Tuesday Aug 11, 2009

JavaFX Scenegraph Performance Revisited

Prior to the release of JavaFX 1.2, an earlier blog post explained how creating an application with a scenegraph containing a large number of nodes can have performance implications.  That entry was subsequently picked up by Java Lobby and recently posted here.  Partly because it was a few months old, it resulted in a rash of, shall we say, interesting comments.

As one commenter pointed out, the initial results represent JavaFX performance in the 1.0/1.1 timeframe.  JavaFX 1.2 has since been released, and performance has improved substantially.  As a case in point, you can click on the image below to run the clock application.  This same application, developed and compiled with JavaFX 1.1 can be run from the previous blog post.  Further instructions are there

This application, compiled with JavaFX 1.2 and run on identical hardware, uses about a third of the CPU resources of the original version.  Specifically, using OpenSolaris vmstat(1M) to monitor CPU usage, the following average statistics were collected for one minute when the clock display is updated every 10th of a second.  The abbreviations have the following meanings:

  • us = percentage usage of CPU time in user space
  • sy = percentage usage of CPU time in system space
  • id = percentage usage of CPU time idling
  • The sum of (us + sy + id) should approximate 100%.

And here are the utilization numbers:

 Version  # Nodes per Digit
 CPU Utilization
 BulbClockNode JavaFX 1.1
 27 BulbNodes
 us: 22%  sy: 2%  id: 76%
 BulbClockNode JavaFX 1.2  27 BulbNodes  us: 7%  sy: 2%  id: 91%

Yes,  performance has improved significantly and will continue to do so.  In fact, the JavaFX team is promising even better results with the advent of JavaFX 1.3 (code named SoMa), when considerable refining of the underlying architecture will take place.  At this stage in it's lifecycle, it's important to "subscribe" to the JavaFX technology.  Advances are coming fast and furious, and they don't promise to slow down anytime soon.

Friday Jul 24, 2009

Getting the JavaFX Screen Size

In a previous post, I showed one way in which you could size the nodes in your scenegraph relative to the screen size without having to use binding, thus eliminating the potential for a bindstorm.  At the time, the suggestion was to get the screen coordiantes via a call to AWT (Advanced Windowing Toolkit) as follows:

 var AWTtoolkit = java.awt.Toolkit.getDefaultToolkit ();
var screenSizeFromAWT = AWTtoolkit.getScreenSize ();

This currently works for JavaFX Desktop deployments but is far from an optimal solution primarily because it is non-portable.  There is no guarantee that AWT will exist-- in fact I'm pretty sure it will not exist -- in the JavaFX Mobile and JavaFX TV space.  So attempting to utilize the code snippet above in a JavaFX Mobile or TV context will result in an error.  Unfortunately at the time of the original post (Java FX 1.1), I didn't know of a better alternative.

With the advent of JavaFX 1.2, this problem is solved.   A new class called javafx.Stage.Screen  is provided which describes the characteristics of the display.  The API definition of the class can be found here.  So now you can query the JavaFX runtime in a portable fashion to get the screen coordinates as follows:

 import javafx.stage.\*;
 //set Stage boundaries to consume entire screen
 Stage {

fullscreen: true width: Screen.primary.bounds.minX as Integer height: Screen.primary.bounds.minY as Integer ... }

 Cheers.

 

Wednesday Jun 10, 2009

Early Access version of Java RTS 2.2 Released

Just in time for the start of JavaOne 2009, Sun released the latest version of its real-time Java platform, Java RTS 2.2, in early access form.  New features of this release include but are not limited to:

  • OS Platform support for Solaris 10 (x86 and Sparc - Update 5 & 6), Red Hat MRG 1.1 and  SUSE Linux Enterprise Real-Time 10 (SP2 Update 4)
  • 64-bit support for the three operating systems mentioned above. In addition, 32-bit support continues to be supported.
  • Faster throughput. Java RTS 2.2 utilizes a new dynamic (JIT) compiler targeted for Java SE 7 which contains several performance improvements including a new register allocator.
  • TSV (Thread Scheduling Visualizer) enhancements.
  • NetBeans plug-in support for Java RTS and TSV. 
  • New Initialization-Time-Compilation (ITC) options including the ability to wild card a set of methods vs listing single methods and support for specifying compilation based on thread type.
  • Auto-tuning and startup improvements made to the Real-time Garbage Collector (RTGC).
  • New documentation including a Java RTS Quick Start Guide to get new users started quickly with minimum programming and configuration.

