An Oracle blog about Openomics

  • sun
    August 3, 2016

SPARC S7 & Java – Wake Up and Smell the Coffee!

Guest Author

Oracle’s recently announced SPARC S7 server (Next Gen Oracle SPARC) has gained BIG news on its screaming performance and cloud infrastructure capabilities, while offering x86 commodity cost points (Not KIDDING!).  But a bit more under the radar with SPARC S7 for our ISVs, is what happens when you run Java applications on it. 
Let's just say its worth pouring a cup of Java and reading on.

Here are 7 reasons to run your Java applications on SPARC S7:

1. The SPARC Core is architected, tested, and tuned to maximize performance on Java
   The SPARC S7 processor has a ton of Oracle intellectual property in it to make it the best platform for your Java applications.  Oracle's relentless investment in the SPARC processor (6 NEW processors in 6 years, with more to come) is part of Larry Ellison's strategy to ensure his large enterprise customers can benefit by running their software applications on SPARC Solaris systems.  For example, the SPARC processor design is one that handles a high level of database transactions, huge memory bandwidth requirements and very low latency -- optimal for the most demanding Java workloads.  In fact, SPARC S7 systems are faster per core running Java workloads than the latest x86 compute systems, BY FAR.  Need some proof?
 - Go to: blogs.oracle.com/BestPerf  and search on "S7"
 - Read this blog (3rd paragraph):  New SPARC S7 Servers for Cloud & Scale-Out

2. Oracle Solaris 11 drives enterprise Java to new heights
Oracle Solaris 11 has been designed for enterprise Java applications and cloud deployments.  Important Solaris technologies which boost Java application performance include virtualization, dynamic threading, large page support, accelerated cryptography, Software Defined Networking (SDN) and more.  For example, on SDN, Oracle Solaris 11 has a socket based API for Application-Driven SDN. This API has been included in the JDK, and can help applications determine service level properties to allow them to control SLAs within the application itself.  And with additions to JDK in Java 8 (see JDK 8 install on Solaris), such as Lambda Expressions and a new Streams API, developers can now more easily write high-performance Java applications that are multi-threaded, and take advantage of the performance of SPARC multi-core platforms. And that is just skimming the foam off the top of your latte -- there is a a lot more behind the synergies of Solaris and Java.

3. Java applications ROCK on S7 -- Kicking butt on x86
The S7 offers industry leading performance and scalability on Java, which saves customers real $$.  Better core efficiency than x86 systems lowers costs for running Java applications and databases by reducing licensing costs.  This means you have to buy LESS CORES that you do with x86!  Recent public benchmark results have demonstrated the performance advantages of SPARC S7 and Java versus x86.  Check this out -- you will see that SPARC chips keep improving the per core performance, while overall x86 core efficiency is stalled:

 - An S7 core is 1.7x faster than Intel x86 on the SPECjEnterprise2010 benchmark.   [details]
 - An S7 core is 1.5x to 1.9x faster than Intel x86 on the SPECjbb2015 benchmark .   [details]
 - An S7 core is 1.5x to 2.1x faster than Intel x86 on the SPECjbb2015 “distributed” benchmark.   [details]

4. Hands-down, Big Data analytics applications will run BEST on SPARC S7
The Data Analytics Accelerators (DAX) capability in SPARC S7 (and M7) provides hardware acceleration to process a variety of big data analytic workloads on the cloud, while leaving the CPU cores free to focus on other tasks.  This 5-minute video shows how DAX work, but net-net is that hardware acceleration of cloud analytic applications on S7 can deliver 3x - 12x faster time to insight.  Here's one ISV who's drinking the Java:

QuartetFS is pushing the limits of Java to take advantage of SPARC systems for analytics.  They provide an in-memory computing analytical platform which delivers interactive calculations for a new generation of credit and market risk applications, and here's what they had to say about S7:
"We have been testing the SPARC M7 and S7 Data Analytics Accelerator (DAX) open APIs and initial results show our queries running an amazing 6-8 times faster. These results were achieved while we had up to 8x lower CPU usage, as the DAX engines offload the cores and free them up to do other work.”

There are similar results with the open source cluster computing framework Apache Spark (here).

5. SPARC systems have unrivaled tools for Java developers
Ask our Java developers what matters and they will quickly set you straight that it is about the tools available for efficient and effective development of the application.  Without them, the coffee never gets out of the cup (so to speak) -- or at least not for long after that first sip.  Managing the Java development process and maintaining the underlying infrastructure for Java services is exactly what the #1 development platform for Oracle SPARC systems does. 

Meet Oracle Developer Studio.  Specific to Java development, developers can get increased observability of their applications using a combination of Oracle Developer Studio Performance Analyzer and Java Mission Control/Flight Recorder. By using integration's with Oracle Solaris DTrace and enhanced DTrace probes in the Oracle Java VM, performance data can be visually analyzed for hot spots, thread deadlocks, and the Java heap including garbage collection right down to the hardware.  On top of that, optimizations for the SPARC processor have been made to the JVM itself, including Java security performance improvements, Java just-in-time compiler code generations, and enhancements in Java scalability. 

6. SPARC S7 is the MOST SECURE platform option, with SCREAMING encryption performance   
We hear about customers delaying the decision to turn on application encryption because they are worried about the performance impact on the application.  That is dangerous...  The S7 processor has the fastest cryptographic acceleration in the industry, delivering end-to-end data encryption and secure
transactions with near-zero performance impact to the Java application.  The incredible efficiency of this design is that these cryptographic
accelerators can rapidly process data for encryption while the S7 processor cycles are available for use by the database and other
applications. This 4-minute video demonstrates how this works.  Here's more data to show you that we really aren't kidding about this:
 - SHA Digest Encryption: SPARC S7 Performance, Beats Intel E5-2699 v4 Per Core Under Load
 - AES Encryption: SPARC S7 Performance, Beats Intel E5-2699 v4 Per Core Under Load
 - Real-Time Enterprise: SPARC S7-2 Advantage Per Core Under Load Compared to 2-Chip x86 E5-2699 v3
 - SPECjEnterprise2010: SPARC S7-2 Secure and Unsecure Results

In addition, the SPARC S7 & M7 processors features the Silicon Secured Memory (SSM) feature, which safeguards against  invalid or stale memory references and buffer overruns.  This 4-min video demonstrates its capabilities. Here is what AsiaInfo, a leading telecom network and software solutions provider in China, had to say about SSM:
" Our solutions are multi-threaded, memory intensive and response time sensitive and demand high performance. Working with Oracle and using the Silicon Secured Memory feature on SPARC M7 and S7 we have been able to shorten time to find and fix bugs by 3 days. This is a huge benefit as detecting and fixing memory access issues is normally a very difficult and time consuming process."

7. Java (JVM) is the language of the cloud -- and the design target for SPARC
Today, Java, and Java Virtual Machine (JVM), are the invisible force behind many of the applications
that power cloud computing
(see chart below). This includes mission critical applications
where performance, scale, and ease of management are vital
requirements.  The design target of SPARC S7 is scale-out processing in the cloud.  And since Oracle has acquired rights to Java from Sun Microsystems, it should come as no surprise
that they are investing heavily to ensure that Oracle's SPARC Solaris systems are optimized to be the
best platform in the world to deploy Java (JVM) applications in the cloud.

More coffee??

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.