The Databricks Field Guide
A free manual for the Databricks platform, written by the engineers who build on it for a living. What it is, what it replaces, what it costs and how to run it in production. Plain language first, and the detail underneath when you want it.
How it is written
There is no prerequisite and no right order. Every chapter starts plain and gets deeper as it goes, so the same page works in your first week on the platform and in your fifth year. Stop when you have what you came for.
42 chapters
Nearly every chapter opens with one: what the thing is and why it matters, in plain language, assuming nothing. Read only these, front to back, and you are about twenty minutes from an accurate picture of the whole platform.
Then the detail
Everything after that opening is for the person who has to build it. Basics first, then the parts most teams never find. Much of the resistance we meet comes from comparing Databricks against a half-remembered feature list, so we go and name them.
44 comparisons
What you would use on AWS, Azure or Google Cloud to do the same job, named rather than implied. We argue for Databricks in these. We also say where the other clouds genuinely win, instead of pretending they do not.
40 objections
The arguments that come up in every architecture review, in the words people really use. Each one is answered by admitting the part of it that is true. If a colleague sent you here to change your mind, start with these.
Two markers run through the guide. Warning flags something that has cost us or a client real time, money or trust, and there are thirty of them. Our position marks a place where reasonable teams disagree and we have picked a side. Those thirty are the paragraphs most worth arguing with.
What is inside
Storing, processing, analysing and serving, then operating it once it exists. Read straight through and you build a platform from the account downward. Almost nobody reads a manual that way, which is what the routes below are for.
01
7 chapters
02
6 chapters
03
4 chapters
04
9 chapters
05
5 chapters
06
6 chapters
07
6 chapters
Where to start
Start with the vocabulary, then spend an hour building something on the free tier. Those two define every term the rest of the manual leans on and get you to a working governed table without spending anything.
Three chapters, written to be read in order, and honest about when the answer is that you should stay where you are.
What Databricks Actually Is Why Not Just Postgres? What It Actually Costs
Start with the two decisions that are cheap now and expensive to change later.
Including how to show each customer only their own data, and the framing underneath both.
Reporting, Semantics and Power BI Multi-Tenant Reporting The Five Pillars
Margin notes
8 October 2026
Databricks acquired Row Zero on 24 September 2026 to put a governed spreadsheet inside Genie, backed by Genie Ontology and Unity Catalog.
8 October 2026
Lakeflow Pipelines is the declarative layer and Lakeflow Jobs is the orchestration layer on Databricks. Both are generally available and they work together.
8 October 2026
Databricks supports Delta Lake and Apache Iceberg as open storage formats. Delta is the default for managed tables and Iceberg clients can read it.
Meet us
Wednesday, October 14, 2026, at the Hilton Anatole. We're sponsoring, our team is there all day, and we host a small dinner for data and engineering leaders that evening.
Who wrote it
TechFabric has been building software since 2017, and a large part of that work now lands on Databricks: migrations off legacy warehouses, governed agent platforms, and more than a few proofs of concept taken from a notebook that worked on a Tuesday to something a business could depend on. The opinions here are what we accumulated doing it. Some are the obvious ones every good team converges on. Some are the result of getting it wrong first.
It is honest about the places a simpler tool or another cloud is the better answer, because a recommendation you cannot trust to say no is not a recommendation.