A 90-day evaluation is available here: http://java.sun.com/javase/technologies/realtime/index.jsp.  For more detailed information on OS support and new functionality, check out the 2.2 Release Notes.

Tuesday Jun 09, 2009

JavaFX Book Arrives!

Since the beginning, the deadlines set forth for producing our book have been aggressive to say the least.  Ultimately, we wanted them to be physically available for sale at the JavaOne 2009 conference.  These things arrived literally hot off the press, having been flown in the opening day of the show.  This effort was not in vain; the results are in.  According to our friends at Digital Guru, JavaFX was definitely the hot topic of the week and JavaFX - Developing Rich Internet Applications was the best seller.  Furthermore as of 9:52AM Eastern Time 09-June-2009, Amazon reports that its sales rank is a very respectable 21,873.  Not bad for just a few days showing.

These sales were no doubt fueled by Larry Ellison's cameo appearance during the first JavaOne keynote.  Clearly hampered by what he could and could not say due to Oracle's pending acquisition of Sun, Larry went out of his way to promote the future of JavaFX.  He must have mentioned it a half dozen times during his 10 minute or so appearance.  On behalf of Eric and the other Jim, thanks bud.



Monday May 11, 2009

You Are What You Eat?

So you may ask, what effect does the food we eat have on our blood pressure?  My experience, documented below, won't likely hold up to whole lot of scientific scrutiny, but it's good enough lesson for me.

A combination of poor eating habits and genetics lands me in that ever-increasing group of individuals who have high blood pressure.  At first I was what you might consider borderline hypertensive.  Being thin and active, in conjunction with my doctor, we only monitored my levels to make sure they didn't get worse.  Unfortunately they did.

In an attempt at avoiding medication, I started seriously watching my sodium intake.  The resulting change in blood pressure was noticeable in pretty short order.   See the table below for daily readings for the last two weeks.  In general, I try to take my reading at or around 8:00AM if possible.

Having spent my whole life eating pretty much what I please, it's not easy to make this lifestyle change.   But with the help of loved ones, things have been going quite well.  But then came Mother's day (5/10).   A combination of brunch with my parents and dinner with my in-laws put the kabash on any diet plans.  I knew this would happen, and was frankly looking forward to the feast.  It only took one day for my pressure to skyrocket.  Amazing.  But man did I enjoy eating the waffles, bacon, crumbcake, bagels, Mimosas (Orange Juice and Champagne) cookies, macaroni, meatballs, turkey, chocoloate cake, Coke, Espresso, wine, Sambuca and Cognac to name a few.


 Date  Time  Blood Pressure
 4/29 8:01 AM
 124/73
 4/30 8:19 AM
 124/73
 5/1 8:18 AM
 123/71
 5/2 7:48 AM
 119/75
 5/3 9:22 AM
 126/74
 5/4 7:53 AM
 129/79
 5/5 8:20 AM
 119/69
 5/6 7:16 AM
 126/75
 5/7 6:19 AM
 126/72
 5/8 6:56 AM
 125/75
 5/9 7:48 AM
 133/71
 5/10 9:14 AM
 129/73
 5/11 7:56 AM
 159/81

Back to the bland diet, that is, until the next feast.

Friday May 01, 2009

JavaFX Book Available Online at Safari.com

Our book, entitled JavaFX: Developing Rich Internet Applications, having gone through the copyedit and proof stages, is now available online at Safari @ http://my.safaribooksonline.com/9780137013524.  Please feel free to provide feedback/comments on any aspect of the book that you've chanced to read.  We're still on schedule to have hardcopies available at the Java ONE show which will start on June 2.  If you plan on attending, please stop by the Addison-Wesley/Pearson/SMI Press booth.  Happy reading.


Monday Apr 06, 2009

Node Count and JavaFX Performance

In a recent blog entry entitled Best Practices for JavaFX Mobile Applications (Part 2), Michael Heinrichs espouses that keeping the scenegraph as small as possible helps JavaFX applications perform optimally. Regardless what version of JavaFX you're using, this is sage advice.  Having spent some time trying to create components for a scoreboard-like application, I was concerned over the amount of CPU time being consumed by the clock component pictured directly below.


