Skip to content

README.md

Latest commit

 

History

History
101 lines (65 loc) · 3.5 KB

README.md

File metadata and controls

101 lines (65 loc) · 3.5 KB
Raw

Interactive Tables Dashboard

A sample dashboard that benchmarks Snowflake Interactive Tables + Interactive Warehouse against Standard Tables using TPC-H workloads. The dashboard displays live KPIs and charts, while a Locust-based load test simulates concurrent dashboard users to measure query latency under load.

Repository Structure

dashboard/
├── api/          # Python FastAPI backend (Snowflake connector, REST endpoints)
├── public/       # Frontend single-page HTML dashboard (Chart.js)
├── sql/          # SQL scripts to create the denormalized LINEITEM_DASHBOARD table
├── locust/       # Locust load test simulating concurrent dashboard users
└── spcs/         # Snowpark Container Services deployment (Dockerfiles, specs, scripts)

api/

FastAPI server that connects to Snowflake and exposes REST endpoints consumed by the dashboard UI. It routes queries to either an Interactive Warehouse or a Standard Warehouse depending on the selected mode.

public/

Static HTML/JS frontend with Chart.js visualizations: KPI cards, time-series line charts, doughnut charts, and bar charts. Served directly by the FastAPI backend.

sql/

Contains create_lineitem_dashboard.sql which builds the denormalized LINEITEM_DASHBOARD table (joining LINEITEM, ORDERS, CUSTOMER, NATION, REGION) in both a standard schema and an interactive-table schema.

locust/

Locust workload definition that simulates real dashboard users. Each virtual user fetches configuration, then repeatedly calls all dashboard API endpoints in parallel with randomized segment filters.

spcs/

Everything needed to deploy the dashboard and load test to Snowpark Container Services: Dockerfiles, service specs, and shell scripts for build, deploy, update, status, logs, and teardown.

Running the Dashboard Locally

  1. Copy .env.example to .env and configure:

    CONNECTION_NAME=<your_connection>
SOLUTION_NAME=<your_solution_name>
DEFAULT_SCALE=100

    The connection must exist in ~/.snowflake/connections.toml.

    All Snowflake object names are derived from SOLUTION_NAME:

    Object Name
    Database <SOLUTION_NAME>_BENCH_DB
    Standard warehouse <SOLUTION_NAME>_BENCH_WH_STD_<scale>
    Interactive warehouse <SOLUTION_NAME>_BENCH_WH_INT_<scale>
    Standard schema TPCH_SF<scale>
    Interactive schema TPCH_SF<scale>_IT

  2. Start the server:

    ./run-dashboard.sh

    This launches the FastAPI backend on port 3000. Open http://localhost:3000 in your browser.

Running the Simulation (Load Test)

Headless mode

./run-users.sh <warehouse>

Where <warehouse> is interactive or standard. This runs Locust against http://localhost:3000 with 5 concurrent users for 5 minutes.

Interactive mode (Locust web UI)

cd locust
uv run locust -f locustfile.py --host http://localhost:3000

Open http://localhost:8089 to configure the number of users, ramp-up rate, and duration.

Deploying to Snowpark Container Services

See spcs/README.md for full deployment instructions.

The deploy.sh script supports two actions:

cd spcs

# Create the LINEITEM_DASHBOARD table (substitutes SOLUTION_NAME and SCALE from .env)
./deploy.sh sql

# Build images, push to registry, create compute pools and services
./deploy.sh services

Run sql first to ensure the denormalized table exists, then services to deploy the dashboard and load-test containers.