Addition of flash memory improves read-write speed in planning and what-if analyses; doubling of RAM improves in-memory processing capacity.
Oracle introduced a second-generation Exalytics In-Memory Machine on Monday that promises not only extended in-memory analysis, thanks to a doubling of RAM to 2 terabytes, but also improved read-write performance for planning and analysis applications with the introduction of 2.4 terabytes of flash memory.
Oracle started shipping the first-generation Exalytics X2-4 in early 2012, so it's only natural that Moore's Law would bring improvements in capacity and performance with the new Exalytics X3-4. RAM doubled just by using the latest Intel chips, and that will obviously increase the capacity for accelerating applications with in-memory processing performance.
Using the latest disk drives increased Exalytics' internal storage capacity to 5.4 terabytes, up from 3.6 terabytes previously, improving the machine's ability to cache data from Oracle Exadata, Essbase databases or Endeca Information Discovery servers.
The inclusion of flash memory in Exalytics is entirely new, and it's an advance that will benefit specific types of applications.
"For read-write applications, like Hyperion Planning, the flash memory really increases the write performance, so we're typically seeing three to 10 times faster performance," said Paul Rodwick, Oracle's VP of product management, in an interview with InformationWeek.
In corporate budgeting and planning applications, for instance, the addition of flash will help companies avoid having to split analyses by region or division for performance reasons. A big, global planning app that might have taken eight hours to run previously might run in two hours with the Exalytics X3-4, Rodwick said.
Exalytics is designed to automatically cache the hot data that is most in demand to accelerate performance when and where needed. That's a contrast to SAP's Hana in-memory platform, which moves entire applications into main memory. SAP's approach is simpler and cleaner from an architectural perspective, but Oracle makes the case that adding Exalytics to an Exadata environment is a simple step that makes everything run faster without having to purchase vast amounts of RAM server capacity.
"Unlike some vendor solutions, you're not constrained by the physical memory of the machine," Rodwick explained. "You can even access data on giant Hadoop clusters, but Exalytics identifies the most important information to put into main memory."
Exalytics performs its automated caching and acceleration role with Oracle BI Foundation Suite middleware, the Oracle Hyperion Enterprise Performance Management System (which runs on Oracle Essbase) and Oracle Endeca Information Discovery. The latest releases of all three of these software products have been certified to run on Exalytics.
The Exalytics X3-4 is $175,000 plus $21,000 per year for support (up from $135,000 and $16,200 for the X2-4). That's just for the hardware. To this is added TimesTen in-memory database license fees of $138,000 ($34,500 for each of the four Exalytics processors,) plus $30,360 per year ($7,590 per processor) for premier support.
With Monday's announcement, Oracle also introduced two upgrade options for existing X2-4 In-Memory Machines: adding 2.4 TB of flash memory or adding flash plus new boards to increase RAM to 2 terabytes. The flash upgrade is $35,000 while the flash-and-board refresh is $105,000.
Rumors recently circulated that Oracle had dropped the price of Oracle BI Foundation Suite, but Rodwick said that's "old news" from roughly a year ago tied to the introduction of Exalytics.
"The cost for BI Foundation Suite on a named-user basis has never been changed, but about nine months ago we adjusted per-CPU pricing in part because we were seeing more customers want to license the full complement of Exalytics," Rodwick explained.
Oracle BI Foundation Suite was previously priced at $450,000 per processor license, but the cost was dropped to $300,000 per CPU. Exalytics adds four CPUs over and above the number of Exadata Database Machine CPUs required, so the software-related costs of the hardware are considerable. The named-user license remains at $3,675, giving smaller firms a more favorable pricing model.
We welcome your comments on this topic on our social media channels, or [contact us directly] with questions about the site.