Oracle Exadata X11M: More performance at the same price
Oracle's new X11M database platform boasts better performance than the X10M. This is particularly noticeable in AI, analytics and OLTP.
(Image: Michael Vi/Shutterstock.com)
- Harald Weiss
Oracle has introduced the 13th generation of its Exadata database platform. The new X11M system offers considerably more performance than its predecessor, X10M – at the same price. The improvements are made possible by a faster vector search, faster IOPS and shorter latency in transaction processing, as well as a faster data scan and greater query throughput in analyses. The X11M is optimized for the latest generation of AMD EPYC processors and can scan data in both flash memory and Exadata RDMA Memory (XRMEM).
A look at the data sheet shows numerous improvements, including up to 55 percent faster persistent vector index queries (IVF) with transparent swapping to intelligent Exadata memory. In addition, in-memory vector index queries (HNSW) are up to 43 percent faster. A new software optimization enables faster AI searches with 4.7 times more data filtering in storage servers and 32 times faster queries when searching for binary vectors.
In addition, serial transaction processing is up to 25 percent faster and SQL 8K I/O read latency is up to 21 percent lower at just 14µs. Analytical queries should also be processed up to 25 percent faster, and Oracle also promises up to 2.2 times faster analytical I/O on storage servers and an increase in database in-memory scanning of up to 500 GB/s.
Vector search 30 times faster
AI Vector Search accelerates AI workloads by up to 30 times by offloading processing to the storage servers. This avoids sending large amounts of data over the network. At the same time, only the CPUs located directly next to the data are used. In addition, the system calculates and sends the result of the vector distance function back to the database server, avoiding additional CPU utilization on the database servers and thus making queries 4.6 times faster.
Videos by heise
Memory bandwidth in Exadata X11M database servers increases by 33 percent compared to Exadata X10M, significantly accelerating bandwidth-sensitive storage operations such as analytics and AI. Local storage on the database servers continues to utilize two 3.84T NVMe SSD drives, which customers can expand to four 3.84T drives.
Finally, the Exadata X11M database servers continue to utilize the PCIe 5 data interface for the dual-port RDMA-over-Converged Ethernet (RoCE) network used for the private network.
(mho)
