Data lake software stores raw files in cheap object storage and adds the layers that make them usable: ingestion, table formats, catalogues, access control and query engines.
This guide covers eleven products, from storage and governance pieces to full lakehouse platforms, ranked on setup effort, real cost with add-ons, ease of leaving, independence from a cloud partner and fit.
Vendors can pay for more visibility on this website, for example a higher place in a ranking or more room on a page. Commercial relationships can also affect which products we feature and how prominently. Every vendor link is marked the same, whatever the arrangement behind it. How that works.
In short
What data lake software does
Data lake software holds structured and unstructured data in open files on object storage, and provides the ingestion, cataloguing, permissions and query tooling needed to analyse it.
Our order comes from five tests, taken in turn: effort to set up, what you really pay, what happens to your data on the way out, how locked in you are, and the intended customer. The ranking then gives places 1 to 3 to challengers outside the best-known brands, if the category has such vendors, and the incumbents follow. Why those five, and why there is no score out of ten, is on the how we work page.
Lakehouse ingestion and table upkeep running inside your own cloud account
Ranked #1 of 11 in Best Data Lake Software in 2026.
Pricing on requestNorth America
Onehouse suits an engineering team that already has object storage and wants ingestion from Kafka, database change capture and file drops turned into maintained Hudi, Iceberg or Delta tables.
The platform deploys in your VPC or Kubernetes, including on-premises clusters. Expect to size and budget it with the vendor, since the site shows a pricing model but no rates, and you still run your own query engines or use its managed ones.
What stands out
Runs in your VPC
Hudi, Iceberg, Delta
Ingestion and table upkeep
Where it costs you
Rates are not listed on the website, so a budget needs a sales conversation.
Its centre is ingestion and table upkeep, which leaves the BI layer to other tools.
Right for
Teams that must keep lake data and compute inside their own cloud account
Wrong for
Buyers wanting one fixed monthly price list
United StatesPay as you go or a flat yearly fee; rates not listed on the site; test drive offered
Trino-based SQL across lakes and other sources, sold by compute credit
Ranked #2 of 11 in Best Data Lake Software in 2026.
Free tierPublished pricingNorth America
Starburst fits organisations whose data already sits in several places and who want one SQL layer on top. Galaxy starts on a free tier, and the 30-day trial exposes Enterprise features such as autoscaling and access controls before the account falls back.
Because it is built on Trino, skills carry over from open Trino. Budgeting needs the credit consumption table, since cluster size drives credits per hour.
What stands out
Built on Trino
Credit-based billing
Federated queries
Where it costs you
Credit rates climb with each tier, and ABAC, SCIM and PrivateLink start above the entry paid tier.
It queries data held elsewhere, so storage and table maintenance remain your task.
Right for
Analysts who need SQL over lake files and other databases without moving data
Wrong for
Teams wanting storage and ingestion in one product
United StatesFree tier with three clusters; Pro, Enterprise and Mission-Critical priced per credit, rates published; 30-day trial
Git-style branching and rollback for data sitting in your object store
Ranked #3 of 11 in Best Data Lake Software in 2026.
Open sourcePublished pricingNorth America
lakeFS answers a narrow question: how do you test a change to lake data without damaging the original? Branches are zero-copy, so a trial run does not duplicate files, and access works through an S3-compatible gateway, SDKs and a REST API.
It sits beside engines such as Spark and Trino rather than replacing them. Larger data volumes, more users and managed hosting require the quoted Enterprise plan.
What stands out
Data version control
Source available
Zero-copy branches
Where it costs you
The Team plan stops at five users and 500GB of data under management.
BUSL licensing means production use is limited to the unmodified software.
Right for
Data and ML teams needing branches, audit and rollback on existing buckets
Wrong for
Buyers looking for a query engine or catalogue alone
United StatesTeam plan at a published monthly rate with a 30-day trial; Enterprise quoted
S3-compatible object storage you run yourself, as the base layer of a lake
Ranked #4 of 11 in Best Data Lake Software in 2026.
Free tierSelf-hostablePublished pricingNorth America
MinIO is the storage tier you place under a lake when data must stay on hardware you control. AIStor lists S3, an Iceberg catalog and encryption features, and a quoted Enterprise subscription adds multi-node scale and support.
The cost shows up in operations: your team sizes the cluster, replaces drives and upgrades it, and any catalogue, query engine or governance tool comes from elsewhere.
What stands out
Object storage
Self-hosted
S3 and Iceberg catalog
Where it costs you
AIStor Free runs on one node only and bars redistribution.
