If you had forgotten that Oracle was in the hardware business, it’s easy to understand why, as Oracle has not exactly promoted the business very well.
Oracle’s hardware is descended from Sun Microsystems servers, which Oracle acquired in 2010. Oracle has since shifted direction with the hardware, dumping its custom Sparc processors for x86 and tuning the hardware specifically to run Oracle software.
While hardware hasn’t been the focus, the company recently updated its Exadata platform, the Oracle Exadata X11M. The box is purpose-built to optimize the performance of Oracle Database for AI, OLTP, and analytics database performance in cloud, multi-cloud, and customer data centers.
The X11M is oriented around online transaction processing, typically the forte of IBM mainframes. But going up against IBM isn’t the idea as Oracle knows databases and knows its software, and if anyone can optimize its hardware/software integration, it’s them, experts say.
Oracle is claiming that the data-optimized hardware components and data-intelligent software enhancements in the X11M allow for quicker transaction processing, accelerated analytics, and AI workloads by up to 30 times over the prior generation of servers.
Exadata can be configured to match current operational needs and automatically scaled up and down as needs grow and shrink, much like the elasticity of the cloud. New power-saving features and optimized resource usage contribute to reduced operational costs.
The new X11M is based on the latest generation of AMD Epyc processors with up to 96 cores, featuring up to 25% faster per-core performance over previous generations and a 33% increase in memory bandwidth over the prior generation. A long-time Intel shop, Oracle shifted from Intel to AMD with the previous generation of servers.
The X11M boosts AI workload performance by offloading intricate vector processing to storage servers, minimizing data transfers and enhancing efficiency. AI queries are up to 32 times faster when using binary vector formats, and optimized vector distance functions cut CPU usage, improving query performance by as much as 4.7 times.
The Exadata X11M supports traditional OLTP and analytics customers by supporting up to 1.25 times more concurrent transactions and up to 21 percent lower SQL 8K I/O read latency. This means it can achieve up to one million write IOPS on storage servers for transactional workloads.
The new X11M has a variety of deployment options, like traditional on-prem deployments, Oracle Cloud@Customer as-a-Service, public cloud partners and Oracle’s own Oracle Cloud Infrastructure. This gives customers the option of on-premises and on-demand.
Source:: Network World