Monday Aug 12, 2013

Compact Profiles Demonstrated

Following up on an article introducing compact profiles, the video that follows demonstrates how this new feature in the upcoming Java 8 release can be utilized.  The video:

  • Describes the compact profile feature and the rationale for its creation.
  • Shows how to use the new jrecreate utility to generate compact profiles that can be readily deployed.
  • Demonstrates that even the smallest of profiles (less than 11MB) is robust enough to support very popular and important software frameworks like OSGi.

The software demonstrated is in early access.  For those interested in trying it out before the formal release of Java 8, there are two options:

  1. Members of the Oracle Partner Network (OPN) with a gold membership or higher can download the early access Java 8 binaries of Java SE-Embedded shown here.  For those not at this level, it may still be possible to get early access software, but it will require a qualification process beforehand.
  2. It's not as intimidating as it sounds, you can pull down the source code for OpenJDK 8, and build it yourself.  By default, compact profiles are not built, but this forum post shows you how.  The reference platform for this software is linux/x86.  Functionally, the generated compact profiles will contain the pared down modules for each compact profile, but you'll find the footprint for each to be much larger than the ones demonstrated in this video, as none of the Java SE-Embedded space optimizations are performed by default.

Not having any premium privileges on YouTube, the maximum allowed length of a video is 15 minutes.  There's actually lots more to talk about with compact profiles, including enhancements to java tools and utilities (javac, jar, jdeps, and the java command itself) that have incorporated intelligence for dealing with profiles.

Hmm.  Maybe there's an opportunity for a Compact Profiles Demonstrated Part II?

Wednesday Jul 31, 2013

An Introduction to Java 8 Compact Profiles

Java SE is a very impressive platform indeed, but with all that functionality comes a large and ever increasing footprint.  It stands to reason then that one of the more frequent requests from the community has been the desire to deploy only those components required for a particular application instead of the entire Java SE runtime environment.  Referred to as subsetting, the benefits of such a concept would seem to be many:

  • A smaller Java environment would require less compute resources, thus opening up a new domain of devices previously thought to be too humble for Java.
  • A smaller runtime environment could be better optimized for performance and start up time.
  • Elimination of unused code is always a good idea from a security perspective.
  • If the environment could be pared down significantly, there may be tremendous benefit to bundling runtimes with each individual Java application.
  • These bundled applications could be downloaded more quickly.

Despite these perceived advantages, the platform stewards (Sun, then Oracle) have been steadfast in their resistance to subsetting.  The rationale for such a stance is quite simple: there was sincere concern that the Java SE platform would fragment.  Agree or disagree, the Java SE standard has remained remarkably in tact over time.  If you need any further evidence of this assertion, compare the state of Java SE to that of Java ME, particularly in the mobile telephony arena.  Better still, look how quickly Android has spawned countless variants in its brief lifespan.

Nonetheless, a formal effort has been underway having the stated goal of providing a much more modular Java platform.  Called Project Jigsaw, when complete, Java SE will be composed of a set of finer-grained modules and will include tools to enable developers to identify and isolate only those modules needed for their application.  However, implementing this massive internal change and yet maintaining compatibility has proven to be a considerable challenge.  Consequently full implementation of the modular Java platform has been delayed until Java 9.

Understanding that Java 9 is quite a ways off, an interim solution will be available for Java 8, called Compact Profiles.  Rather than specifying a complete module system, Java 8 will define subset profiles of the Java SE platform specification that developers can use to deploy.  At the current time three compact profiles have been defined, and have been assigned the creative names compact1, compact2, and compact3. The table that follows lists the packages that comprise each of the profiles.  Each successive profile is a superset of its predecessor.  That is to say, the compact2 profile contains all of the packages in compact1 plus those listed under the compact2 column below.  Likewise, compact3 contains all of compact2 packages plus the ones listed in the compact3 column.

compact1                     compact2                    compact3
--------------------------   -----------------------     --------------------------                      java.rmi                    java.lang.instrument
java.lang                    java.rmi.activation
java.lang.annotation         java.rmi.registry 
java.lang.invoke             java.rmi.server             java.util.prefs
java.lang.ref                java.sql                    javax.annotation.processing
java.lang.reflect            javax.rmi.ssl               javax.lang.model
java.math                    javax.sql                   javax.lang.model.element                     javax.transaction           javax.lang.model.type
java.nio                     javax.transaction.xa        javax.lang.model.util
java.nio.channels            javax.xml         
java.nio.channels.spi        javax.xml.datatype
java.nio.charset             javax.xml.namespace
java.nio.charset.spi         javax.xml.parsers 
java.nio.file.spi                  javax.xml.transform           javax.xml.transform.dom     javax.xml.transform.sax     javax.naming           javax.xml.transform.stax
java.text            javax.naming.event
java.text.spi                javax.xml.validation        javax.naming.ldap
java.util                    javax.xml.xpath             javax.naming.spi
java.util.concurrent         org.w3c.dom                 javax.script
java.util.concurrent.atomic  org.w3c.dom.bootstrap
java.util.jar                    javax.sql.rowset
java.util.logging            org.xml.sax                 javax.sql.rowset.serial
java.util.regex              org.xml.sax.ext             javax.sql.rowset.spi
java.util.spi                org.xml.sax.helpers                                            javax.xml.crypto
javax.crypto                                             javax.xml.crypto.dom
javax.crypto.interfaces                                  javax.xml.crypto.dsig
javax.crypto.spec                                        javax.xml.crypto.dsig.dom                                                javax.xml.crypto.dsig.keyinfo                                            javax.xml.crypto.dsig.spec                                      org.ieft.jgss