You can click on the preceding image to run this mini application via Java WebStart.   By placing your mouse over any of the digits and typing in, via keyboard input, a valid number, you can set the clock.  Clicking on the "START/STOP" text will toggle the clock on and off.  Like many scoreboard clocks, when the time remaining is less than one minute, 10ths of seconds are displayed.  It is during this phase, when digits are being updated every tenth of a second, that this application can be especially taxing.  You can imagine how troublesome this clock might be if it were to be part of say a hockey scoreboard which could have an additional 4 penalty clocks ticking simultaneously.

The major factor affecting performance appears to be the sheer number of nodes in the scenegraph that require recalculation for every clock tick.  For this first implementation, each of the five clock digits is comprised of 27 BulbNodes, (my naming) which are switched on or off depending upon what value needs to be displayed.

In an effort to see how decreasing the node count might effect performance, this second implementation of the clock component uses the same underlying framework, except that each digit is now composed of 7 LED SegmentNodes (my naming again) instead of 27 BulbNodes.   You can run this version of the clock component by clicking on the image that follows.


For our final example, in order to truly minimize node count, each digit is represented by a single ImageView node. When the value of a digit changes, a new Image is displayed.  By caching all of the possible digit values (blank, 0-9) you can very quickly switch images.  No doubt, prettier images can be created, but I think you get the point.  Click on the image that follows to try this version.


The Results

The slower the compute platform, the more pronounced the differences in performance should be.  Thinking along those lines, a very modest 1.4 GHz Pentium M laptop was chosen as the test environment to compare CPU utilization for these applications.  OpenSolaris provides an easy-to-use well-known command-line tool called vmstat(1M), which was chosen as the mechanism to analyze the individual applications. In contrast, the Performance Tab which is part of the Windows Task Manager, seemed to produce wild performance variations.

For each run,  the clocks were set to one minute, and run until the time expired.  The data shown below represents the average CPU utilization, after startup, for each of the three implementations.  In particular we'll look at the following fields returned by vmstat:

  • us - percentage usage of CPU time in user space
  • sy - percentage usage of CPU time in system space
  • id - percentage usage of CPU time idling
The sum of (us + sy + id) should approximate 100%.


Number of Nodes per Digit
CPU Utilization
Implementation 1: BulbClockNode
 27 BulbNodes
 us: 22%  sy: 2%  id: 76%
Implementation 2: LEDClockNode
 7 SegmentNodes
 us: 9%    sy: 2%  id: 89%
Implementation 3: ImageClockNode
 1 ImageNode
 us: 3%    sy: 2%  id: 95%


The JavaFX engineering team is well aware of this limitation, and hopes to redesign the underlying scenegraph plumbing in the future.  Regardless, it's still a good idea to take into consideration the size of your scenegraph.

JavaFX book status:  Our upcoming book, JavaFX: Developing Rich Internet Applications, is in copy edit.  Coming soon: Rough cuts of certain chapters will be available on Safari.

Tuesday Mar 24, 2009

Hooray! Acrobat Reader for [Open]Solaris x86

Today truly marks a milestone in the history of Solaris for the x86/x64 platform.  One of the most ubiquitous applications, Adobe Reader (notably Acrobat), is now available on [Open]Solaris for both Sparc and x86.

It's been a long time coming, one argument for the lack of Solaris x86 support till now was that [Open]Solaris didn't have the critical mass.  That's a hard one to swallow for the following reasons:

  • There's been a Solaris Sparc version for a long time.  Solaris x86 downloads outnumber Sparc downloads by a large factor.
  • Versions for AIX and HP-UX, which have a significantly smaller installed base, are available.  Maybe IBM and HP payed a lot of money for this?
  • Until recently there were versions even for the likes of OS/2 and UnixWare.

Perhaps open source alternatives are becoming good enough to pose a threat?  Whatever the reason, we welcome the arrival of Acrobat Reader and Adobe's change of heart.

You can get your copy here.

Happy Downloading!

Thursday Mar 19, 2009

Bindstorming

It is within our nature, even in the most infinitesimal way, to leave our mark on this world before we exit it.  I'd like to coin the following term, heretofore unseen in the JavaFX space, and submit it as my humble contribution to the human collective:

bindstorm \\'bïnd•storm\\ (noun): condition where a multitude of JavaFX bind recalculations severely hampers interactive performance

Yeah, I know, using the word you wish to define inside its definition is bad, but there is precedent for this: (1) Fancy-schmancy, hoity-toity college dictionaries do it all the time. (2) Mathematicians and computer scientists call this recursion: that mysterious concept which developers use to impress others of their programming prowess.

