Oracle News | June 30, 2016

Oracle’s New Processor for ‘Scale-Out’ Environments Doubles Down on SPARC M7 Innovations

By: John Soat | Senior Writer


Oracle received widespread acclaim in 2015 for its revolutionary SPARC M7 microprocessor, which featured the most innovative design elements since Oracle acquired Sun Microsystems and its SPARC processor architecture in 2010. Now Oracle is rolling out the next iteration, SPARC S7, which reiterates that revolutionary effort with breakthrough efficiency, built-in security, competitive pricing, and a cloud-first orientation.

That’s because SPARC S7 doubles down on the M7’s innovative engineering while at the same time anchoring a family of servers priced at the commodity end of the hardware sector. The microprocessor also will become the basis of a high-end server cluster and an infrastructure-as-a-service offering in Oracle Cloud.

The S7 is engineered specifically for today’s “scale-out” computing environments, which employ very large numbers of low-cost servers networked together for powerful yet efficient distributed processing. It’s an architecture often employed in cloud computing, either public or private clouds. But it also is very effective running “horizontal” applications, such as data analytics, results indicate.

“That’s where S7 is really targeted,” says Marshall Choy, vice president of product management in Oracle’s systems division. The new processor is “derivative” of its predecessor, he says, but is intended to help businesses “use SPARC for a whole new class of workloads.”

Ready to Rumble

SPARC M7 is a computing heavyweight, incorporating up to 32 fourth-generation CPU cores per processor. The M7’s record-setting performance is most often used in “scale-up” environments, to power compute-intensive applications such as enterprise resource planning.

SPARC S7, by comparison, is a middleweight. It runs the same advanced cores as the M7 but fewer of them—eight per processor.

SPARC M7 caught the IT world’s attention by incorporating an innovative engineering method called Software in Silicon. The process mitigates latency and increases performance by hardwiring significant software functions directly into the processor. In the M7, those functions include data analytics acceleration engines, a process intended for application data integrity and malware-attack prevention, and data encryption without performance degradation.

SPARC S7 features integrated data analytics and data security functions, but the new processor also benefits from engineering advances “specifically on the efficiency side,” Choy says. For instance, the S7 is 50% to 100% more efficient in its core operations than commodity x86 processors, according to industry standard benchmarks, he says.

Price Parity

Oracle is introducing several S7-based server configurations: the SPARC S7-2 Server and S7-2L Server; the ruggedized Netra SPARC S7-2 Server; and the SPARC MiniCluster S7-2, which extends Oracle’s SuperCluster machines to the midrange.  

Those products represent “new economic price points” for Oracle’s SPARC server line, Choy says. The goal: “price parity” with commodity x86-based servers, which populate a great many scale-out architectures, while also delivering enterprise functionalities for scale-out and cloud applications.

The SPARC S7 is “not just a pricing exercise, but an engineering exercise,” Choy says. Its built-in efficiency, along with its hardwired data analytics accelerators, make the S7 particularly well-suited for today’s big data workloads, which increasingly incorporate artificial intelligence capabilities that enable the computer to interpret larger and larger data sets at deeper and deeper levels. The new processor allows for “up to 10 times better, more efficient data analytics and machine learning” than x86-based systems, based on test results, Choy says.

That’s especially important when running unwieldy, resource-intensive open source frameworks, such as Hadoop and Spark or NoSQL and Cassandra. SPARC S7 can “take a previously batch-run operation and transform it into near real-time or interactive analytics,” Choy says.

The processor’s application data integrity function, called Silicon Secured Memory, is intended to simplify and improve protection against bad programming practices and the malware that exploits them, Choy says. Ditto for encryption: The performance overhead for “running fully encrypted, end-to-end in your cloud environment” is less than 2%, benchmark results confirm, Choy says.

Finely-Tuned Cluster

SPARC MiniCluster S7-2 is an engineered system, a finely-tuned bundle of hardware, software, and storage packaged as a single unit—similar to, and derivative of, Oracle SuperCluster M7.

Making the most of S7’s efficient operations, SPARC MiniCluster S7-2 represents “a massive reduction of administration tasks” usually associated with such a complex system, Choy says. “We reduced up to 90% of the administration time and effort required to implement things like security, high availability, databases, and general system administration,” he says.

The two engineered systems are aimed at different areas of enterprise computing. While Oracle SuperCluster M7 is intended for very high-end computing chores—ERP is its #1 workload, Choy says—MiniCluster S7-2 “is an ideal platform for a remote office, branch office type of database deployment—smaller database deployments,” he says.

MiniCluster S7-2 was born from customer feedback. “What customers wanted was a smaller addition to the family, an extension of the SuperCluster into the midrange,” Choy says. And it reflects a stair-step engineering strategy, whereby the M7 evolved into the S7. “We designed all of these systems to be complementary,” he says.

SPARC S7 is an obvious exception to the conventional wisdom that the sequel rarely lives up to the original. “S7 builds on the M7 innovation by taking all the key features, such as embedded security and analytics acceleration, and redesigning them into an integrated processor for much lower cost,” says John Fowler, Oracle's executive vice president of systems. “This makes it possible to use S7 in a much broader range of applications than M7, with great economics.”

It also might make the S7 a very hard act to follow, right? Maybe—maybe not.

Senior Writer

John Soat is senior writer with Oracle's Content Central organization.

More about John Soat