Oracle has introduced the Exadata Database Service on Exascale (ExaDB-XS) with the goal of enhancing performance for database workloads while lowering costs. This new architecture is based on Oracle’s multi-tenant Exascale architecture which intelligently allocates compute resources for its databases using X8M, X9M, and X10M appliances. Exadata, Oracle’s infrastructure for relational databases, has been around for a decade.
The focus of Exadata Exascale is on AI workloads and vector databases, with a new pay-to-use cost structure. The X8M appliances use RoCE networking to connect servers to storage, with Oracle’s storage servers equipped with XRMEM memory accessed via RDMA. The shared architecture now allows a common pool of Exascale compute and storage to cater to thousands of tenants and millions of databases.
Through a specific Exascale control plane, VMs and storage are managed, replacing the previous logic of managing assets through files within the OS. With a focus on redundancy and performance optimization, smaller 8MB fragments called “extents” are utilized to store data across multiple storage servers in a redundant manner. Performance is optimized through data tiering and parallelization of SQL processing.
ExaDB-XS offers similar performance to dedicated infrastructure but at a significantly lower cost, up to 95% less compared to the Exadata Cloud Infrastructure X9M Dedicated. The pricing is based on a pay-per-use model and offers significant savings for users. In the future, Exadata Exascale will be the foundational architecture for all database services in Oracle Cloud Infrastructure.
Overall, Oracle’s Exadata Exascale represents a step forward in innovation, offering improved performance and cost savings for database workloads.