You may ask what savings can be realized by using compact profiles?  As Java 8 is in pre-release stage, numbers will change over time, but let's take a look at a snapshot early access build of Java SE-Embedded 8 for ARMv5/Linux.  A reasonably configured compact1 profile comes in at less than 14MB.  Compact2 is about 18MB and compact3 is in the neighborhood of 21MB.  For reference, the latest Java 7u21 SE Embedded ARMv5/Linux environment requires 45MB.

So at less than one-third the original size of the already space-optimized Java SE-Embedded release, you have a very capable runtime environment.  If you need the additional functionality provided by the compact2 and compact3 profiles or even the full VM, you have the option of deploying your application with them instead.

In the next installment, we'll look at Compact Profiles in a bit more detail.

Saturday Mar 16, 2013

Is it armhf or armel?

Arm processors come in all makes and sizes, a certain percentage of which address a market where cost, footprint and power requirements are at a premium.  In this space, the inclusion of even a floating point unit would be considered an unnecessary luxury.  To perform floating point operations with these processors, software emulation is required.

Higher-end Arm processors come bundled with additional capability that enables hardware execution of floating point operations.  The difference between these two architectures gave rise to two separate Embedded Application Binary Interfaces or EABIs for ARM: soft float and VFP (Vector Floating Point).  Although there is forward compatibility between soft and hard float, there is no backward compatibility.  And in fact, when it comes to providing binaries for Java SE Embedded for Arm, Oracle provides two separate options: a soft float binary and a VFP binary.  In the Linux community, releases built upon both these EABIs are refereed to as armel based distributions.

Enter armhf.  Although a big step up in performance, the VFP EABI utilizes less-than-optimal argument passing when a floating point operations take place.  In this scenario, floating point arguments must first be passed through integer registers prior to executing in the floating point unit.  A new EABI, referred to as armhf optimizes the calling convention for floating point operations by passing arguments directly into floating point registers.  It furthermore includes a more efficient system call convention.  The end result is applications compiled with the armhf standard should demonstrate modest performance improvement in some cases, and significant improvement for floating point intensive applications.

Alas, armhf represents yet another binary incompatible standard, but one that has already gained considerable traction in the community. Although still relatively early, the transition from armel to armhf is underway.  In fact, Ubuntu has already announced that future releases will only be built to the armhf standard, effectively obsoleting armel. As mentioned in Henrik's Stahl's Blog, an armhf version of Java SE Embedded is in the works, and we have already made available a armhf-based developer Preview of JDK 8 with JavaFX.

In the interim, we will have to deal with the incompatibilities between armel and armhf.  Most recently we've seen a rash of failed attempts to run the ArmV7 VFP Java SE Embedded binary on top of an armhf-based Linux distro.  During diagnosis, the question becomes, how can I determine whether my Linux distribution is based on armel or armhf?  Turns out this is not as straightforward as one might think.  Aside from experience and anecdotal evidence, one possible way to ascertain whether you're running on armel or armhf is to run the following obscure command:

$ readelf -A /proc/self/exe | grep Tag_ABI_VFP_args

If the Tag_ABI_VFP_args tag is found, then you're running on an armhf system.  If nothing is returned, then it's armel.  To show you an example, here's what happens on a Raspberry Pi running the Raspbian distribution:

pi@raspberrypi:~$ readelf -A /proc/self/exe | grep Tag_ABI_VFP_args
  Tag_ABI_VFP_args: VFP registers

This indicates an armhf distro, which in fact is what Raspbian is.  On the original, soft-float Debian Wheezy distribution, here's what happens:

pi@raspberrypi:~$ readelf -A /proc/self/exe | grep Tag_ABI_VFP_args

Nothing returned indicates that this is indeed armel.

Many thanks to the folks participating in this Raspberry Pi forum topic for providing this suggestion.

