Exadata Exascale

Exascale changes the way compute and storage resources are provisioned and consumed in Exadata. Moving storage management out of database servers and into storage servers enables Exadata to adopt a loosely-coupled architecture, where traditionally siloed resource management – compute, memory, and storage capacity – are no longer rigidly defined or restricted.

This new software architecture for Exadata uses the hardware-based RDMA and Smart SQL Offload capabilities of Exadata and applies cloud characteristics to deliver both the extreme performance of Exadata and cloud-scale, multi-tenant resource pooling.

Exadata and Exascale's direct I/O architecture removes the need for intermediate storage management tiers common in traditional on-premises and cloud storage offerings.

Oracle AI Database 26ai sends intelligent data requests – RDMA for OLTP and Smart Scan for AI and Analytics – directly to pooled Exascale storage servers, taking advantage of Exadata's extremely low microsecond read latency, millions of read IOPS, and extraordinary scan throughput.

Exascale creates compute and storage resource pools that can be quickly and easily allocated to meet the needs of any size workload.

Exascale file and extent management is built into Exadata System Software and is self-managed and automated. Coupled with Oracle AI Database 26ai, Exascale frees database servers from managing database files and extents, releasing CPU and memory that can be used for database processing.

Databases files on Exascale are spread across all the storage servers within a storage pool, enabling databases from multiple virtual machine clusters to take full advantage of all the resources of an Exadata system, without having to carve out individual disk groups, all while maintaining fine-grained access control.

Exascale introduces unique RDMA-enabled Block Volumes (Exascale Volumes) to Exadata. Exascale Volumes are used by Virtual Machines (VMs) in place of using local storage in the database servers. Moving the VM volumes to shared, high-performance storage creates new opportunities for performance, scale, and functionality never before seen in Exadata.

A VM running on Exadata can now take advantage of the collective performance of the the dozens or hundreds of pooled disk drives in the storage servers, rather than the two or four drives in the individual database server on which it runs. Before Exascale, the limiting factor for the number of VMs that could run on database server was almost always the storage space available on the local disk drives. Eliminating this bottleneck enables users to scale the number of VM clusters by over four times, boosting the maximum number of VMs per database server from 12 to 50 on a quarter rack or larger X10M or newer Exadata.

Exascale reimagines database snapshots and cloning on Exadata, removing all upstream dependencies with new redirect-on-write technology.

Exascale can instantly create database (or PDB) clones for development or testing. Clones can be either a full copy or a thin clone, and can be created from a live database or an existing database snapshot. Clones on Exascale have no upstream dependencies, such as a read-only test master – and use shared data blocks and redirect-on-write technology for extreme space-efficiency. This drastically reduces storage capacity needs for cloning, and provides every database and clone native Exadata performance, regardless of the use case.