Summit 2026 Interactive Lab

Real-Time Arcade Score Streaming with Snowpipe Streaming + Interactive Tables

Stream thousands of arcade game scores from across the globe into Snowflake in real time, materialise them into an Interactive Table, and query with an Interactive Warehouse — experiencing sub-second latency at scale.

Architecture

 Python Generator
 ─────────────────
  Arcade Score Events          Snowpipe Streaming SDK
  (unlimited rows/sec) ──────► StreamingIngestClient
  20 games                      └─ Channel 0      ──► ARCADE_SCORES
  45 cities                                            (Interactive Table)
  500 players                                          CLUSTER BY (GAME_ENDED_AT)
                                                            │
                                                            ▼
                                                     SUMMIT_INT_WH
                                                  (Interactive Warehouse, XS)
                                                  Always-on · sub-second queries

Snowpipe Streaming uses the channel API, not SQL DML, so it writes rows directly into the Interactive Table — no intermediate landing table needed.

Warehouse design

Warehouse	Type	Purpose
`SUMMIT_TRAD_WH`	Standard XS	Traditional warehouse for comparison benchmarks
`SUMMIT_INT_WH`	Interactive XS	All lab queries; always-on for instant low-latency responses

Prerequisites

Requirement	Details
Python	3.9 – 3.13
Snowflake account	In a supported region for Interactive Tables/Warehouses
Role	ACCOUNTADMIN (or CREATE WAREHOUSE + CREATE DATABASE privileges)
OpenSSL	For RSA key-pair generation
JMeter	For concurrency testing (optional)

Supported regions (Interactive Tables GA)

AWS: us-east-1, us-west-2, us-east-2, ca-central-1, ap-northeast-1, ap-southeast-2, eu-central-1, eu-west-1, eu-west-2 GCP: us-central1, us-east4, europe-west2/3/4, australia-southeast2 Azure: All Azure regions

Lab Setup

1 — Run the Snowflake setup script

Open sql/01_setup.sql in Snowsight and run it using a standard warehouse session.

2 — Register the RSA public key for the service user

bash sql/02_service_auth.sh

Generates rsa_key.p8 / rsa_key.pub if they don't exist, then prints the ALTER USER statement. Paste it into Snowsight and run it as ACCOUNTADMIN.

rsa_key.p8 is in .gitignore and must never be committed.

The script provisions (in order):

ARCADE_STREAMING_ROLE + ARCADE_STREAMING_USER + RSA keypair auth policy
ARCADE_DB database + PUBLIC schema + SUMMIT_TRAD_WH standard warehouse
ARCADE_SCORES Interactive Table (CLUSTER BY (GAME_ENDED_AT), initially empty)
SUMMIT_INT_WH Interactive Warehouse (XS, always-on)
ARCADE_REPORTING_POOL compute pool (XS, for the optional Streamlit dashboard)
Grants for ARCADE_STREAMING_ROLE and ARCADE_LAB_READER

3 — Create `profile.json`

cp profile.json.example profile.json

Edit with your account identifier and the full path to rsa_key.p8 (the script from Step 2 prints the exact path):

{
    "user":             "ARCADE_STREAMING_USER",
    "account":          "YOUR_ORG-YOUR_ACCOUNT",
    "url":              "https://YOUR_ORG-YOUR_ACCOUNT.snowflakecomputing.com:443",
    "private_key_file": "/full/path/to/rsa_key.p8",
    "role":             "ARCADE_STREAMING_ROLE"
}

4 — Install Python dependencies

python -m venv .venv
source .venv/bin/activate       # Windows: .venv\Scripts\activate
pip install -r requirements.txt

5 — Start the arcade streamer

cd python
python arcade_streamer.py

============================================================
 Summit 2026 – Arcade Scores Snowpipe Streamer
============================================================
  Account   : YOUR_ORG-YOUR_ACCOUNT
  Database  : ARCADE_DB.PUBLIC
  Pipe      : ARCADE_SCORES-STREAMING
  Channels  : 1
  Target    : unlimited rows/sec
============================================================

  [14:22:05]  rows:      512  |  512.0 rows/sec  |  errors: 0  |  elapsed:    1s
  [14:22:05]  [latency] ARCADE_CHANNEL_0_A3F2B1C4: 540 ms avg
  [14:22:10]  rows:    1,024  |  512.0 rows/sec  |  errors: 0  |  elapsed:    6s
  [14:22:10]  [latency] ARCADE_CHANNEL_0_A3F2B1C4: 512 ms avg

6 — Wait for cache warm-up

The SUMMIT_INT_WH Interactive Warehouse starts warming its local SSD cache as soon as it resumes. Wait 2–3 minutes after the streamer starts before running interactive queries — the first few queries after resume will be slower while the cache populates.

