Field-validated on-premises storage integration for analytics, engineering, and AI
Enterprises are moving analytics and AI workloads on-premises to avoid public cloud egress costs and data-sovereignty risk, but that only works if the platform is proven on the infrastructure underneath it. Cloudera and Nutanix jointly validated this combination in the field: Cloudera Base and its Data Services, including Data Engineering, AI, and Data Warehouse, deployed on Nutanix, using Nutanix Objects as S3-compatible storage and Nutanix Files for NFS.
Running Cloudera Runtime 7.3.2 on Nutanix AOS 7.5.1 and AHV 11.0.1, the validation exercised Hive, Spark, Cloudera Data Engineering, Cloudera AI Workbench, and Cloudera Data Warehouse against Nutanix Objects and Nutanix Files. Four of five integration scenarios, except the RAZ scenario, passed full validation, giving customers a de-risked reference architecture for Cloudera analytics and AI on-premises.
Cloudera Base and its Data Services, including Cloudera Data Engineering, Cloudera AI, and Cloudera Data Warehouse, were deployed on Nutanix AHV 11.0.1 with Nutanix AOS 7.5.1, running Cloudera Runtime 7.3.2 managed through Cloudera Manager 7.13.2, with Cloudera Data Services (ECS) 1.5.5 SP2 hosting the data services tier. All Cloudera cluster VMs ran RHEL 9.6.
Nutanix Volume Groups were used for HDFS DataNode storage rather than standard vDisks, since Volume Groups persist independently of VM lifecycle and protect HDFS data if a VM is accidentally deleted. A custom, DNS-resolvable FQDN was registered via Prism Central for all S3A connections, with the Nutanix Objects SSL certificate distributed to every cluster node and the Cloudera Manager truststore for HTTPS.
AOS RF1 with HDFS RF3 is supported for storage efficiency, whereas AOS RF2 with HDFS RF2 was used during testing for layered resiliency. Active Directory served as both DNS and identity provider, with AutoTLS, Ranger, Knox, and Atlas forming the security and governance stack.
Nutanix Objects was configured as the S3-compatible backend via Cloudera Manager's Safety Valve, with endpoint, path-style access, and signing algorithm set directly in core-site.xml, requiring no custom connectors or code changes. Hive and Spark both validated full CRUD against the S3A bucket, including external table creation and cross-engine consistency.
Cloudera Data Engineering log offload to Nutanix Objects was confirmed via the Management Console API (HTTP 200), and Cloudera Data Warehouse's managed Parquet table storage in Nutanix Objects was confirmed via DESCRIBE FORMATTED. Cloudera AI Workbench was provisioned against a Nutanix Files NFS share, reaching Ready status.
The one exception: Ranger Authorization Service (RAZ) could not be implemented with Nutanix Objects, since RAZ requires Knox IDBroker to call the AWS STS AssumeRole API, which Nutanix Objects does not currently support. Non-RAZ S3A with static credentials is the validated path today; RAZ support is a Nutanix Objects roadmap item. WORM, replication, and versioning should remain disabled, as each conflicts with Cloudera's normal write and lifecycle operations.
This joint validation gives customers a tested, production-ready blueprint for running Cloudera on premises on Nutanix, fully on-premises with no cloud dependency. Read the full technical blog and white paper for configuration detail and validation evidence.
Learn more at www.nutanix.com.