Oracle 26ai can join external tables backed by S3 Parquet files with regular internal tables in a single SQL statement, without data movement, ETL pipelines, or Spark. The optimizer treats S3-backed external tables as another data source and picks the right join strategy automatically. This post shows how to build a customer dimension table from […]
Hybrid Analytics with Oracle 26ai and S3 Data Read More »
I spent months building a tool that answers a question nobody else was asking: how many databases can your infrastructure actually handle before something breaks? HammerDB-Scale orchestrates parallel database benchmarks on Kubernetes, correlates application metrics with storage behaviour, and finds consolidation limits. I also overengineered it for the wrong user experience. But I thought it
HammerDB-Scale 2.0 and Why I Rebuilt the Whole Thing Read More »
TL;DR: DBMS_CLOUD.CREATE_EXTERNAL_TABLE creates an external table that reads Parquet directly from S3. Define the table once, query it with standard SQL. No data loading, no ETL. With S3 connectivity working (Part 1 and Part 2), this post shows how to create DBMS_CLOUD external tables over Parquet files in S3 and query them from Oracle 26ai.
Querying Parquet Files on S3 with DBMS_CLOUD in Oracle 26ai Read More »
TL;DR: DBMS_CLOUD fails with non-AWS S3 because it parses hostnames to select the signing algorithm and extract the AWS region. Fix: point s3.us-east-1.amazonaws.com at your endpoint via DNS, add it as a SAN on the endpoint’s TLS cert, and import the cert into an Oracle wallet. Full DBMS_CLOUD functionality, no file downloads. In Part 1,
Solving DBMS_CLOUD Connectivity with S3-Compatible Storage Read More »
TL;DR: DBMS_CLOUD requires valid HTTPS certificates with matching hostnames. Self-signed certs or HTTP-only? DBMS_CLOUD won’t work. This post shows what breaks and why. Version Note: This uses Oracle 26ai (23.26.1.0.0). SSL handling behavior changed in 23ai. On 19c/21c, sqlnet.ora settings like SSL_SERVER_DN_MATCH=NO still work, so self-signed certs may be viable with additional configuration. The steps
Connecting Oracle 26ai to S3-Compatible Object Storage Read More »
Finding which lakehouse architecture performs best on your infrastructure is expensive. Deploying a single stack on Kubernetes by hand takes days (ask me how I know). A different catalog, a different query engine, a different version, different resource profiles. The cost of the first deployment is high. The cost of repeating it for every permutation
Deploy, Score, and Compare Data Infrastructure on K8s with Lakebench Read More »
Part 2 of The Economics of Data Architectures in 2026 Part 1 surveyed seven architecture patterns, each with its own economics. Warehouse bets on centralization. Lakehouse bets on eliminating the ETL tax. Mesh bets on distributed ownership. And so on. But here’s the question nobody asks at selection time: what will it cost to leave?
Data Architectures and Change Management Read More »
Part 1 of The Economics of Data Architectures in 2026 The Modern Data Stack is Dead The Modern Data Stack (MDS) was never an architecture. It was a philosophy: best-of-breed tools, loosely coupled via APIs, separation of concerns. Fivetran for ingestion. Snowflake for storage. dbt for transformation. Looker for visualization. Buy the pieces, assemble yourself.
Data Architectures and Their Economics Read More »