Wednesday Feb 06, 2013

Source Code for JavaFX Scoreboard Now Available

For the last few years, many of the JavaFX related blog posts found here have made reference to all or parts of a Scoreboard application written in JavaFX.  For example, the entry prior to this contains a video demonstrating how this Scoreboard application can be run on an embedded device such as a Raspberry Pi and displayed on an ordinary flat screen TV.

Alongside those posts, some have asked for access to the source code to this application.  I've always felt uncomfortable releasing the code, not under the delusion that it has any commercial value, but rather because, let us say, it was not developed with the strictest software engineering standards in mind.  Having gotten over those insecurities (not really), I've decided to place it in a project called javafx-scoreboard.

If you are interested in gaining access to this code you'll need to do the following:

  1. Register on  to get a account.  It's free and painless.  If you already have an account, terrific! You can skip this step.
  2. Send an email to me at  asking to be added to the javafx-scoreboard project.  Include your username in the email.
  3. Once you're added to the project, you'll have download access to the contents of the project.

For grins, here's document that gives an overview of the Scoreboard application.


Saturday Dec 22, 2012

A Raspberry Pi / JavaFX Electronic Scoreboard Application

As evidenced at the recently completed JavaOne 2012 conference, community excitement towards the Raspberry Pi and its potential as a Java development and deployment platform was readily palpable.  Fast forward three months, Oracle has announced the availability of a JDK 8 (with JavaFX) for Arm Early Access Developer Preview where the reference platform for this release is none other than the Raspberry Pi.

What makes this especially interesting to me is the addition of JavaFX to the Java SE-Embedded 8 platform.  It turns out that at $35US, the (not so) humble Raspberry Pi has a very capable graphics processor, opening up a Pandora's box of graphics applications that could be applied to this beloved device.  As a first step in becoming familiar with just how this works, I decided to dust off a two year old JavaFX scoreboard application, originally written for a Windows laptop, and see how it would run on the Pi.  Low and behold, the application runs unmodified (without even a recompile).

The video that follows shows how an ordinary flat screen TV can be converted into a full screen electronic scoreboard driven by a Raspberry Pi.  The requirements for such a solution are incredibly straightforward: (1) the TV needs access to a power receptacle and (2) it must be within range of a WiFi network in order to receive scoreboard update packets.  The device is so compact and miserly from a power perspective, that we velcro the Pi to the back of the TV and get our power from the TV's USB port. If you can spare a few moments, it just might be worth your while to take a look.

Monday Nov 05, 2012

Sprinkle Some Magik on that Java Virtual Machine

GE Energy, through its Smallworld subsidiary, has been providing geospatial software solutions to the utility and telco markets for over 20 years.  One of the fundamental building blocks of their technology is a dynamically-typed object oriented programming language called Magik.  Like Java, Magik source code is compiled down to bytecodes that run on a virtual machine -- in this case the Magik Virtual Machine.

Throughout the years, GE has invested considerable engineering talent in the support and maintenance of this virtual machine.  At the same time vast energy and resources have been invested in the Java Virtual Machine. The question for GE has been whether to continue to make that investment on its own or to leverage massive effort provided by the Java community? Utilizing the Java Virtual Machine instead of maintaining its own virtual machine would give GE more opportunity to focus on application solutions.  

At last count, there are dozens, perhaps hundreds of examples of programming languages that have been hosted atop the Java Virtual Machine.  Prior to the release of Java 7, that effort, although certainly possible, was generally less than optimal for languages like Magik because of its dynamic nature.  Java, as a statically typed language had little use for this capability.  In the quest to be a more universal virtual machine, Java 7, via JSR-292, introduced a new bytecode called invokedynamic.  In short, invokedynamic affords a more flexible method call mechanism needed by dynamic languages like Magik.

With this new capability GE Energy has succeeded in hosting their Magik environment on top of the Java Virtual Machine.  So you may ask, why would GE wish to do such a thing?  The benefits are many:

  • Competitors to GE Energy claimed that the Magik environment was proprietary.  By utilizing the Java Virtual Machine, that argument gets put to bed.  JVM development is done in open source, where contributions are made world-wide by all types of organizations and individuals.
  • The unprecedented wealth of class libraries and applications written for the Java platform are now opened up to Magik/JVM platform as first class citizens.
  • In addition, the Magik/JVM solution vastly increases the developer pool to include the 9 million Java developers -- the largest developer community on the planet.
  • Applications running on the JVM showed substantial performance gains, in some cases as much as a 5x speed up over the original Magik platform.
  • Legacy Magik applications can still run on the original platform.  They can be seamlessly migrated to run on the JVM by simply recompiling the source code.
  • GE can now leverage the huge Java community.  Undeniably the best virtual machine ever created, hundreds if not thousands of world class developers continually improve, poke, prod and scrutinize all aspects of the Java platform.  As enhancements are made, GE automatically gains access to these.
  • As Magik has little in the way of support for multi-threading, GE will benefit from current and future Java offerings (e.g. lambda expressions) that aim to further facilitate multi-core/multi-threaded application development.
  • As the JVM is available for many more platforms, it broadens the reach of Magik, including the potential to run on a class devices never envisioned just a few short years ago.  For example, Java SE compatible runtime environments are available for popular embedded ARM/Intel/PowerPC configurations that could theoretically host this software too.
As compared to other JVM language projects, the Magik integration differs in that it represents a serious commercial entity betting a sizable part of its business on the success of this effort.  Expect to see announcements not only from General Electric, but other organizations as they realize the benefits of utilizing the Java Virtual Machine.