Don't get me wrong, JavaFX binding is incredibly powerful.  Heck, we dedicated a whole chapter to it in our soon-to-be-released book JavaFX: Developing Rich Internet Applications.  But binding does come with a price, and like most anything else, over-consumption can lead to abuse.

Consider this use case: you've got a JavaFX application with dozens or maybe even hundreds of Nodes that are part of the scenegraph.  Each of the Nodes are ultimately sized and positioned in proportion to height and width instance variables that are passed on down.  If you define width and height at startup and have no interest in a resizeable interface, then you stand a good chance of avoiding the use of many bind expressions.  The one potential twist here is that if you're sincerely interested in a non-resizeable application, but want it to consume the entire screen, what do you do?  As screens come in all shapes and sizes, you may not know what the resolution is at start time.  JavaFX has an elegant solution for this which uses binding.

Here's a simple application which defines a Rectangle and Circle that fill the entire screen.  You can click anywhere within the Circle to exit the application.  Notice the number of binds required to get this to work.

import javafx.stage.\*;
import javafx.scene.\*;
import javafx.scene.shape.\*;
import javafx.scene.paint.\*;
import javafx.scene.input.\*;

function run() : Void {
    var stage: Stage = Stage {
        fullScreen: true
        scene: Scene {
            content: [
                Rectangle {
                    width: bind stage.width
                    height: bind stage.height
                    fill: Color.BLUE
                }
                Circle {
                    centerX: bind stage.width / 2
                    centerY: bind stage.height / 2
                    radius: bind if (stage.width < stage.height) then
                            stage.width / 2 else stage.height / 2
                    fill: Color.RED
                    onMouseClicked: function(me: MouseEvent) {
                        FX.exit();
                    }
                }
            ]
        }
    }
}

Imagine what this would look like if you had lots of complex custom components with many more dependencies on height and width.  In addition to the potential performance impact, this could be error-prone and cumbersome to code.  To avoid the over usage of binding and the potential for a bindstorm, applications of this sort could be re-written as follows:

import javafx.stage.\*;
import javafx.scene.\*;
import javafx.scene.shape.\*;
import javafx.scene.paint.\*;
import javafx.scene.input.\*;

function run() : Void {
    var AWTtoolkit = java.awt.Toolkit.getDefaultToolkit ();
var screenSizeFromAWT = AWTtoolkit.getScreenSize (); Stage { fullScreen: true scene: Scene { content: [ Rectangle { width: screenSizeFromAWT.width height: screenSizeFromAWT.height fill: Color.BLUE } Circle { centerX: screenSizeFromAWT.width / 2 centerY: screenSizeFromAWT.height / 2 radius: if (screenSizeFromAWT.width <
screenSizeFromAWT.height) then
screenSizeFromAWT.width / 2
else screenSizeFromAWT.height / 2 fill: Color.RED onMouseClicked: function(me: MouseEvent) { FX.exit(); } } ] } } }

We achieve the same effect as the first example by first making a call to a method in the java.awt.Toolkit package.  With this information we can statically define our scenegraph without the use of binding.

There is one caveat to this solution.  As the AWT (Advanced Windowing Toolkit) is an integral part of Java SE, this code should be portable across all JavaFX desktops.  However, if you wish to deploy a JavaFX Mobile solution, the AWT calls would likely change.  Is there a mechanism that might work across both models?

As a final thought, while we're on this theme of coining terms, my compadres Jim Clarke and Eric Bruno, co-authors of the aforementioned JavaFX book, jokingly asked what word could be used to describe this scenario:

"Condition where binds lead to binds that leads back to the original bind, ending up in a Stack fault?"

BindQuake? BindTsunami? Bindless? BindSpin? BindHole (BlackHole)? BindPit?


        
    

Tuesday Feb 24, 2009

Registering Multiple Actions (or Handlers) in JavaFX

Java developers, especially those performing any type of GUI work, will ultimately encounter Java's event-driven programming paradigm.  In short, if programmers want to act upon some kind of event they bundle up a chunk of code into a Java method, typically referred to as a handler, and register the handler with that event.  Whenever that event occurs, the handler code will automatically be executed.

