Oracle recently launched its Exadata X11M product to the market. Exadata is an integrated hardware and software platform designed from the ground up to optimize Oracle Database workloads. Exadata X11M zeroes in on a few areas — performance across AI, OLTP and analytics, scale, operational efficiency, and cost. Further, Oracle simultaneously announced X11M innovations for public cloud, multicloud, hybrid cloud and on-premises deployments.
While these are some big items, it is worth digging into what Oracle did and didn’t deliver with this Database Machine.
Oracle, Enterprise Data And The Database Machine
Oracle’s huge footprint in the enterprise means that it possibly handles more enterprise data than any other technology vendor. This is data that runs organizations — the very definition of mission- and business-critical. Because of these dynamics, Oracle introduced Exadata in 2008. Exadata was designed to deliver a tightly integrated platform tuned for performance, scale and reliability for in-database analytics. Over its thirteen generations of development, Exadata has been enhanced to deliver these capabilities across the spectrum of Oracle Database workloads: AI, analytics, transaction processing, document databases, graph databases, spatial, time series and more.
Exadata is a tightly integrated hardware and software platform — truly a database machine. It is designed from the ground up for performance, scale and reliability. For example, Exadata RDMA memory, also called XRMEM, is a shared read accelerator that uses remote direct memory access to provide incredibly fast access to data cached in storage servers. XRMEM is a prime example of how Oracle has created a bespoke architecture to drive differentiated performance.
Oracle has built many such innovations into the Exadata platform since its launch. Further, while Oracle Database will run on virtually any server platform, its performance on Exadata is significantly better, and running workloads faster means that they can be run using fewer hardware and database licenses — further increasing efficiency. Examples of Exadata optimizations that help run crucial workloads faster include Exadata Smart Scan, which offloads SQL queries to intelligent storage, and AI Smart Scan, which does the same for AI vector search.
Exadata X11M Redesigned From The Ground Up
While Exadata has evolved over the years, Exadata X11M may be the most significant update since Oracle acquired Sun Microsystems in 2010 and replaced HP as its server platform. As I mentioned earlier, X11M is about delivering better performance, operational efficiency and cost. It achieves this through a refreshed hardware platform anchored by dual-socket AMD EPYC servers (with up to 96 cores per CPU) and enhanced system software. Oracle uses these servers together with fast DDR5 memory and PCIe Gen5 flash storage to create high-performance database servers and intelligent storage servers that make Exadata’s unique capabilities possible.
By combining Oracle’s XRMEM with these EPYC capabilities, the X11M storage servers realize a performance increase of up to 33% compared to the X10M and can scan data stored in flash storage 2.2x faster than before — enabling more value to be delivered from data far faster. Oracle also claims that AI vector searches on database servers (in memory) can achieve results 43% faster than on the X10M. For vector searches executed on storage servers (in Flash), the performance increase is 55%.
Further, when looking at how Exadata software leverages hardware, Oracle claims even more incredible results. For customers operationalizing generative AI on larger (private) datasets, more complex vector searches can be run considerably faster due to improvements in storage server data filtering (4.7x increase) and binary vector searches (up to 32x increase).
This all adds up to Exadata X11M delivering real-world value for organizations operationalizing AI with their enterprise data. It is difficult to quantify exactly how much the X11M delivers in terms of realized (real-world) gains in performance and cost — which will inevitably be a little different from the benchmark improvements cited above — but it is clear that these gains are significant. And with no change in price from the prior generation, the inherent business value is only more compelling.
Performance Matters — And Not Just With AI
While AI is the topic every IT pundit likes to discuss, there are other workloads that drive businesses today. One of those is online transaction processing, and this is another area where Oracle has delivered significant improvements with Exadata X11M. Handling OLTP effectively is all about low latency, high throughput and concurrency — in other words, how many transactions can be committed per second. This requires a database with super-efficient indexing capabilities coupled with optimized queries and supported by an infrastructure that can handle extreme concurrency.
However, when performing OLTP at scale, architectural designs of a database residing on servers go out the window due to that extreme concurrency. Because of this, Oracle has spent a lot of time working on software implementations that can fully exploit the underlying infrastructure.