Friday Jun 22, 2012

Healthcare Mobile Database Synchronization Demonstration

Like many of you, I learn best by getting my hands dirty.  When confronted with the task of understanding a new set of products and technologies and figuring out how they might apply to a vertical industry like healthcare, I set out to create a demonstration.  The video that follows aims to show how the Oracle embedded software portfolio can be applied to a healthcare application.  The demonstration utilizes among others, Java SE Embedded, Berkeley DB, Apache Tomcat, Oracle 11gR2 and Oracle Database Mobile Server.

Eric Jensen gives a great critique and description of the demo here.  To sum it up, we aim to show how live medical data can be collected on a medical device, stored in a local database, synchronized to a master database and furthermore propagated to a mobile phone (Android) application.  Come take a look!

Monday Mar 19, 2012

Take Two: Comparing JVMs on ARM/Linux

Although the intent of the previous article, entitled Comparing JVMs on ARM/Linux, was to introduce and highlight the availability of the HotSpot server compiler (referred to as c2) for Java SE-Embedded ARM v7,  it seems, based on feedback, that everyone was more interested in the OpenJDK comparisons to Java SE-E.  But there were two main concerns:

  • The fact that the previous article compared Java SE-E 7 against OpenJDK 6 might be construed as an unlevel playing field because version 7 is newer and therefore potentially more optimized.
  • That the generic compiler settings chosen to build the OpenJDK implementations did not put those versions in a particularly favorable light.

With those considerations in mind, we'll institute the following changes to this version of the benchmarking:

  1. In order to help alleviate an additional concern that there is some sort of benchmark bias, we'll use a different suite, called DaCapo.  Funded and supported by many prestigious organizations, DaCapo's aim is to benchmark real world applications.  Further information about DaCapo can be found at
  2. At the suggestion of Xerxes Ranby, who has been a great help through this entire exercise, a newer Linux distribution will be used to assure that the OpenJDK implementations were built with more optimal compiler settings.  The Linux distribution in this instance is Ubuntu 11.10 Oneiric Ocelot.
  3. Having experienced difficulties getting Ubuntu 11.10 to run on the original D2Plug ARMv7 platform, for these benchmarks, we'll switch to an embedded system that has a supported Ubuntu 11.10 release.  That platform is the Freescale i.MX53 Quick Start Board.  It has an ARMv7 Coretex-A8 processor running at 1GHz with 1GB RAM.
  4. We'll limit comparisons to 4 JVM implementations:
    • Java SE-E 7 Update 2 c1 compiler (default)
    • Java SE-E 6 Update 30 (c1 compiler is the only option)
    • OpenJDK 6 IcedTea6 1.11pre 6b23~pre11-0ubuntu1.11.10.2 CACAO build 1.1.0pre2
    • OpenJDK 6 IcedTea6 1.11pre 6b23~pre11-0ubuntu1.11.10.2 JamVM build-1.6.0-devel

Certain OpenJDK implementations were eliminated from this round of testing for the simple reason that their performance was not competitive.  The Java SE 7u2 c2 compiler was also removed because although quite respectable, it did not perform as well as the c1 compilers.  Recall that c2 works optimally in long-lived situations.  Many of these benchmarks completed in a relatively short period of time.  To get a feel for where c2 shines, take a look at the first chart in this blog.

The first chart that follows includes performance of all benchmark runs on all platforms.  Later on we'll look more at individual tests.  In all runs, smaller means faster.  The DaCapo aficionado may notice that only 10 of the 14 DaCapo tests for this version were executed.  The reason for this is that these 10 tests represent the only ones successfully completed by all 4 JVMs.  Only the Java SE-E 6u30 could successfully run all of the tests.  Both OpenJDK instances not only failed to complete certain tests, but also experienced VM aborts too.

One of the first observations that can be made between Java SE-E 6 and 7 is that, for all intents and purposes, they are on par with regards to performance.  While it is a fact that successive Java SE releases add additional optimizations, it is also true that Java SE 7 introduces additional complexity to the Java platform thus balancing out any potential performance gains at this point.  We are still early into Java SE 7.  We would expect further performance enhancements for Java SE-E 7 in future updates.

