Oracle unveils Big Memory Machines with new SPARC M6 processors for in-memory applications, to fight the data flood.

Companies are awash in data. In fact, Science Daily reported that 90% of the world’s data was generated over the last two years. Capturing, integrating, and analyzing massive amounts of data from both social media interactions and traditional in-house systems demands a shift to in-memory processing.

With Oracle’s new SPARC M6-32 systems, customers can now run entire applications in memory. Based on the SPARC M6 12-core processor, these systems have a staggering 32 terabytes of system memory and up to 384 processor cores with 8-threads per core. The SPARC M6-32 provides twice the memory capacity of IBM Power servers and delivers extreme in-memory compute performance for large-scale, mission-critical workloads and databases. And it offers the highest levels of system availability for core business applications.

In-Memory Application Modules

To take advantage of the real-time capabilities of the SPARC M6, Oracle is redesigning specific modules in its ERP stack, including Oracle E-Business Suite, PeopleSoft, JD Edwards EnterpriseOne, and Siebel applications by adding new in-memory analytic features. This allows customers to analyze unprecedented volumes of data in real-time to create forecasting models and “what-if” scenarios to better understand and manage their business.

Ganesh Ramamurthy, Oracle vice president of Engineering explains, “We have taken essential features of the new hardware platform and married that with in-memory application capabilities. There are 3072 hardware threads on this M6-32 platform combined with 32 terabytes of memory. That is at the peak in the industry— nobody else is doing this stuff.”

Most Efficient Virtualization Technologies

There are two models for utilizing a machine with this kind of massive capability: First, a single image of the operating system where all 32 terabytes of memory can be allocated to a single application. Second, and more typically, customers can virtualize the machine by carving it up and assigning different parts of the machine to different applications. This allows you to put multiple tiers of an application on the same machine, or put different applications, different users, and different departments on the same machine and isolate them from each other by virtualizing them.

Using the industry’s most comprehensive suite of built-in, no cost virtualization technologies, these systems help maximize IT resources and bring new levels of server utilization and efficiency to market. They can be virtualized using Oracle VM Server for SPARC “LDoms” (which is hypervisor-based virtualization); using Oracle Solaris Zones (OS virtualization); and through electrically isolated Dynamic Domains (hardware partitioning), which is the highest level of isolation possible.

Benefits of Software Defined Networking

With the adoption of cloud computing, customers are employing virtualization widely in their data centers and starting to understand the benefits of Software Defined Networking (SDN). High-speed networking design affects how easily applications connect, how fast they can talk to each other, and how much that costs. Building a low-latency, high-bandwidth network is extremely complex and expensive to acquire and maintain. Even simple 10Gb Ethernet enterprise networks can run in the millions of dollars.

With Oracle’s SPARC M6-32, customers have a 24 terabit-per-second interconnect inside a single system, with 150 ns latency, to network virtualized applications. Oracle also provides the virtual networking tools to control the architecture. This class of networking is not possible with the conventional out-of-the-server approach. This is just one more example of the efficiency and cost savings that come from the technology advances in Oracle’s SPARC M6-32 server.

SPARC/Solaris Roadmap Continuity

Masood Heydari, senior vice president of Engineering for Oracle’s SPARC systems development, says that customers purchase these high-end machines with the intent of using them over a long period of time and with the expectation that they can be upgraded to improve performance.

“So from a capability perspective, in terms of throughput, in terms of memory bandwidth, in terms of I/O capability, it was designed with that upgrade in mind,” says Heydari. “With the SPARC M6-32 system, we have doubled the number of cores and doubled the performance of the machine. Plus, customers with SPARC M5-32 servers can add SPARC M6 processors to their existing systems.”

Oracle also continues a long-standing tradition and commitment to SPARC customers with its Oracle Solaris Binary compatibility application guarantee. That guarantee ensures that a customer’s existing Solaris-based applications will run unmodified on the new SPARC M6 systems.

Oracle SuperCluster M6-32 Engineered System

The SPARC M6-32 server is also the core building block of Oracle’s new SuperCluster M6-32 engineered system. This top-of-the-line Oracle SuperCluster system is the biggest, most scalable, and fastest engineered system from Oracle.

Combining powerful virtualization technologies and unique Oracle Exadata storage and Oracle Exalogic software optimizations, the Oracle SuperCluster M6-32 is designed to run the full range of mission-critical database and enterprise applications on a single system in a consolidated or private cloud environment. SuperCluster M6-32 scales up to 32 processors and 32 terabytes of memory for breakthrough in-memory application performance.

In describing the SuperCluster M6-32, Ramamurthy adds, “What is unique to the M6-32 platform is the combination of big memory hardware and in-memory application software. This engineering allows customers to run a number of Oracle applications in memory on the SuperCluster platform and achieve performance levels never seen before.”

Transforming Social Data into Business Intelligence

Nearly 2.8 zettabytes of data was created in 2012 and that number will double again by 2015, according to the consultancy firm IDC.

Having the ability to sort through social media data from an unstructured data source like Hadoop and meld key bits of information with a traditional OLTP database is one more emerging trend for in-memory systems like the SPARC M6-32.

The financial industry is are already mining its website logs to understand what their customers are interested in and combining that data with their online trading activities to provide their customers with timely recommendations and personalized tips. That process usually involves moving some portion of the unstructured data into a structured database, like Oracle Database, and then joining that data to do some analytics—a process easily handled by the new SPARC M6 systems.

[colored_box color=red]

The Most Powerful SPARC Server Family

In the past, Symmetric Multiprocessing (SMP) systems have had substantial price premiums due to the technology required to achieve high performance across dozens of processors and associated shared memory. Oracle’s SPARC M6-32 is the SMP server that changes all that. By leveraging a common set of technologies across product lines, the price/performance metric of a SPARC M6-32 server with 32 processors is similar to Oracle’s SPARC T5 server with 2, 4 or 8 processors. This kind of price/performance is unprecedented and allows customers to choose the best system for their applications without having to weigh the financial burden of the big system memory option.

This represents a major shift in the paradigm of large-scale computing. Customers interested in deploying large data-intensive applications might consider a cluster of X86-based systems thinking that would be the most cost effective way to build their infrastructure. But a scaled-out architecture like that, with dozens of 2-processor systems in multiple racks connected with high-speed network switches would be considerably more complex, inefficient, and costly than a single SMP server. Just one of Oracle’s SPARC M6-32 servers with 32 processors has 5 times the performance of a cluster with 64 two-processor X86-systems, but still costs 20% less. This type of price/performance will completely change the way CIOs look at their computing infrastructures.

The new Oracle SPARC M6-32 server doubles the throughput performance of the previous generation SPARC M5-32 systems to easily accommodate large-scale business critical workloads. With the new SPARC M6-32 server and Oracle SuperCluster M6-32 engineered systems, customers have the ability to host entire applications or databases in system memory to realize breakthrough performance results for their Oracle software and large-scale mission critical workloads.

[/colored_box]