Lab Exercises

Open sql/03_lab_queries.sql in Snowsight while the streamer is running.

Tip: Use SUMMIT_INT_WH for exercises marked ⚡ and SUMMIT_TRAD_WH for exercises marked 🔧. Each USE WAREHOUSE statement is already in the query file.

⚡ Exercise 1 — Pipeline throughput

Watch row counts grow in real time and measure live ingest throughput (rows/sec). The SDK-reported avg processing latency per channel is also printed to the streamer console.

⚡ Exercise 2 — Data freshness

Each row carries GAME_ENDED_AT (when the score was generated and sent, in UTC). The query measures freshness: how many seconds ago was the most recently committed row? The streamer's console also logs Snowflake-reported avg processing latency per channel via the SDK's get_channel_statuses() API.

⚡ Exercise 3 — Global leaderboard (Interactive Warehouse)

Top 20 scores of the last 24 hours. Notice how the CLUSTER BY (GAME_ENDED_AT) clustering key allows the Interactive Warehouse to skip irrelevant partitions and return results in under a second.

⚡ Exercise 4 — Per-game top 5 (window function)

Uses QUALIFY ROW_NUMBER() scoped to the last hour. Stays well within the interactive warehouse 5-second query limit.

⚡ Exercise 5 — Country heat map

Which countries are playing most right now? Japan and South Korea should dominate — the data generator weights cities by real gaming culture.

⚡ Exercise 6 — Game popularity

Pac-Man and Tetris lead; Joust and Tron are rare finds.

⚡ Exercise 7 — Platform breakdown

Arcade cabinets are most common for classic titles; Tetris goes mobile.

⚡ Exercise 8 — Live score feed

Re-run this every few seconds. Each execution hits data that arrived within the last 5 minutes — demonstrating end-to-end freshness from generator to Interactive Table.

🔧 Exercise 9 — Achievement rarity

Shows how rare badges ("Pacifist", "Triple Threat") require specific skill tier

game mode + score conditions — rarely seen in thousands of rows.

⚡ Exercise 10 — Interactive vs Traditional warehouse speed comparison

Run the identical GROUP BY on both warehouses and compare in Query History. The Interactive Warehouse wins at high concurrency thanks to local SSD caching and pre-computed index metadata.

⚡ Exercise 11 — Concurrency demo (JMeter load test)

Use the JMeter load testing tool to simulate 50 concurrent users hitting the warehouse. See Concurrency Testing with JMeter below.

🥚 Bonus A — Find the ghost player

Someone has been logging perfect scores. Can you find them?

⚡ Bonus B — Time Travel on an Interactive Table

SELECT COUNT(*) FROM ARCADE_SCORES AT(OFFSET => -300);

Interactive Tables support Time Travel even with streaming ingestion.

Streamlit Dashboard (Optional)

After completing the lab exercises, deploy a live dashboard against the same ARCADE_SCORES data.

See STREAMLIT.md for full setup instructions — install the Snowflake CLI and Cortex CLI, generate a PAT via sql/04_generate_pat.sql, and deploy the dashboard.

Concurrency Testing with JMeter (Optional)

The jmeter/ directory contains a load test plan that simulates concurrent users querying the warehouse. This demonstrates the Interactive Warehouse's ability to handle high concurrency without queuing.

Install JMeter

macOS:

brew install jmeter

Linux (Debian/Ubuntu):

sudo apt-get update
sudo apt-get install jmeter

Windows or manual install: Download from https://jmeter.apache.org/download_jmeter.cgi

Download Snowflake JDBC driver

Download the driver and install it to JMeter's lib directory:

cd jmeter

# Download the driver
curl -L -o snowflake-jdbc.jar https://repo1.maven.org/maven2/net/snowflake/snowflake-jdbc/3.16.1/snowflake-jdbc-3.16.1.jar

# Copy to JMeter lib directory (macOS with Homebrew)
cp snowflake-jdbc.jar $(brew --prefix)/Cellar/jmeter/*/libexec/lib/

# Or for manual JMeter installation
# cp snowflake-jdbc.jar $JMETER_HOME/lib/

The test runner script will attempt to install the driver automatically, but manual installation ensures it works correctly.