In comparing Java SE-E to OpenJDK performance, among both OpenJDK VMs, Cacao results are respectable in 4 of the 10 tests.  The charts that follow show the individual results of those four tests.  Both Java SE-E versions do win every test and outperform Cacao in the range of 9% to 55%.

For the remaining 6 tests, Java SE-E significantly outperforms Cacao in the range of 114% to 311%

So it looks like OpenJDK results are mixed for this round of benchmarks.  In some cases, performance looks to have improved.  But in a majority of instances, OpenJDK still lags behind Java SE-Embedded considerably.

Time to put on my asbestos suit.  Let the flames begin...

Tuesday Feb 14, 2012

Comparing JVMs on ARM/Linux

For quite some time, Java Standard Edition releases have included both client and server bytecode compilers (referred to as c1 and c2 respectively), whereas Java SE-Embedded binaries only contained the client c1 compiler.  The rationale for excluding c2 stems from the fact that (1) eliminating optional components saves space, where in the embedded world, space is at a premium, and (2) embedded platforms were not given serious consideration for handling server-like workloads.  But all that is about to change.  In anticipation of the ARM processor's legitimate entrance into the server market (see Calxeda), Oracle has, with the latest update of Java SE-Embedded (7u2), made the c2 compiler available for ARMv7/Linux platforms, further enhancing performance for a large class of traditional server applications. 

These two compilers go about their business in different ways.  Of the two, c1 is a lighter optimizing compiler, but has faster start up.  It delivers excellent performance and as the default bytecode compiler, works extremely well in almost all situations.  Compared to c1, c2 is the more aggressive optimizer and is suited for long-lived java processes.  Although slower at start up, it can be shown to achieve better performance over time.  As a case in point, take a look at the graph that follows.

One of the most popular Java-based applications, Apache Tomcat, was installed on an ARMv7/Linux device.   The chart shows the relative performance, as defined by mean HTTP request time, of the Tomcat server run with the c1 client compiler (red line) and the c2 server compiler (blue line).  The HTTP request load was generated by an external system on a dedicated network utilizing the ab (Apache Bench) program.  The closer the response time is to zero the better, you can see that for the initial run of 25,000 HTTP requests, the c1 compiler produces faster average response times than c2.  It takes time for the c2 compiler to "warm up", but once the threshold of 50,000 or so requests is met, the c2 compiler performance is superior to c1.  At 250,000 HTTP requests, mean response time for the c2-based Tomcat server instance is 14% faster than its c1 counterpart.

It is important to realize that c2 assumes, and indeed requires more resources (i.e. memory).  Our sample device with 1GB RAM, was more than adequate for these rounds of tests.  Of course your mileage may vary, but if you have the right hardware and the right workload, give c2 a further look.

While discussing these results with a few of my compadres, it was suggested that OpenJDK and some of its variants be included in on this comparison.  The following chart shows mean http request times for 6 different configurations:

  1. Java SE Embedded 7u2 c1 Client Compiler
  2. Java SE Embedded 7u2 c2 Server Compiler
  3. OpenJDK Zero VM (build 20.0-b12, mixed mode) OpenJDK 1.6.0_24-b24 (IcedTea6 1.12pre)
  4. JamVM (build 1.6.0-devel, inline-threaded interpreter with stack-caching) OpenJDK 1.6.0_24-b24 (IcedTea6 1.12pre)
  5. CACAO (build 1.1.0pre2, compiled mode) OpenJDK 1.6.0_24-b24 (IcedTea6 1.12pre)
  6. Interpreter only: OpenJDK Zero VM (build 20.0-b12, interpreted mode) OpenJDK 1.6.0_24-b24 (IcedTea6 1.12pre)

Results remain pretty much unchanged, so only the first 4 runs (25K-100K requests) are shown.  As can be seen, The Java SE-E VMs are on the order of 3-5x faster than their OpenJDK counterparts irrespective of the bytecode compiler chosen.  One additional promising VM called shark was not included in these tests because, although it built from source successfully, it failed to run Apache Tomcat.  In defense of shark, the ARM version may still be in development (i.e. non-stable) mode.

Creating a really fast virtual machine is hard work and takes a lot of time to perfect.  Considering the resources expended by Oracle (and formerly Sun), it is no surprise that the commercial Java SE VMs are excellent performers.  But the extent to which they outperform their OpenJDK counterparts is surprising.  It would be no shock if someone in the know could demonstrate better OpenJDK results.  But herein lies one considerable problem:  it is an exercise in patience and perseverance just to locate and build a proper OpenJDK platform suitable for a particular CPU/Linux configuration.  No offense would be taken if corrections were presented, and a straightforward mechanism to support these OpenJDK builds were provided.

