Skip to content

SolutionsServerless data platforms

A data lake you don't have to babysit.

A governed lake on Amazon S3, catalogued by AWS Glue and queried with Amazon Athena or Amazon Redshift Serverless. No clusters to size or patch.

Plan your data platform

Definition

What is a serverless data platform?

A data lake and warehouse built from managed services that scale and bill with use. Data sits in Amazon S3 in open formats, the AWS Glue Data Catalog describes it, AWS Lake Formation decides who reads which columns, and Amazon Athena or Amazon Redshift Serverless answer the queries.

Credentials

AWS Services Partner
Advanced Tier
AWS service deliveries
8
AWS certifications
25+
AWS certified
Every engineer

What is inside the lake? Three views.

Every table is catalogued, every grant is code, and every query reads only what it needs.

Serverless lakehouse · artefacts

01 Catalogue

aws glue get-tables --database-name sales_curated --query 'TableList[].Name'[ "orders", "customers", "returns" ]aws glue get-table --database-name sales_curated --name orders \    --query 'Table.[PartitionKeys[].Name, StorageDescriptor.SerdeInfo.Name]'[ [ "dt" ], "parquet" ]# raw zone: as delivered · curated zone: Parquet, partitioned by dayPASScrawler sales-curated: 3 tables, schema unchanged

Tables, partitions and schemas live in the AWS Glue Data Catalog.

02 Govern

lakeformation/analysts.tfTerraform+1−0
resource "aws_lakeformation_permissions" "analysts_customers" {  principal   = var.analyst_role_arn  permissions = ["SELECT"]  table_with_columns {    database_name         = "sales_curated"    name                  = "customers"    wildcard              = trueAdded:     excluded_column_names = ["email", "phone"]PII  }}
Policy checkPII columns readable by analystsFAIL · 2 columns→PASS · 0 columns

AWS Lake Formation grants columns to roles, in Terraform.

03 Query

SELECT region, sum(amount) AS revenue  FROM sales_curated.orders  WHERE dt = '2026-10-16'  GROUP BY region;# reads one partition and two columns of a Parquet tablePASSquery succeeded · workgroup scan limit not reached

A partition filter means Athena reads one day, not the whole table.

Is serverless the right fit?

Fit check

Think twice

  • Heavy BI load, steady around the clockPrice provisioned
  • Many sub-second dashboards at onceTest first
  • Unpartitioned CSV scanned in fullFix layout first

Good fit

  • Spiky or unpredictable query loadBills by use
  • Many teams reading one copy of the dataLake Formation
  • A small team with no time to run clustersManaged

How is it billed?

Each engine bills on its own unit, and rates vary by Region. Parquet and partitioning cut what Athena scans, and so what it costs.

How each service in a serverless data platform is billed
ServiceYou pay forUnitRates
Amazon AthenaData each query scansPer TB scannedAthena pricing
Amazon Redshift ServerlessCompute while queries runRPU-hourRedshift pricing
AWS GlueETL jobs and crawlersDPU-hourGlue pricing
Amazon S3Storage and requestsGB-monthS3 pricing

How do we build the lake?

  1. Step 01

    Land and catalogue.

    Raw and curated zones in Amazon S3, catalogued by Glue.

  2. Step 02

    Govern access.

    Lake Formation grants by role, down to the column.

  3. Step 03

    Serve every team.

    Athena, Redshift Serverless or SageMaker, per team.

FAQ

Common questions.

It depends on how steady your load is. Serverless bills per query or per compute-hour used, which suits spiky and ad hoc work. A steady, heavy load can cost less on provisioned capacity. We model both against your query history first.

Yes. Streams land in Amazon S3 through Amazon Data Firehose or AWS Glue streaming jobs. Designing the stream itself is data engineering.

No. Amazon Redshift Serverless can query the lake next to its own tables, and Athena can reach other sources through federated queries. We move one workload at a time.

Next step

Plan your data platform with an engineer.

Tell us where your data lives. We sketch the lake with you.

Plan your data platform