Oracle this week revealed it is taking advantage of the latest release of its system software and the latest generation of AMD EPYC to significantly boost the performance of the Oracle Exadata platform both in the cloud and in on-premises IT environments.
The Oracle Exadata X11M platform, in addition to providing up to 25% faster serial transaction processing and concurrent transaction throughput, can reduce by up to 21% SQL 8K I/O read latency, according to Oracle.
At the same time, analytic query processing is now up to 25% faster, along with there being 2.2X faster analytic I/O on storage servers, and an increase of up to 500GB/sec database in-memory scan. Those capabilities are enabled in part by an ability to scan data residing in both flash and Exadata RDMA Memory (XRMEM).
Additionally, Oracle reports persistent vector index (IVF) searches are up to 55% faster, with transparent offloading to intelligent Exadata storage, while in-memory vector index queries are now up to 43% faster. New software optimizations available across all Exadata platforms provide even faster search for artificial intelligence (AI) applications, with 4.7X more data filtering in storage servers and 32X faster queries when searching binary vectors.
Ashish Ray, vice president for mission-critical database technologies for Oracle, said that latter capability is especially critical as more organizations look to integrate large language models (LLMs) with business data that is typically processed as structured data. The vector search capability makes it possible to convert unstructured data that typically resides outside of a relational database into a format that can be consumed by an LLM, noted Ray.
That approach eliminates the need for IT organizations to hire a separate team to set up and manage a vector database, he added.
In general, Oracle is making a case for a database capable of processing a wide range of types of data that can be deployed either in multiple cloud services or in an on-premises IT environment. Regardless of approach, because Oracle has access to both the database and Linux kernels, the performance of the Exadata platform will always be significantly higher than if an IT organization deployed a database themselves, said Ray.
In fact, many organizations find they no longer have to dedicate resources to optimizing databases that are now managed by Oracle on their behalf, he added.
Oracle, at the same time, also reduces the amount of energy that would otherwise be consumed by enabling more workloads to optimally run on a fewer number of smaller systems, noted Ray.
Many IT teams today need to deploy and manage databases in both on-premises and cloud computing environments. Using the Oracle Exadata platform, it also becomes much simpler to share data across those hybrid cloud computing environments, said Ray. That’s becoming more important as organizations seek to ensure the right data is being made available in the right place at the right time, he noted.
It’s not clear to what degree organizations are looking to rationalize the number of databases they need, but as the number of types of databases employed increases, so too does the total cost of IT as more infrastructure and staff are needed to manage and support those platforms.
In an era where the volume of data that needs to be processed and analyzed only continues to grow, the efficiency at which all that data is being managed now matters more than ever.