Tuesday Jan 03, 2012

Tomcat Micro Cluster

The term Micro Server has been bandied about recently as a means to provide a certain class of server functionality. As embedded systems continue their inexorable drive towards better performance, and standard hardware/software architectures become ubiquitous, the notion of using low-cost, low-power, small-footprint devices as servers becomes quite realistic.  Just as data center managers have utilized multitudes of affordable rack mount servers to provide scalability, why not duplicate that effort with these off-the-shelf devices?

The video that follows takes the Micro Server to its next logical evolution: The Micro Cluster.  Built from commodity hardware (and by commodity I mean The Home Depot), the cluster board has a rack mount form factor that can house 12 Plug Computers.  As the Java SE HotSpot Virtual Machine is available for the Plug Computer (ArmV5/Linux), we'll utilize Apache Tomcat to demonstrate a Tomcat Micro Cluster.  Over time, as the individual compute nodes increase in performance and capacity, this should become even more compelling.

Monday Oct 24, 2011

Java ONE 2011 Hands on Lab for Java SE Embedded

Now that the dust has settled, sincere thanks go out to my compadres (you know who you are) for helping make The Java One 2011 Java SE Embedded Hands on Lab such a success.  In fact it was so well received, our peers in Asia are already planning on replicating the effort for JavaOne Tokyo in April 2012.  In addition to the Tokyo event,  we hope to provide future opportunities for this venue elsewhere.  In the interim, we'd seriously consider hosting this lab, albeit on a smaller scale (Java One had 105 networked devices and workstations), for interested customers.  To give you a feel for the lab contents, here's a synopsis:

Java One 2011 Hands on Lab 24642: Java SE Embedded Development Made Easy

This Hands-on Lab aims to show that developers already familiar with the Java develop/debug/deploy lifecycle can apply those same skills to develop Java applications, using Java SE Embedded, on embedded devices.  Each participant in the lab will:

  • Have their own individual embedded device to gain valuable hands on experience
  • Turn their embedded device into a Java Servlet container
  • Learn how to deploy embedded Java applications, developed with the NetBeans IDE, onto their device
  • Learn how embedded Java applications can be remotely debugged from their desktop NetBeans IDE
  • Learn how to remotely monitor and manage embedded Java applications from their desktop
If the logistics for setting up a lab prove to be a bit too much, as an alternative, we've given quite a few live presentations/demonstrations with similar flair.  So please, by all means, contact me at, if you're interested in learning more.  For those of you who run developer user groups, most notably Java User Groups, and are looking for a speaker at your next meeting, please consider us.  We will not disappoint.

Friday Aug 19, 2011

Serial Port Communication for Java SE Embedded

The need to communicate with devices connected to serial ports is a common application requirement.  Falling outside the purview of the Java SE platform, serial and parallel port communication has been addressed with a project called RXTX.  (In the past, you may have known this as javacomm).  With RXTX,  Java developers access serial ports through the RXTXcomm.jar file.  Alongside this jar file, an underlying native layer must be provided to interface with the operating system's UART ports.  For the usual suspects (Windows, Linux/x86, MacOS, Solaris/Sparc), pre-compiled binaries are readily available.  To host this on an alternative platform, some (hopefully minimal) amount of work is required.

Here's hoping the following notes/observations might aid in helping you to build RXTX for an embedded device utilizing one of our Java SE Embedded binaries.  The device used for this particular implementation is my current favorite: the Plug Computer.

Notes on Getting RX/TX 2.1-7-r2 Working on a Plug Computer

1. At this early juncture with Java 7, be wary of mixing Java 7 with code from older versions of Java. The class files generated by the JDK7 javac compiler contain an updated version byte with a value that results in older (Java 6 and before) JVMs refusing to load these classes.

2. The RXTX download location has binaries for many platforms including Arm variants, but none that worked for the Plug Computer, so one had to be built from source.

3. Using the native GCC for the Plug Computer and the RXTX source, binaries (native shared objects) were compiled for the armv5tel-unknown-linux-gnu platform.

4. The RXTX "stable" source code found at the aforementioned site is based on version rxtx 2.1-7r2.  This code appears to be pretty long in the tooth, in that it has no knowledge of Java 6.  Some changes need to be made to accommodate a JDK 6 environment.  Without these modifications, RXTX will not build with JDK6

SUGGESTED FIX, most elegant, not recommended:
Edit the file in the source directory and look for the following:

    case $JAVA_VERSION in

and change the second line to:


Upon modification, the script found in the rxtx source directory must be re-run to recreate the ./configure script.  Unfortunately, this requires loading the autoconf, automake and libtool packages (plus dependencies) and ended up resulting in libtool incompatibilies when running the resultant ./configure script.