JavaFX provides a similar mechanism.  For a straightforward example, the code below defines a simple timer in JavaFX with a resolution of 1 second.  Each time a second expires, the function specified by the action instance variable will be executed.  Here's what it looks like:

import javafx.animation.\*;

public class SimpleTimer {
    public def timeline = Timeline {
        repeatCount: 5
        interpolate: false
        keyFrames: [
            KeyFrame {
                time: 1s
                action: function () : Void {
                    println("tick");
                }
            }
        ]
    }
}

Adding a run() function, as follows, to the bottom of this source will enable you run an instance of this timer:

function run() : Void {
    var s = SimpleTimer{};
    s.timeline.playFromStart();
} 

The output from this run looks like this:

tick
tick
tick
tict
tick 

It's all well and good if you only need a single action.  What if you wanted to perform multiple actions and/or dynamically add or subtract a number of actions?  We can enhance our previous SimpleTimer class to dynamically register and unregister handlers by taking advantage of two of JavaFX's features: sequences and function pointers.

Our new class provides more flexibility:

  • It defines an instance variable called duration, which enables the user to specify the resolution of a clock tick at object instantiation.
  • It defines two additional public functions called registerHandler() and unRegisterHandler() which take a function pointer (a handler) as an argument.  By registering a handler, the function will be included in the list of handlers to be executed each time the specified duration expires.
  • A handler is registered by inserting it's function pointer argument into an internal sequence of function pointers called handlers[].
  • A handler is similarly unregistered by deleting it's function pointer argument from the handlers[] sequence.
  • The action instance variable, which is part of the KeyFrame instance, now calls an internal function called runHandlers()runHandlers() sequentially executes the functions found in the handlers[] sequence.
Here's the new class:
import javafx.animation.\*;

public class Timer {
    /\*\*
     \* User-definable:  specifies the length of time for one cycle.
     \*/
    public var duration = 100ms;

    public def timeline = Timeline {
        repeatCount: Timeline.INDEFINITE
        interpolate: false
        keyFrames: [
            KeyFrame {
                time: duration
                action: runHandlers
            }
        ]
    }

    // Holds the list of handlers to run
    protected var handlers: function() [];

    /\*\*
     \* Add the function, represented by the handler argument, to the list
     \* of handlers.  These will run when the elapsed time, specified
     \* by the duration instance variable, expires.
     \*/
    public function registerHandler(handler : function()) : Void {
        for (func in handlers) {
            if (handler == func) {
                return;  // handler already registered -- skip
            }
        }
        insert handler into handlers;
    }

    /\*\*
     \* Remove the function, represented by the handler argument, from
     \* the list of handlers.
     \*/
    public function unRegisterHandler(handler : function()) : Void {
        delete handler from handlers;
    }

    protected function runHandlers() : Void {
        for (handler in handlers) {
            handler();
        }
    }
} 

To test this class out, we'll add a run() function at the script level.  The run() function creates a Timer instance and registers two handler functions, decrementTenthsRemaining() and processTicks().  Here's the code:

function run() : Void {
    var t = Timer {};
    var tenthsRemaining = 100;
    var decrementTenthsRemaining = function() : Void {
        tenthsRemaining -= 1;
    }
    var processTick = function() : Void {
        if (tenthsRemaining mod 10 == 0) {
            println("seconds left: {tenthsRemaining / 10}");
        }
        if (tenthsRemaining == 0) {
            t.timeline.stop();
        }
    };
    t.registerHandler(decrementTenthsRemaining);
    t.registerHandler(processTick);
    t.timeline.play();
}

And this is the output from the run:

seconds left: 9
seconds left: 8
seconds left: 7
seconds left: 6
seconds left: 5
seconds left: 4
seconds left: 3
seconds left: 2
seconds left: 1
seconds left: 0

Shameless Promotion:  Keep up to date with the latest status of our upcoming JavaFX Book entitled JavaFX: Developing Rich Internet Applications at jfxbook.com.


Tuesday Feb 17, 2009

JavaFX Book Coming to a Theatre Near You

Despite the considerable attention JavaFX has garnered, publications (i.e. books) that discuss JavaFX in any detail are few and far between, and anything that has been published, as good as it may have been, is unfortunately hopelessly out of date.  The reality is that up until recently, the JavaFX platform has been a moving target.  With the advent of JavaFX 1.1 however, the platform has stabilized to the point where you should begin to see legitimate books appearing on the subject.

Jim Clarke, Eric Bruno and I have been working steadfastly on a book entitled JavaFX: Developing Rich Internet Applications, which should be among the first -- if not the first -- of these new books.  From our standpoint the content is nearly finished.  What remains is the editing and publication process which takes a few additional months to complete.  Plans call for this book to be available in time for the JavaOne 2009 Conference, if not earlier.