Run the concurrency test

Set up environment variables:

export SNOWFLAKE_ACCOUNT=xy12345
export SNOWFLAKE_PRIVATE_KEY_FILE=/path/to/rsa_key.p8

# Optional - defaults to ARCADE_STREAMING_USER from setup
# export SNOWFLAKE_USER=ARCADE_STREAMING_USER

Note: The test uses the ARCADE_STREAMING_USER created in the setup script with RSA key authentication.

Example:

export SNOWFLAKE_ACCOUNT=YOUR_ORG-YOUR_ACCOUNT
export SNOWFLAKE_PRIVATE_KEY_FILE=/full/path/to/rsa_key.p8

Test Interactive Warehouse (SUMMIT_INT_WH):

cd jmeter
./run_concurrency_test.sh SUMMIT_INT_WH

Test Traditional Warehouse (SUMMIT_TRAD_WH):

cd jmeter
./run_concurrency_test.sh SUMMIT_TRAD_WH

The script will:

Run 50 concurrent threads for 30 seconds
Execute 5 different queries randomly
Generate an HTML report in results_WAREHOUSE_TIMESTAMP/
Display summary statistics

View the HTML report:

open results_SUMMIT_INT_WH_*/index.html

Expected results

Metric	SUMMIT_INT_WH (Interactive)	SUMMIT_TRAD_WH (Traditional)
Throughput	higher	lower
Avg Latency	sub-second	several seconds
Min Latency	sub-second	1+ seconds
Concurrency	Handles 50 concurrent users smoothly	Queue delays with high concurrency

The Interactive Warehouse achieves significantly higher throughput and lower latency because it uses a shared SSD cache and pre-computed indexes optimized for the associated Interactive Tables.

Customize the test

Edit jmeter/concurrency_test.jmx to adjust:

Thread count: ThreadGroup.num_threads (default 50)
Test duration: ThreadGroup.duration (default 30 seconds)
Queries: Add or modify JDBCSampler elements

Troubleshooting

Connection errors:

Verify SNOWFLAKE_PRIVATE_KEY_FILE is an absolute path (output by bash sql/02_service_auth.sh)
Ensure ARCADE_STREAMING_USER has been created (run sql/01_setup.sql)
Confirm RSA public key is registered on the user (run bash sql/02_service_auth.sh and paste SQL into Snowsight)

JDBC driver not found:

Download the driver: cd jmeter && curl -L -o snowflake-jdbc.jar https://repo1.maven.org/maven2/net/snowflake/snowflake-jdbc/3.16.1/snowflake-jdbc-3.16.1.jar

No data returned:

Ensure the arcade streamer is running (python/arcade_streamer.py)
Verify data exists: SELECT COUNT(*) FROM ARCADE_DB.PUBLIC.ARCADE_SCORES

Interactive Warehouse Key Facts

Property	Value
Query timeout (SELECT)	5 seconds (cannot be increased)
Auto-suspend	No — runs continuously
Cache warm-up	Required after resume (2–5 min for small tables)
Compatible table types	Interactive Tables only
Sizing guidance	XS = working set < 500 GB
Billing	Minimum 1 hour; per-second after that

Streamer Options

# Default: 1 channel, unlimited rows/sec, runs until Ctrl-C
python arcade_streamer.py

# Throttled demo (50 rows/sec)
python arcade_streamer.py --rate 50

# Multi-channel high-throughput (4 channels, unlimited)
python arcade_streamer.py --channels 4

# Insert exactly 10,000 rows then stop
python arcade_streamer.py --rows 10000

# Preview generated data without connecting to Snowflake
python arcade_streamer.py --dry-run --rows 5

Data Model

`ARCADE_SCORES` (Interactive Table)