Instead, edit ./configure and search for the occurrences (there are more than one) of

    case $JAVA_VERSION in

and change the second line to:


Run './configure', then 'make' to generate the RXTXcomm.jar and platform specific .so shared object libraries.

5. You may also notice in the output of the make, that there were compilation errors for source files which failed to find the meaning of "UTS_RELEASE".  This results in some of the shared object files not being created.  These pertain to the non-serial aspects of RXTX.  As we were only interested in, this was no problem for us.

6. Once built, move the following files into the following directories:

    # cd rxtx-2.1-7-r2/
    # cp RXTXcomm.jar $JAVA_HOME/lib/ext
    # cd armv5tel-unknown-linux-gnu/.libs/
    # cp $JAVA_HOME/lib/arm
    # cd $JAVA_HOME/lib/arm
    # ln -s

Now Java applications which utilize RXTX should run without any java command-line additions.

The RXTXcomm.jar file can be downloaded here.  To spare you the effort, a few pre-built versions of are provided at this location:

If you've gone through this exercise on any additional architectures, send them my way and I'll post them here.

Tuesday Jun 21, 2011

Observations in Migrating from JavaFX Script to JavaFX 2.0

Observations in Migrating from JavaFX Script to JavaFX 2.0


Having been available for a few years now, there is a decent body of work written for JavaFX using the JavaFX Script language. With the general availability announcement of JavaFX 2.0 Beta, the natural question arises about converting the legacy code over to the new JavaFX 2.0 platform. This article reflects on some of the observations encountered while porting source code over from JavaFX Script to the new JavaFX API paradigm.

The Application

The program chosen for migration is an implementation of the Sudoku game and serves as a reference application for the book JavaFX – Developing Rich Internet Applications. The design of the program can be divided into two major components: (1) A user interface (ideally suited for JavaFX design) and (2) the puzzle generator. For the context of this article, our primary interest lies in the user interface. The puzzle generator code was lifted from a project and is written entirely in Java. Regardless which version of the UI we choose (JavaFX Script vs. JavaFX 2.0), no code changes were required for the puzzle generator code.

The original user interface for the JavaFX Sudoku application was written exclusively in JavaFX Script, and as such is a suitable candidate to convert over to the new JavaFX 2.0 model. However, a few notable points are worth mentioning about this program. First off, it was written in the JavaFX 1.1 timeframe, where certain capabilities of the JavaFX framework were as of yet unavailable. Citing two examples, this program creates many of its own UI controls from scratch because the built-in controls were yet to be introduced. In addition, layout of graphical nodes is done in a very manual manner, again because much of the automatic layout capabilities were in flux at the time. It is worth considering that this program was written at a time when most of us were just coming up to speed on this technology. One would think that having the opportunity to recreate this application anew, it would look a lot different from the current version.

Comparing the Size of the Source Code

An attempt was made to convert each of the original UI JavaFX Script source files (suffixed with .fx) over to a Java counterpart. Due to language feature differences, there are a small number of source files which only exist in one version or the other. The table below summarizes the size of each of the source files.

JavaFX Script source file

Number of Lines

Number of Character

JavaFX 2.0 Java source file

Number of Lines

Number of Characters






































































A few notes about this table are in order:

  • The number of lines in each file was determined by running the Unix ‘wc –l’ command over each file.
  • The number of characters in each file was determined by running the Unix ‘ls –l’ command over each file.
  • The examination of the code could certainly be much more rigorous. No standard formatting was performed on these files.  All comments however were deleted.

There was a certain expectation that the new Java version would require more lines of code than the original JavaFX script version. As evidenced by a count of the total number of lines, the Java version has about 22% more lines than its FX Script counterpart.

Furthermore, there was an additional expectation that the Java version would be more verbose in terms of the total number of characters.  In fact the preceding data shows that on average the Java source files contain fewer characters per line than the FX files.  But that's not the whole story.  Upon further examination, the FX Script source files had a disproportionate number of blank characters.  Why?  Because of the nature of how one develops JavaFX Script code.  The object literal dominates FX Script code.  Its not uncommon to see object literals indented halfway across the page, consuming lots of meaningless space characters.

RAM consumption

Not the most scientific analysis, memory usage for the application was examined on a Windows Vista system by running the Windows Task Manager and viewing how much memory was being consumed by the Sudoku version in question. Roughly speaking, the FX script version, after startup, had a RAM footprint of about 90MB and remained pretty much the same size. The Java version started out at about 55MB and maintained that size throughout its execution.

What About Binding?

Arguably, the most striking observation about the conversion from JavaFX Script to JavaFX 2.0 concerned the need for data synchronization, or lack thereof. In JavaFX Script, the primary means to synchronize data is via the bind expression (using the “bind” keyword), and perhaps to a lesser extent it’s “on replace” cousin. The bind keyword does not exist in Java, so for JavaFX 2.0 a Data Binding API has been introduced as a replacement.