Its public GitHub repository carries a notice that nobody maintains it any more.
Right for
Organisations building an on-premises lake on S3-compatible storage
Wrong for
Buyers who want hosted storage with no hardware to run
United StatesFree single-node edition; Enterprise Lite and Enterprise quoted; 60-day trial
Hybrid data platform spanning data centres, public clouds and the edge
Ranked #5 of 11 in Best Data Lake Software in 2026.
Self-hostablePublished pricingNorth America
Cloudera is aimed at organisations that cannot put every workload in one cloud. It presents one runtime for on-premises and cloud, cloud bursting for peak demand and a shared layer for security, compliance and lineage.
Iceberg tables are supported across those environments. The effort is in the first ninety days, because a hybrid estate needs platform staff and the sales process precedes any hands-on evaluation.
What stands out
Hybrid deployment
Iceberg tables
Shared security layer
Where it costs you
On-premises pricing is by quote and the pricing page lists no trial.
Published cloud rates exclude infrastructure and networking charges from your provider.
Right for
Large organisations running lakes across data centres and several public clouds
Wrong for
Small teams wanting self-service signup and a card payment
United StatesCloud rates per compute unit published; on-premises quoted; no trial listed
Lakehouse query engine on Iceberg, now part of SAP
Ranked #6 of 11 in Best Data Lake Software in 2026.
Self-hostablePublished pricingNorth America
Dremio is a query and semantic layer for Iceberg lakes, with an open catalogue based on Apache Polaris. The Community Edition lets a team try it on a laptop or server, Cloud bills per compute unit with a 30-day credit, and Enterprise is self-managed on Kubernetes.
Since SAP bought the company, check support terms and whether your contract sits with Dremio or with SAP.
What stands out
Iceberg and Polaris
Free Community Edition
Part of SAP
Where it costs you
SAP ownership may change packaging, support and contract terms.
Dremio Cloud currently runs on AWS only, with Azure listed as coming soon.
Right for
Teams standardising on Iceberg who want a fast SQL layer
Wrong for
Buyers who need a managed cloud on Azure today
United StatesCloud billed per compute unit, price published; Enterprise quoted; Community Edition free; 30-day credit trial
Compute-metered lakehouse platform with a public price list
Ranked #7 of 11 in Best Data Lake Software in 2026.
Published pricingNorth America
For a data lake buyer, Databricks is the compute-first option. Work is metered in DBUs by the second, storage in DSUs, and a price list gives the undiscounted figure for each SKU by cloud and region. Committed-use discounts are arranged through sales.
That openness helps comparison, yet the total remains uncertain until real jobs run, since storage, networking and provider fees are charged separately. A free Community Edition exists, limited to learning Spark.
What stands out
Per-second billing
DBU pricing
Committed-use discounts
Where it costs you
Unit prices are public per SKU, but the final bill also depends on the cloud provider's own charges.
A trial linked to your own cloud account still runs up provider costs once resources start.
Right for
Organisations that want a lakehouse with a published price list and per-second metering
Wrong for
Buyers needing a European-headquartered supplier, as the company is based in San Francisco
United StatesPay as you go per Databricks Unit, price list published; committed-use discounts; trial offered
Warehouse platform that also reads Iceberg tables from your cloud storage
Ranked #8 of 11 in Best Data Lake Software in 2026.
Pricing on requestNorth America
Snowflake enters this ranking as a warehouse that has opened up to the lake. Iceberg tables can sit in Snowflake storage or in an S3, GCS or Azure bucket through an external volume.
The further you move toward your own storage and catalog, the more you give up: lifecycle management, Fail-safe and full platform support. Billing is by credits and compressed storage across four editions, and trial length is not stated, so check what the chosen edition covers.
What stands out
Iceberg tables
Four editions
Credit consumption
Where it costs you
Using your own bucket for Iceberg tables leaves protection and recovery with you, and Fail-safe does not apply.
Tables on an external catalog get limited platform support and no lifecycle management.
Right for
Teams that keep curated warehouse data in Snowflake and only read open table formats beside it
Wrong for
Teams building a lake-first design in which every table lives on an external catalog
United StatesConsumption-based credits and monthly storage across four editions; per-credit rates sit in a separate table
Central permissions and cataloguing for data lakes held in Amazon S3
Ranked #9 of 11 in Best Data Lake Software in 2026.
Pricing on requestNorth America
Lake Formation is the governance layer for lakes built on S3 and the Glue Data Catalog. Tag-based and attribute-based access control, cross-account sharing and CloudTrail audit logs are documented.