We also plan on making rough cuts of certain chapters available on Safari.  As soon as these are ready, we'll let you know.  Finally, check out our website, jfxbook.com, dedicated to the book.  There you will find additional resources that accompany the book, including sample code and applications.  One such application is a JavaFX version of the popular Sudoku game, pictured below.

Visit jfxbook.com and give it a try.

Wednesday Jan 14, 2009

Overhead in Increasing the Solaris System Clock Rate

In a previous entry entitled Real-Time Java and High Resolution Timers, we discussed how Sun's Java Real-Time System requires access to timers with a resolution greater than the default 10ms to do anything really interesting.   It was also stated that most modern processors have an APIC or Advanced Programmable Interrupt Controller which supports much finer-grained clock tick rates.

Unfortunately there are many instances where a system does indeed contain an APIC, but it is not exposed by the BIOS.  Furthermore, we've found that some of the embedded, low-power x86-based processors do not contain an APIC at all.  For an example, take a look at the AMD Geode LX 800 based fit-PC Slim.

So if you wanted to utilize higher resolution timers for this class of system, you'd have to resort to alternative methods.  Solaris and OpenSolaris provide two /etc/system parameters called hires_tick and hires_hz to facilitate increasing your default system clock tick.  By adding the following line to /etc/system, you'll increase the system clock tick rate from the default of 100 per second to 1,000 per second, effectively changing the clock resolution from 10ms to 1ms.

   set hires_tick=1

If you want to further increase the clock resolution, you can do so via the hires_hz system tunable parameter.  Although formally unsupported, it does work.   In order to, for example, increase the clock tick rate to 10,000, add this to /etc/system:

    set hires_tick=1
    set hires_hz=10000

To achieve the desired effect above, you must include both the hires_tick assignment in addition to setting the hires_hz parameter.

These modifications do not come without side-effects, and depending upon the hardware in question and the granularity of the desired timer resolution, they could be significant.  In short, it takes additional CPU cycles to field all those timer interrupts.  So I thought it'd be interesting to see what effect changing the clock tick rate had on two separate systems.   Here they are:

 System  fit-PC Slim
 Panasonic Toughbook CF-30 (Revision F)
 CPU  AMD Geode LX 800 (500 Mhz)
 Intel Core 2 Duo L7500 1.60GHz
 OpenSolaris Version
 snv_98  snv_101b

The measuring tool used for this simple experiment is vmstat(1m).  Solaris aficionados will likely point out that there are much more accurate alternatives, but I think vmstat(1m) gives a decent feel for what's going on without having to expend a whole lot of extra energy.  In particular we'll look at the following fields returned by issuing a 'vmstat 5', and picking one of the interim samples:

  • in - interrupts per second
  • us - percentage usage of CPU time in user space
  • sy - percentage usage of CPU time in system space
  • id - percentage usage of CPU time idling

The sum of (us + sy + id) should approximate 100%.  The table below shows sample vmstat output on various clock tick settings for our two hardware platforms.

Clock tics/sec
 100
 1000  10000  100000
/etc/system settings
 none (default)
 set hires_tick=1
set hires_tick=1
set hires_hz=10000
set hires_tick=1
set hires_hz=100000
vmstat(5) sample fit-PC
 in: 201
 us: 0
 sy: 1
 id: 99
 in: 2001
 us: 0
 sy: 5
 id: 95
 in: 19831
 us: 0
 sy: 43
 id: 57 
n/a

vmstat(5) sample CF-30

 in: 471
 us: 0
 sy: 0
 id: 99
 in: 2278
 us: 0
 sy: 1
 id: 99
 in: 20299
 us: 0
 sy: 5
 id: 95
 in: 200307
 us: 0
 sy: 21
 id: 79

Notes/Conclusions:

  • The vmstat(5) was let run for about a minute.  The output above shows one of the typical 5 second samples.  The other 5 second samples are almost identical.
  • The user (us) CPU time numbers give us a reasonable idea that these systems were predominantly in an idle state when being sampled.
  • The number of interrupts serviced per second is directly proportional to the clock tick rate.
  • And of course, the larger the number of interrupts, the more system CPU time is required.
  • The amount of overhead taken up by increasing the clock rate is a function of system capability.  The CF-30 not only has a much faster processor, it also has two cores to share the interrupt load.  As such it could accommodate a much higher clock tick rate.  For the fit-PC, performance is impacted profoundly even at modest clock tick rates.