To give a feel for the difference between the two versions of the Sudoku program, the table that follows indicates how many binds were required for each source file. For JavaFX Script files, this was ascertained by simply counting the number of occurrences of the bind keyword. As can be seen, binding had been used frequently in the JavaFX Script version (and does not take into consideration an additional half dozen or so “on replace” triggers). The JavaFX 2.0 program achieves the same functionality as the original JavaFX Script version, yet the equivalent of binding was only needed twice throughout the Java version of the source code.

JavaFX Script source file

Number of Binds

JavaFX Next Java source file

Number of “Binds”













































As the JavaFX 2.0 technology is so new, and experience with the platform is the same, it is possible and indeed probable that some of the observations noted in the preceding article may not apply across other attempts at migrating applications. That being said, this first experience indicates that the migrated Java code will likely be larger, though not extensively so, than the original Java FX Script source. Furthermore, although very important, it appears that the requirements for data synchronization via binding, may be significantly less with the new platform.

Wednesday Jun 01, 2011

Java SE Embedded Development Made Easy

Slowly but surely this old dog (who can learn new tricks, but at a snail's pace) came to the realization that although still quite relevant, a whole generation of people prefer not to read lengthy writings, but would rather digest information in small pieces using new media formats.  Thus the rationale for the following blog...

Certainly no thespian when it comes to public speaking, I will say this:  based upon my experience demonstrating Java SE on embedded devices, people have definitely expressed genuine interest.  Maybe it was the cool device (i.e. Plug Computer) which was used, or maybe this combination of hardware and software inspired the audience to think of the possibilities presented by this new platform.  Either way, I thought it might make sense to capture a shortened presentation/demonstration session.  Following is a 30 minute session broken down into two 15 minute videos (because YouTube only allows videos of 15 minutes or less for mere mortals). They aim to demonstrate how developers already familiar with the Java SE development paradigm can leverage that knowledge to seamlessly develop on very capable embedded processors.  Happy viewing!

Tuesday Mar 15, 2011

The Unofficial Java SE Embedded SDK

Developing applications for embedded platforms gets simpler all the time, thanks in part to the tremendous advances in microprocessor design and software tools.  And in particular, with the availability of Java SE compatible Virtual Machines for the popular embedded platforms, development has never been more straightforward.

The real beauty behind Java SE Embedded development lies in the fact that you can use your favorite IDE (Eclipse, NetBeans, JDeveloper ...) to create, test and debug code in the identical fashion in which you'd develop a standard desktop or server application.  When the time comes to try it out on a Java SE Embedded capable device, it's just a matter of shipping the bytecodes over to the device and letting it run.  There is no need for complicated emulators, toolchains and cross-compilers.  The exact same bytecodes that ran on your PC, run unmodified on the embedded device.

In fact, because all versions of Java SE (embedded or not) share a considerable amount of common code, we have plenty of anecdotal evidence which supports the notion that behavior -- correct or incorrect -- manifests itself identically across platforms.  We refer specifically here to bugs.  Now no one wants bugs, but believe it or not, our customers like the fact that behavior is consistent across platforms whether it's right or not. "Bug for bug compatibility" has actually become a strong selling point!

Having espoused the virtues of transparently developing off device, many still wish to test and debug on-device regularly as part of their development cycle.  If you're the touchy/feely type, there are ample examples of affordable and supported off-the-shelf devices that could fit the bill for an Unofficial Java SE Embedded SDK.  One such platform is the Plug Computer.

The reference platform for the Plug Computer is supplied by Marvell Technology Group. Manufacturers then license the technology from Marvell to create their own specific implementations.  Two such vendors are GlobalScale and Ionics.  These are incredibly capable devices that include Arm processors in the 1.2GHz to 2.0GHz range, and sport 512MB of RAM and flash.  There are a host of external port and interface options including USB, µUSB, SATA, GBE, SD, WiFi, ZigBee, Z-Wave and soon HDMI.  Additionally, several Linux distros are available for these systems too.  The typical cost for a base model is $99, and perhaps the most disruptive aspect of these systems, they consume on average about 5 watts of power.

Alongside developing in the traditional manner, the ability to step through and examine state on these devices via remote debugging comes as a standard feature with the Java SE-E VM.  Furthermore, you can use the JConsole application from your desktop to remotely monitor performance and resource consumption on the device.

So what would a bill of materials look like for The Unofficial Java SE Embedded SDK?  Pretty simple actually:

That's about it.  Of course, for higher level functionality, you can add additional packages.  For example, Apache runs beautifully here.  Could anyone imagine a large number of these devices acting as a parallel web server?


Jim Connors-Oracle


« June 2016