Pentaho Alternative: The Community Edition Licence Changed Under You

The Licence Moved

If you built on Pentaho Community Edition, you built on Apache 2.0. That held through version 9.5.

From 10.2, Pentaho ships under the Business Source License. BSL lets you read and modify the source, and restricts what you may do commercially with it for a defined period. Whether that restriction bites depends on how you use it — which is itself the problem, because it turned a settled question into a legal one. An engineer used to be able to answer "can we use this?" alone. Now someone from legal has to.

Nothing was switched off and nothing stopped working. But the terms you accepted are not the terms you are on, and that changed without any action on your part.

If this feels familiar, it should. JasperReports Server's free edition was withdrawn outright in January 2024. Different vendor, different mechanism, same lesson: an open-source licence held by a single commercial owner is a decision that owner can revisit.

Where Pentaho Sits Now

Hitachi acquired Pentaho in 2015 and has steadily folded it into a wider data-management portfolio. Community Edition releases have grown less frequent. Neither fact is an announcement, and Hitachi has not said Pentaho is finished — but the pattern is legible enough that the wider data engineering community has drawn its own conclusion, and increasingly treats Pentaho as something to migrate off rather than start on. Teams publicly describe replacing Pentaho jobs with Airflow and watching 40-minute runs drop under 5 minutes.

That is the reputational position, fairly stated: still shipping, still supported commercially, no longer what anyone reaches for first.

The Enterprise Edition Is Not a Soft Landing

The obvious answer to a restrictive community licence is to buy the commercial one. Look at what that costs before you assume it.

Hitachi publishes no price list — every deal is a custom quote, which is itself a warning sign about where negotiating leverage sits. Reported figures land between roughly $43,000 a year for a modest 24-core data integration deployment and $98,000 for a 64-core enterprise configuration. First-year totals typically run two to three times the licence once implementation, connectors, and training are counted.

The detail that says the most comes from a customer describing a base licence around $35,000, followed by another $45,000 for the connectors they actually needed. Complex deployments pass $150,000.

So the community edition asks you to consult a lawyer, and the enterprise edition asks you to open a negotiation with no published anchor price. Neither is where you were when you chose Pentaho.

What DataPallas Does Instead

Everything sits under one roof here: getting data in, shaping it, building documents from it, delivering those documents, and putting dashboards and analytics in front of people. Open source, self-hosted on machines you own, reaching any database through JDBC, on Linux, Windows, or macOS.

A licence change is what brought you to this page, so the commercial side should be stated straight. DataPallas is open source with fair commercial pricing. That pricing is built around the people who build and administer the pipelines and reports — not your readers, your cores, or your data volume.

There is no free tier, and that is the point rather than an omission. A community edition lasts exactly as long as the commercial interest propping it up, and you have just watched that interest change its mind. Paying for the product is what gives a vendor a reason to keep shipping it.

Data Integration Without the Designer

Pentaho Data Integration — Kettle — is where most Pentaho estates actually live: transformations and jobs drawn in Spoon, scheduled, and quietly holding the business together.

DataPallas splits that work in two.

Moving data is handled by CDC replication, streaming changes out of your operational databases into ClickHouse within seconds. MySQL and PostgreSQL work directly; SQL Server, Oracle, Db2, and MongoDB arrive through Debezium. Switch on CDC at the source, fill in connection details, start the containers. Nothing to draw, nothing to schedule, and no nightly window — it is a stream rather than a batch.

Reshaping data is handled by dbt Core, which builds staging views, dimensions, facts, and analytical views in plain SQL. Version-controlled, testable, reviewable by anyone who reads SQL. A .ktr is XML behind a canvas, thick with coordinates nobody wrote on purpose; a dbt model is a query you can reason about.

That pattern runs through the whole product. Reports are configured on screen — source, template, parameters, schedule — with AI available beside anything that wants code and dormant until you call on it. All of it is then written out as plain text you can read without opening the tool that produced it. Spoon gave you a canvas and left you with files that will not merge; here you keep the canvas and the files merge.

Reporting That Goes Further Than Pentaho Report Designer

Reports are driven from SQL, Groovy, or files — CSV, TSV, fixed-width, Excel, XML, Google Sheets, Office 365 — and come out as PDF, XLSX, DOCX, HTML, or any text format.

Layouts are written in HTML and CSS: true page breaks, headers and footers that repeat, exact placement, web fonts, and a preview that refreshes when you save. XSL-FO handles documents with demanding typography. Nobody has to learn a bespoke report designer, and AI can write the template outright.

Then the part Pentaho never really had: bursting. Split one report by customer, employee, or branch and push each piece out by email, FTP/SFTP, cloud storage, HTTP/WebDAV, or a portal — on a timetable, retrying on failure, and verified before anything reaches a recipient. Plus a Document Portal where each person finds only their own files.

Mondrian's Job, Done Differently

Pentaho's OLAP came from Mondrian and MDX. Here the modelling lives in Cubes: dimensions, measures, joins, segments, and hierarchies stated once in a text DSL that diffs and merges like code. Reports, dashboards, AI questions, and embedded components all read the same definition, and AI drafts the first version from your live schema. No MDX, no cube processing window.

Underneath, DuckDB covers roughly 100K to 100M rows with no administration at all, and ClickHouse carries it to billions.

Dashboards You Can Put In Your Own Product

Live dashboards, and data tables, charts, and pivot tables that slot into your own application behind one script tag — React, Vue, Angular, or plain HTML.

And Chat2DB turns ordinary questions into SQL, with the AI crew helping across reporting, pipelines, portals, and web work.

The Honest Bit

Your .ktr and .kjb files do not run here, and .prpt report definitions do not import. Those have to be rebuilt.

Two things shrink that job. Most Pentaho transformations, examined honestly, are mainly moving tables from one place to another — CDC replaces that outright rather than reimplementing it. What is left is genuine business logic, and that becomes dbt SQL you can actually test.

Where PDI is deeper: its transformation step library is enormous, and the visual debugger with row-level inspection is genuinely useful when a pipeline misbehaves. Jobs leaning on unusual steps or intricate error routing need rewriting rather than translating. Look at those first — they set the size of the project.

Phase One: Replicate in Parallel, Change Nothing

Point CDC at the same databases your Pentaho jobs already read. Your existing transformations keep running on their normal schedule, loading your current warehouse, while DataPallas streams the same tables into ClickHouse continuously.

Build the dbt models, define a cube over them, and compare the results against what Pentaho produces for the same period. Two stacks, your own data, differences visible.

Cut over one subject area at a time, once the numbers agree. Nothing is switched off until you are satisfied.

Where to Start

1. Check which Pentaho version you are on and which licence came with it. 9.5 and earlier is Apache 2.0; 10.2 onward is BSL.
2. Count your jobs by what they really do. Pure table movement is replaced, not rewritten.
3. Run CDC against one source database alongside your existing pipeline.
4. Build dbt models for one subject area and put a cube on top.
5. Compare the numbers, then weigh what you have built against a Hitachi quote you would have to ask for.