Thursday Dec 04, 2008

Why JavaFX is Relevant

This week marks the formal release of JavaFX 1.0.  During the interval between the early marketing blitz and now, we've heard a lot from our friends in the press and the blogosphere, and in many instances what they had to say was not very pretty.  Some think the Rich Internet Application platform battle lines are already drawn between Adobe and Microsoft, and dismiss Sun as having arrived too late to the party.  Others opine that JavaFX's underlying Java platform is so yesterday.  In fact Java is the primary reason why JavaFX will, much to the chagrin of many, receive serious consideration.  Here's why:

  • Java is ubiquitous.  It is the proven, de-facto platform for web-based deployment.  On the desktop, it is estimated that approximately 90% of PCs have Java installed. In fact the major PC OEMs have seen fit to install it for you out of the box.  In the mobile world, Java is the dominant deployment platform.  Billions (that's with a 'b') of devices run Java.
  • The Java development community is arguably the largest on the planet.  Java gained initial widespread acclaim as a productive development environment, and continues to do so.  As JavaFX is an evolution of Java and seamlessly integrates with it, human nature tells us that individuals will naturally want to work with and leverage that which they already know and are familiar with.
  • Alternatives are still no match for the Java Virtual Machine.  It has been extensively studied, vetted, scrutinized, poked, prodded, abused, cloned, and optimized more than any other virtual machine in the history of computing. And just in case you're under the impression that the Java Virtual Machine is limited only to the Java (and now JavaFX script) programming languages, think again.  At last count there were over 200 projects integrating modern dynamic languages to the Java VM.  That list includes the usual suspects like PHP, Ruby, JavaScript, Python, and [insert your favorite language here].
  • The amount of Java Standard Edition online updates is staggering.  We know.  We supply the downloads.  And once a desktop is upgraded, it will be able to take full advantage of the features JavaFX brings to the table, effectively trivializing the barriers to entry.
Many of our most talented folks have been working feverishly to reach this milestone.  That being said, there's still lot's more work to do.  But we're off to a real nice start.  Check out http://javafx.com.  Hmm.  looks like the site is a little sluggish right now.  Maybe we underestimated all the interest?

Wednesday Nov 05, 2008

OpenSolaris on the Smallest System Yet?

One of my compadres forwarded me this link from CompuLab, an Israeli company which has created this real small, very energy efficient PC.  It may just be the smallest full-featured system to date. 

I'm a sucker for these types of things, and thought it would be interesting to get a reduced footprint version of OpenSolaris up and running on it.  Haven't gotten around to playing with wi-fi, or for that matter the graphics (as the reduced footprint version of OpenSolaris has no windowing system), but every indication points to a system that doesn't require any magic incantations to get OpenSolaris up and running.  Here are some of the specs

CPU: AMD Geode LX800 500MHz     
Chipset: AMD CS5536
Display: Integrated Geode LX display controller up to 1920x1440  
Memory: 512MB DDR 333MHz soldered on-board
Hard disk: 2.5” IDE 60GB
Ports:
 RJ45 Ethernet port 100Mbps
 WLAN 802.11b/g 54Mbps
 3 x USB 2.0 HiSpeed 480Mbps
 mini RS-232 (cable available from CompuLab)
 VGA DB15
 Stereo line-out audio (headphones)
 Stereo line-in audio / Mic


The system has a serial port, and upon request, CompuLab will provide the required custom serial cable.  By instructing GRUB to redirect the console to the serial port, you can connect the serial cable to another system and communicate directly with the console.  To get a rough idea of how to accomplish this, check here.  The screenshots below show that from an OpenSolaris terminal window, you can use the tip(1) command to accomplish this.

So the question is, how minimal is this configuration?  Issuing `ps -ef' shows that only 14 processes are currently running, of which I'm sure there's an opportunity to eliminate one or two if need be.

To give you an idea of how much RAM is currently consumed, here's what issuing the 'memstat' macro under mdb(1m) yields:

Who says Solaris can't fit in small places?

Wednesday Jul 23, 2008

Fast Booting Solaris

A veteran Java ONE keynote presenter, Perrone Robotics has developed some real interesting technologies which take the concept of using autonomous (i.e. unmanned) vehicles to a whole new level.  One of their key ingredients is the MAX software platform which utilizes common commercially available components to enable Perrone to very quickly and cost-effectively retrofit nearly any vehicle in short order.