Hybrid access mode lets you adopt it one use case at a time. Because enforcement happens through Athena, Redshift Spectrum, EMR and Glue, an engine outside that list is not covered by these permissions.
What stands out
S3 permissions
Row and cell filters
AWS only
Where it costs you
It enforces permissions only through integrated AWS analytics services.
Extra meters apply for Governed Tables and storage API calls beyond S3 and Glue costs.
Right for
AWS-based teams that need column, row and cell-level control on S3 data
Wrong for
Teams running lakes outside AWS or on several clouds
United StatesNo charge for permissions; underlying services billed separately; storage API and Governed Tables metered
Blob Storage with a hierarchical namespace for analytics workloads on Azure
Ranked #10 of 11 in Best Data Lake Software in 2026.
Pricing on requestNorth America
Azure Data Lake Storage is the file-system layer for lakes on Azure, with Azure RBAC and POSIX ACLs set at directory or file level. Premium, hot, cool, cold and archive tiers are priced separately, and reserved capacity is sold in blocks.
Spark and Presto read it through the ABFS driver. Everything above storage, such as tables, catalogue and SQL, comes from other Azure or third-party services.
What stands out
Hierarchical namespace
POSIX ACLs
Azure only
Where it costs you
Pricing adds metadata and directory-rename charges when the hierarchical namespace is on.
Moving data out of cooler tiers early triggers prorated deletion fees.
Right for
Azure customers wanting Hadoop-compatible lake storage with directory-level permissions
Wrong for
Teams needing a catalogue or query engine in the same product
United StatesPer-GB storage plus operation charges by access tier; reserved capacity in fixed blocks
Open lakehouse with Presto and Spark engines, as SaaS or installed software
Ranked #11 of 11 in Best Data Lake Software in 2026.
Self-hostablePublished pricingNorth America
watsonx.data offers three routes: managed SaaS, bring-your-own-cloud in a VPC, and on-premises software on IBM Power and Fusion HCI. It supports Iceberg, Hive and Delta tables and lists Presto, Gluten with Spark, Milvus and Cassandra.
The first ninety days go on picking a route and engines. Pay per use and subscription pricing both exist, so model both before committing.
What stands out
Presto and Spark
SaaS or on-premises
Iceberg, Hive, Delta
Where it costs you
Rates exclude a per-account hourly core-support charge.
Prices are indicative and vary by country.
Right for
Enterprises wanting Presto and Spark engines with a software option for their own hardware
Wrong for
Small teams wanting a single simple tariff
United StatesPay per resource unit, rates published; monthly subscription option; on-premises entitlement; trial offered
Data lake software holds structured and unstructured data in open files on object storage, and provides the ingestion, cataloguing, permissions and query tooling needed to analyse it. Work through these questions in order; they separate the products better than a feature matrix does.
01
Decide whether you need a published price
The split is 7 published against 4 quoted. If you want to compare numbers first, start with Starburst, lakeFS, MinIO, Cloudera, Dremio, Databricks, IBM watsonx.data.
02
Decide how much the jurisdiction matters
Headquarters range over 1 countries and 1 regions: North America 11. Legal exposure follows the vendor, not only the data centre, so read the contracting entity before you sign.
03
Consider whether you want the source
1 of these are open source: lakeFS. Hosting one yourself trades a subscription for maintenance.
Choosing between a lake with its own pipelines and a lake you assemble
The first split in data lake software is how much you buy as one product. Onehouse bundles ingestion, table optimisation and format sync, while MinIO or Azure Data Lake Storage give you storage only, leaving ingestion and cataloguing to you.
Cloudera and IBM watsonx.data sit in between, selling a platform that spans engines and governance. Count the people needed to connect the parts: an assembled lake looks cheap on the licence line and expensive in engineering time during the first ninety days. Ask each vendor which jobs it does not cover, and write down who owns them.
Which of ingestion, table maintenance, catalogue and query does this product leave to you?
Who is on call when a pipeline between two vendors breaks?
Can you start with storage only and add layers later?
Table formats and what leaving costs
A lake keeps files in your own storage, so exit should be easy, but table formats decide how easy. Onehouse writes Hudi, Iceberg and Delta, Dremio is built around Iceberg, and Snowflake can read Iceberg tables from your bucket. Check whether a catalogue is open, whether another engine can read the same tables, and what the vendor lets you export.
Snowflake's own documentation separates tables it manages from those on an external catalogue, with fewer features for the latter. Test a read from a second engine during the trial, because that single test shows how portable the data really is.