Column	Type	Notes
`SCORE_ID`	VARCHAR(36)	Session UUID
`PLAYER_ID`	VARCHAR(36)	Stable player UUID (500-player pool)
`PLAYER_NAME`	VARCHAR(64)	Display name
`PLAYER_COUNTRY` / `PLAYER_CITY`	VARCHAR	Geographic origin
`LATITUDE` / `LONGITUDE`	FLOAT	GPS coordinates
`GAME_NAME`	VARCHAR(64)	One of 20 classic arcade titles
`GAME_MODE`	VARCHAR(32)	classic (50%) · tournament · co-op · survival · speed-run
`PLATFORM`	VARCHAR(32)	arcade · console · mobile · pc (game-specific weights)
`SCORE`	NUMBER(12,0)	Power-law distributed by skill tier
`LEVEL_REACHED`	NUMBER(4,0)	Correlated with score
`DURATION_SECONDS`	NUMBER(6,0)	Correlated with level
`LIVES_REMAINING`	NUMBER(2,0)	Inversely correlated with score
`ACCURACY_PCT`	FLOAT	Nullable; correlated with skill tier
`ACHIEVEMENT`	VARCHAR(64)	Rare contextual badge (nullable)
`GAME_ENDED_AT`	TIMESTAMP_NTZ	When the score was generated and sent — UTC, client clock (clustering key)

Cleanup

Open sql/05_cleanup.sql in Snowsight and run it.

Drops (in order): the Streamlit dashboard, compute pool, both warehouses, database (cascades to all tables and pipes), service user, and roles.

File Structure

Summit26-InteractiveLab/
├── README.md
├── STREAMLIT.md                     Optional Streamlit dashboard setup
├── requirements.txt                 Snowpipe Streaming SDK dep
├── profile.json.example
├── .gitignore
├── sql/
│   ├── 01_setup.sql                 Full Snowflake provisioning (incl. compute pool)
│   ├── 02_service_auth.sh           Generates RSA key pair + outputs ALTER USER SQL
│   ├── 03_lab_queries.sql           11 exercises + bonus queries
│   ├── 04_generate_pat.sql          Generate PAT for Cortex CLI + snow connection add command
│   └── 05_cleanup.sql               Teardown (dashboard, pool, warehouses, DB, users)
├── python/
│   ├── config.py                    Game catalogue, cities, skill tiers
│   ├── generator.py                 Realistic score generator
│   └── arcade_streamer.py           Snowpipe Streaming SDK ingest
└── jmeter/
    ├── concurrency_test.jmx         JMeter test plan
    └── run_concurrency_test.sh      Test runner script

Name		Name	Last commit message	Last commit date
Latest commit History 27 Commits
docs		docs
jmeter		jmeter
python		python
sql		sql
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
STREAMLIT.md		STREAMLIT.md
profile.json.example		profile.json.example
requirements.txt		requirements.txt

Folders and files

Latest commit

History

Repository files navigation

Summit 2026 Interactive Lab

Real-Time Arcade Score Streaming with Snowpipe Streaming + Interactive Tables

Architecture

Warehouse design

Prerequisites

Supported regions (Interactive Tables GA)

Lab Setup

1 — Run the Snowflake setup script

2 — Register the RSA public key for the service user

3 — Create profile.json

4 — Install Python dependencies

5 — Start the arcade streamer

6 — Wait for cache warm-up

Lab Exercises

⚡ Exercise 1 — Pipeline throughput

⚡ Exercise 2 — Data freshness

⚡ Exercise 3 — Global leaderboard (Interactive Warehouse)

⚡ Exercise 4 — Per-game top 5 (window function)

⚡ Exercise 5 — Country heat map

⚡ Exercise 6 — Game popularity

⚡ Exercise 7 — Platform breakdown

⚡ Exercise 8 — Live score feed

🔧 Exercise 9 — Achievement rarity

⚡ Exercise 10 — Interactive vs Traditional warehouse speed comparison

⚡ Exercise 11 — Concurrency demo (JMeter load test)

🥚 Bonus A — Find the ghost player

⚡ Bonus B — Time Travel on an Interactive Table

Streamlit Dashboard (Optional)

Concurrency Testing with JMeter (Optional)

Install JMeter

Download Snowflake JDBC driver

Run the concurrency test

Expected results

Customize the test

Troubleshooting

Interactive Warehouse Key Facts

Streamer Options

Data Model

ARCADE_SCORES (Interactive Table)

Cleanup

File Structure

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

3 — Create `profile.json`

`ARCADE_SCORES` (Interactive Table)

Packages