Included

File Description
GCO_PRESENTATION.pptx Editable slide deck template — customize in PowerPoint, Keynote, or LibreOffice Impress
GCO_PRESENTATION.pdf PDF export of the slide deck (regenerate after editing the .pptx)
DEMO_WALKTHROUGH.md Step-by-step demo script covering infrastructure, jobs, health, and API
INFERENCE_WALKTHROUGH.md End-to-end inference demo: deploy, invoke, scale, autoscale, stop/start, model weights, Valkey cache
DEMO_WALKTHROUGH.pdf PDF version of the main walkthrough
INFERENCE_WALKTHROUGH.pdf PDF version of the inference walkthrough
live_demo.sh Automated live demo script — runs through costs, schedulers, storage, inference, and EFS
lib_demo.sh Shared function library sourced by live_demo.sh, record_demo.sh, and BATS tests
LIVE_DEMO.md Documentation for the live demo script: usage, customization, and maintenance
record_demo.sh Records live_demo.sh as an animated GIF using asciinema + agg
record_deploy.sh Records gco stacks deploy-all -y as an animated GIF (for fresh deploy demos)
record_destroy.sh Records gco stacks destroy-all -y as an animated GIF (for teardown demos)
live_demo.cast Asciinema recording (generated by record_demo.sh)
live_demo.gif Animated GIF of the demo (generated by record_demo.sh, embedded above)
deploy.cast Asciinema recording of a fresh deploy (generated by record_deploy.sh)
deploy.gif Animated GIF of the deploy (generated by record_deploy.sh)
destroy.cast Asciinema recording of teardown (generated by record_destroy.sh)
destroy.gif Animated GIF of teardown (generated by record_destroy.sh)
md_to_pdf.py Script to regenerate PDFs after editing the markdown

Quick Start

  1. Edit DEMO_WALKTHROUGH.md and INFERENCE_WALKTHROUGH.md — replace placeholder names
  2. Regenerate the PDFs:
pip install weasyprint
brew install pandoc
python demo/md_to_pdf.py DEMO_WALKTHROUGH
python demo/md_to_pdf.py INFERENCE_WALKTHROUGH

Which Resource to Use

GCO_PRESENTATION is an editable slide deck (.pptx) for presenting GCO to stakeholders, leadership, or at events. Open it in PowerPoint, Keynote, Google Slides, or LibreOffice Impress to customize it with your name, team, and talking points. Export to PDF when done (see below).

DEMO_WALKTHROUGH covers the core platform: deploying stacks, submitting jobs, health monitoring, the authenticated API, and multi-region routing. Good for a general technical audience.

INFERENCE_WALKTHROUGH focuses on the inference serving feature: deploying vLLM endpoints, invoking through API Gateway, scaling, autoscaling with HPA, rolling image updates, stop/start, model weight management, and Valkey caching. Best for ML/AI-focused audiences.

live_demo.sh is an automated terminal script for live presentations. It detects which features are enabled in cdk.json and runs through costs, schedulers, storage, and caching with narrated output. See LIVE_DEMO.md for full docs.

All walkthroughs can be run independently. The inference walkthrough assumes infrastructure is already deployed.

Regenerating PDFs

Walkthroughs (from Markdown):

python demo/md_to_pdf.py DEMO_WALKTHROUGH
python demo/md_to_pdf.py INFERENCE_WALKTHROUGH

Requirements: weasyprint (Python) and pandoc (system package).

Slide deck (from PowerPoint):

  1. Open GCO_PRESENTATION.pptx in PowerPoint, Keynote, Google Slides, or LibreOffice Impress
  2. Edit slides — update your name on slide 2, add/remove content as needed
  3. Export to PDF:
    • PowerPoint: File → Export → PDF
    • Keynote: File → Export To → PDF
    • LibreOffice: File → Export as PDF
    • Google Slides: File → Download → PDF
  4. Save as demo/GCO_PRESENTATION.pdf

Pre-Demo Checklist

  • Infrastructure deployed (gco stacks deploy-all -y)
  • EKS endpoint set to PUBLIC_AND_PRIVATE for local kubectl access
  • Cluster access configured (./scripts/setup-cluster-access.sh)
  • kubectl get nodes works
  • GPU quota available in your demo region
  • Valkey enabled in cdk.json if demoing the cache walkthrough
  • Walkthrough personalized with your name
  • Slide deck personalized (edit GCO_PRESENTATION.pptx, re-export PDF)
  • Change bio slide on slide 2 to your bio