Can a second engine read your tables without conversion?
Is the catalogue an open specification or a vendor product?
Do files stay in a bucket you own, and is that stated in the contract?
Reading the price of Databricks, Starburst and the rest
Lake pricing is metered, and meters differ. Databricks charges per unit with a public price list, Starburst sells credits that rise by tier, and IBM watsonx.data counts resource units. Amazon Lake Formation adds little itself, but Athena, Glue and S3 bill on their own. Cloud providers' infrastructure charges often sit outside the vendor's rate, as Cloudera's pricing page notes.
Ask each vendor to estimate a real month of queries and ingestion rather than quoting a list price, and ask which line items fall outside its invoice. A sensible test is to run one representative pipeline during the trial, read the meters afterwards and multiply by the number of pipelines you expect in a year.
Which charges land on your cloud bill instead of the vendor invoice?
What does an idle cluster cost?
Are discounts tied to a minimum commitment?
Licences, ownership and buying alone
Several products here carry licence or ownership changes that matter to a buyer. lakeFS is under Business Source 1.1, MinIO's community repository is source only and no longer maintained, and Dremio says it is now part of SAP. None of these stops a purchase, but each changes what you can modify, redistribute or expect in support.
Read the licence grant, find who signs your contract and ask how roadmap commitments are written. Independence also means running the product without a cloud partner, which suits Onehouse, Starburst and lakeFS, but not Lake Formation, which exists only inside AWS.
Does the licence allow your intended production use without modification?
Which legal entity signs the contract and who owns it?
Does the product run outside one cloud provider?
What goes wrong most often when buying data lake software
Buying a storage service and assuming it includes a catalogue, access policies and a query engine.
Comparing list prices without counting the separate cloud bill for compute, storage and network traffic.
Skipping a read test from a second engine, then finding tables are tied to one vendor's catalogue.
Treating a licence or acquisition notice as background noise instead of reading what it changes for production use.
07
Frequently asked questions
8 answers
What is the best data lake software in 2026?
Our ranking of 11 puts Onehouse first. Onehouse, founded by the creators of Apache Hudi, deploys into your own VPC or Kubernetes cluster, so data stays in your account and any cloud discounts you negotiated still apply.
Its core is ingestion, table optimisation and syncing across Hudi, Iceberg and Delta. The drawback is cost visibility: the site names pay-as-you-go or a flat yearly fee but lists no rates, and your own cloud bill sits on top.
Which data lake tools publish their pricing?
7 of the 11, with the pricing model each one publishes:
Starburst: Free tier with three clusters; Pro, Enterprise and Mission-Critical priced per credit, rates published; 30-day trial.
lakeFS: Team plan at a published monthly rate with a 30-day trial; Enterprise quoted.
MinIO: Free single-node edition; Enterprise Lite and Enterprise quoted; 60-day trial.
Cloudera: Cloud rates per compute unit published; on-premises quoted; no trial listed.
Dremio: Cloud billed per compute unit, price published; Enterprise quoted; Community Edition free; 30-day credit trial.
Databricks: Pay as you go per Databricks Unit, price list published; committed-use discounts; trial offered.
IBM watsonx.data: Pay per resource unit, rates published; monthly subscription option; on-premises entitlement; trial offered.
The remaining 4 sell by quote.
Is there a free data lake tool?
Starburst, MinIO have a free plan or free self-hosted edition.
Where are these data lake software vendors established?
In 1 countries across 1 regions: North America 11.
Onehouse: United States.
Starburst: United States.
lakeFS: United States.
MinIO: United States.
Cloudera: United States.
Dremio: United States.
Databricks: United States.
Snowflake: United States.
Amazon Lake Formation: United States.
Azure Data Lake Storage: United States.
IBM watsonx.data: United States.
Which data lake tools are open source?
lakeFS.
Which data lake tools can you host yourself?
MinIO, Cloudera, Dremio, IBM watsonx.data. 7 products offer no self-hosted option.
What should you use instead of Onehouse?
Look at Starburst (made for Analysts who need SQL over lake files and other databases without moving data) and lakeFS (made for Data and ML teams needing branches, audit and rollback on existing buckets), which hold places two and three.
Who should not buy Onehouse?
Buyers wanting one fixed monthly price list. Rates are not listed on the website, so a budget needs a sales conversation..
Missing from this data lake software ranking? Contact us with a short description of the product and its users. Adding it is an editorial choice, and software advice explains what a listing does and does not involve.