The MAX robotics platform runs on a (roughly 4" x 6") low-power PC board atop Solaris and Sun's Java Real-Time System (Java RTS).  This combination gives Perrone the ability to leverage the huge Java development community, and assures that their critical software components behave in a predictable and deterministic fashion.

During the Java ONE 2007 conference, I was speaking with Paul Perrone about the notion of creating a minimized version of Solaris over which his platform might run.  The helicopter pictured above, boots from a relatively small (4-8Gb)  IDE flash drive, where standard Solaris takes up a substantial chunk.  It leaves them precious little space to collect valuable information like telemetry or terrain data.  Paul asked to revist this idea for a future project.  That's where we left off.

Not that we've ignored them since, but it wasn't until a year later that small Solaris reared its head again.  In addition to saving space, their main interest in this environment was in seeing how much faster Solaris might boot up.  The ability to be fully functional from power-up in as short a time as possible is of critical importance.

So before investigating what advantages there might be, let's provide some background information:

Hardware

Two separate systems were used, and for argument's sake, represent two ends of the x86 compute spectrum. 


Embedded Profile
Modern Profile 
System iGologic i84810
Panasonic Toughbook CF-30 (Rev. F)
CPU 1GHz Celeron M
Core 2 Duo L7500 1.60GHz
RAM 512MB 1GB
Disk 4GB Flash IDE Drive
Solid State SATA Drive

Operating Systems

Minimal configurations were created for Solaris 10 08/07 (Update 4) and OpenSolaris Nevada build 93.  These configurations boot entirely into RAM and consume less than 100MB ramdisk space.  With a little more effort they can be may significantly smaller.  The original blog post describing the environment is here.   You can download the framework for these hardware/OS combinations here, and can get a feel for the build environment by taking a look at this README.

Definitions

Within the context of this discussion, here are the key terms along with their meanings.

Total Boot Time: This is the time it takes from power-up till a user is prompted to log in.  Typically for a full Solaris installation, the windowing system must first start up and present the user with a login screen.  In a minimal Solaris environment, there is no windowing system.  Instead, the total boot time is defined as the time it takes from power-up till a user is prompted with the console login: prompt.

POST Time: POST or Power On Self Test is the time taken by the system at pre-boot to handle things like diagnostics,  BIOS and device initialization.  For this discussion, we'll define POST time as the time it takes from power-up until the user is presented with a GRUB boot menu.  We call out this segment of the total time because in many cases we are at the mercy of the PC/BIOS manufacturer and can't overly influence how quickly or slowly this proceeds.

Solaris Boot Time: The time it takes from being presented with a GRUB boot menu till a Solaris user is prompted to log in.  Again, depending upon whether a windowing system is configured or not, this may represent the time it takes to be prompted with a login screen or console login: prompt respectively.  This represents the segment of time that we can influence.

We can conclude from these definitions that:

   Total Boot Time = POST Time + Solaris Boot Time

Results

Embedded Profile: iGoLogic i84810 system 

OS
Post Time
Solaris Boot Time
Total Boot Time
 Solaris 10 08/07
13 sec
26 sec
39 sec
 OpenSolaris Nevada Build 93
13 sec
18 sec
31 sec 

Modern Profile: Panasonic Toughbook CF-30

OS POST Time
Solaris Boot Time
 Total Boot Time
 Solaris 10 08/07
 6 sec
 18 sec
 24 sec
 OpenSolaris Nevada Build 93
 6 sec
 9 sec
 15 sec

Conclusions/Notes

1. These times were taken by hand with the stopwatch feature on my Timex.  If anything, the times might be slightly shorter than actually recorded as there is a human delay in reacting to seeing the necessary prompts.  I ran the tests several times, and the same numbers consistently appear.

2. The version of the OS appears to matter a lot, as OpenSolaris nvx_93 boots up nearly twice as fast as Solaris 10 U4 on the same hardware.

3. The type of I/O device subsystem seems to be a big factor too.  For example, by switching out the IDE Flash Drive with a 5400 RPM IDE hard disk, i84810 total boot time decreased by about 6 seconds for both Solaris 10 and OpenSolaris. 

4. The minimal Solaris environment is currently only available in 32-bit mode.

5. With relative ease, Solaris can be configured to boot in less that 10 seconds on modern x86 hardware.  My unofficial record stands at 9 seconds (or slightly less).   No doubt it could boot faster on more robust hardware (eliminating POST time).  Any takers?

About

Jim Connors

Search

Categories
Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today