The result of Oracle’s work in fully leveraging the X11M hardware means more transactions on fewer servers. For example, Exadata X11M can handle 25% more concurrent or serial transactions faster than its predecessor. While Oracle did not disclose configurations in its testing scenarios, I suspect this is happening on fewer servers due to the high core counts in AMD’s EPYC processors.
For customers with an older version of Exadata deployed, consider this: Exadata X11M delivers 4x faster concurrent transaction throughput and 62% faster serial transaction processing than Exadata X7.
The company also claims significant performance gains in analytics, with up to 25% faster query processing than the X10M. This is again achieved not just by hardware or software but by the tight integration between the two. In this case, Exadata utilizes several unique capabilities to offload tasks (Smart Scan) and accelerate database I/O (Smart Flash Cache and Storage Indexes) for better results.
As with OLTP, what is designed on paper to support analytics doesn’t always translate into real-world performance as complexity and scale are introduced. This is where the value of an integrated data stack like X11M can deliver value where database platforms cobbled together using commodity hardware just can’t.
Autonomy Drives Innovation
While performance numbers and claims typically (and rightfully) steal the show at any product launch, I tend to look more at any associated impacts on enterprise IT. Technology — generative AI, for instance — is amazing and can theoretically deliver great value to an organization. However, if an enterprise IT organization struggles to deploy, tune and manage that technology, the impact turns negative on many levels. Worse, failed POCs tied to a lack of IT skills can make business leaders skittish about operationalizing technology.
Oracle seems to truly understand this and, as a company, has invested a lot in abstracting both complex and mundane tasks away from IT staffers through automation. With Exadata running Autonomous Database, database administrators and their teams spend far less time tuning and managing environments and more time delivering value to the business. Whether on-prem or in the cloud (or multiple clouds), Oracle has set a pace for self-running, self-tuning, self-managing and self-healing that the rest of the market is following. As an ex-IT executive who had to deal with the budget line items associated with doing all these tasks manually, I can tell you this is a breath of fresh air.
It is hard to quantify how much of an impediment maintenance and management of complex systems is to innovation in an IT organization. It’s not simply the hours spent on a weekly basis. It’s the blinders that get put on staffers, along with the inability to see anything beyond the next trouble ticket. This is a compounding issue that bogs down far too many enterprise IT organizations. Because of this, I believe that what Oracle has been doing with Exadata, Exadata Cloud and Autonomous Database is worthy of far more recognition than it gets.
Exadata In The Cloud — And Multiple Clouds
Oracle has spent a lot of time and energy driving native multicloud support for Exadata. This means that Oracle Cloud Infrastructure (and the Exadata platform) physically resides not only in OCI data centers but also in the datacenters of every major cloud service provider. More importantly, it means that that customers’ applications and data can natively and securely integrate with their cloud instances.
This is worth repeating: for Oracle’s customers that use other CSPs, OCI and the Exadata platform physically reside in AWS, Microsoft Azure and Google Cloud datacenters. They are consumed as a native service of each CSP and sit on that CSP’s network, integrated with other cloud services.
This is a wise strategy on Oracle’s part — one that I’ve written about a lot over the past two years (for example here, here and here). I’m a big fan of what Oracle is doing with this for many reasons, but let me dig into a couple. First and foremost, Oracle meets customers where they are. The company is smart enough to realize that not all Oracle Database customers will run their workloads in OCI datacenters. Because of this, Oracle has put a premium on Oracle Database residing where the customer’s applications and services are deployed.
Second, the company is forcing the cloud market to radically change from a bunch of stovepiped CSP environments to one in which apps and data can move quickly, securely and natively from CSP “A” to CSP “B” without fear of exploitation — and without crazy egress fees. We’re still early in this game, but I see Oracle making moves that are going to benefit every enterprise IT organization, whether they are Oracle customers or not.
The Enterprise Impact
It is sometimes difficult to quantify the value and impact of technology on an organization precisely. I’ve never been a fan of studies and tests that show such exact quantifications of value delivered. Value is too subjective, and each customer is too unique to take any claims as gospel.
However, Oracle Exadata X11M is delivering huge improvements from an order-of-magnitude perspective. Further, it is doing so without a cost increase compared to previous-generation cloud and on-premises systems. Because of this, it seems to me that any existing Exadata customer would be wise to take a look at X11M and what it can deliver in terms of operationalizing AI and improving core business capabilities, and any Oracle Database customer that’s not currently using Exadata should take a deep look into it as well.
