Ceph is an open-source distributed storage platform for object, block, and file storage. In an Apache Iceberg lakehouse, its RADOS Gateway exposes an S3-compatible object interface for private data centers and hybrid environments, preserving control without depending on public-cloud storage. section 5.4.7
Use Ceph when architectural control and distributed durability outweigh operational simplicity:
- Build object storage through RGW while RADOS distributes, replicates, and maintains consistency across cluster nodes. section 5.4.7
- Choose replication or erasure coding, self-healing, and rebalancing to support durability and recovery from hardware failures. section 5.4.7
- Plan topology, replication factors, hardware, network bandwidth, and IOPS carefully; performance depends substantially on deployment and tuning. section 5.4.7
- Expect more operational complexity than simpler self-managed object stores, particularly when interactive workloads demand consistently low latency. section 5.4.7 section 5.5.1
Private Iceberg lakehouse
A mature infrastructure team can deploy Ceph in a private data center, use RGW as the Iceberg object-storage backend, and retain unified object, block, and file capabilities. The team must provision cluster capacity and tune topology, replication, networking, and IOPS before relying on high-throughput concurrent access. section 5.4.7
Restricted internal R&D data
Ceph can support multitenant Iceberg deployments with role-based access controls and centralized identity through Keystone or LDAP, although enforcement may require more manual configuration than cloud-native platforms. section 5.5.2
