Apache Superset and Apache Iceberg

Apache Superset and Apache Iceberg

Apache Superset is the most widely adopted open-source business intelligence (BI) and data visualization platform. With over 50,000 GitHub stars and deployments at hundreds of organizations, Superset is the standard self-hosted alternative to commercial BI tools like Tableau and Looker.

Superset queries Apache Iceberg tables via SQL connections to Iceberg-compatible query engines — most commonly Dremio, Trino, or Spark Thrift Server. There is no native Iceberg connector required in Superset: Superset sends SQL to the connected query engine, which handles Iceberg query planning and execution.

Connection Architecture

Superset (dashboards, charts, SQL Lab)
  → SQLAlchemy connection string (HTTP or Arrow Flight)
    → Query engine (Dremio / Trino / Spark)
      → Iceberg tables (via catalog)

Connecting Superset to Dremio (Recommended)

Dremio provides the richest Superset integration via its AI Semantic Layer — virtual datasets, column descriptions, and metric definitions are accessible in Superset’s dataset explorer.

Dremio ODBC/JDBC Connection

Database URI: dremio+flight://my-dremio.example.com:32010

In Superset’s database configuration:

1. Go to Settings → Database Connections → + Database.
2. Select Dremio from the supported databases list.
3. Enter the Arrow Flight SQL endpoint.
4. Configure authentication (username/password or OAuth2).
5. Test connection.

Superset Dataset from Dremio Virtual Dataset

Once connected, Superset can create datasets from:

• Dremio Virtual Datasets (semantic layer views over Iceberg tables).
• Direct SQL queries against Iceberg tables.

-- Superset custom SQL dataset: revenue trend over Iceberg data
SELECT
    DATE_TRUNC('month', order_date) AS month,
    region,
    SUM(total) AS revenue,
    COUNT(*) AS orders,
    COUNT(DISTINCT customer_id) AS customers
FROM analytics.orders
WHERE order_date >= '2025-01-01'
  AND status IN ('SHIPPED', 'DELIVERED')
GROUP BY 1, 2

Connecting Superset to Trino

Database URI: trino://trino-user@trino.example.com:443/iceberg_catalog/analytics

Superset configuration for Trino:

# Superset: Trino connection via superset_config.py
DATABASES = {
    "trino_iceberg": {
        "SQLALCHEMY_URI": "trino://user:password@trino.example.com:443/polaris_catalog/analytics",
        "OPTIONS": {
            "connect_args": {"http_scheme": "https"}
        }
    }
}

Building Dashboards on Iceberg Data

Step 1: Create a Dataset

In Superset, datasets are the link between the query engine and charts. Create a dataset backed by an Iceberg table or view:

1. Datasets → + Dataset → Select database (Dremio/Trino) → Select schema → Select table.
2. Or: SQL Lab → Write SQL → Save as Dataset.

Step 2: Define Metrics and Dimensions

Superset allows you to define reusable metrics within the dataset:

• SUM(total) → “Total Revenue”
• COUNT(DISTINCT customer_id) → “Unique Customers”
• COUNT(*) → “Order Count”

And dimensions (group-by columns):

• region, order_date, status, product_category

Step 3: Build Charts

Select metric + dimension → Chart type → Render. Superset generates the SQL, sends it to the query engine, and plots the result.

Step 4: Combine into Dashboard

Assemble charts into a dashboard with filters, cross-chart interaction, and scheduled refreshes.

Time Travel in Superset + Iceberg

Superset doesn’t natively support Iceberg time travel syntax, but you can use it via SQL Lab or parameterized SQL datasets:

-- Superset SQL dataset with time travel (Trino syntax)
SELECT * FROM iceberg.analytics.orders
FOR VERSION AS OF 8027658604211071520
WHERE order_date >= CURRENT_DATE - INTERVAL '30' DAY;

Or via Dremio:

-- Time travel via Dremio's AT SNAPSHOT syntax (VDS)
SELECT * FROM analytics.orders AT SNAPSHOT '8027658604211071520'
WHERE order_date >= CURRENT_DATE - 30;

Performance Optimization for Superset + Iceberg

Dremio Reflections for Dashboard Performance

For dashboards that run the same aggregations repeatedly:

1. In Dremio, create a Reflection on the virtual dataset powering the Superset dashboard.
2. Dremio pre-computes the aggregation as a materialized Iceberg table.
3. Superset queries hit the reflection instead of raw Iceberg data.
4. Dashboard load time: seconds → milliseconds.

Superset Cache Layer

Configure Superset’s Redis cache to cache query results for dashboards that don’t need real-time freshness:

# superset_config.py
CACHE_CONFIG = {
    "CACHE_TYPE": "RedisCache",
    "CACHE_DEFAULT_TIMEOUT": 3600,  # 1 hour cache for stable metrics
    "CACHE_KEY_PREFIX": "superset_",
    "CACHE_REDIS_URL": "redis://redis:6379/0",
}
EXPLORE_FORM_DATA_CACHE_CONFIG = CACHE_CONFIG

This reduces load on the Iceberg query engine for high-traffic dashboards.