JavaOne 2014: Java vs JavaScript

© 2014 IBM Corporation

Chris Bailey – IBM Runtime Monitoring and Diagnostics

1st October 2014

Java vs JavaScriptHead to Head

Document number


Important Disclaimers

THE INFORMATION CONTAINED IN THIS PRESENTATION IS PROVIDED FOR INFORMATIONAL PURPOSES ONLY.

WHILST EFFORTS WERE MADE TO VERIFY THE COMPLETENESS AND ACCURACY OF THE INFORMATION CONTAINED IN THIS PRESENTATION, IT IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED.

ALL PERFORMANCE DATA INCLUDED IN THIS PRESENTATION HAVE BEEN GATHERED IN A CONTROLLED ENVIRONMENT. YOUR OWN TEST RESULTS MAY VARY BASED ON HARDWARE, SOFTWARE OR INFRASTRUCTURE DIFFERENCES.

ALL DATA INCLUDED IN THIS PRESENTATION ARE MEANT TO BE USED ONLY AS A GUIDE.

IN ADDITION, THE INFORMATION CONTAINED IN THIS PRESENTATION IS BASED ON IBM’S CURRENT PRODUCT PLANS AND STRATEGY, WHICH ARE SUBJECT TO CHANGE BY IBM, WITHOUT NOTICE.

IBM AND ITS AFFILIATED COMPANIES SHALL NOT BE RESPONSIBLE FOR ANY DAMAGES ARISING OUT OF THE USE OF, OR OTHERWISE RELATED TO, THIS PRESENTATION OR ANY OTHER DOCUMENTATION.

NOTHING CONTAINED IN THIS PRESENTATION IS INTENDED TO, OR SHALL HAVE THE EFFECT OF:

- CREATING ANY WARRANT OR REPRESENTATION FROM IBM, ITS AFFILIATED COMPANIES OR ITS OR THEIR SUPPLIERS AND/OR LICENSORS

2

© 2014 IBM Corporation3

A Quick Survey


What languages do you program in?

Java JavaScript Both0

10

20

30

40

50

60

Pe

rce

nta

ge

of A

ud

ien

ce


What runtimes do you use?

Server Java Applets JS in Browser Node.js Rhino Nashorn Avatar.js0

10

20

30

40

50

60

70

80

90

100

Pe

rce

nta

ge

of A

ud

ien

ce


Introduction to the speaker Chris Bailey

IBM Runtime Monitoring and Diagnostics Architect- 14 years working with Java and JVM technologies- 6 months working with Node.js and V8

Recent work focus:- Java monitoring, diagnostics and troubleshooting- Java integration into the cloud- JavaScript monitoring, diagnostics and troubleshooting

My contact information:- [email protected] http://www.linkedin.com/in/chrisbaileyibm- http://www.slideshare.net/cnbailey/- @Chris__Bailey

mailto:[email protected]

http://www.linkedin.com/in/chrisbaileyibm

http://www.slideshare.net/cnbailey/


Goals of this talk

● Langage Adoption

● Deployment Modes

● Introduction to Node.js

● Asynchronous IO

● WebApplication Performance

● Under the Hood of Node.js

● JavaScript on the JVM


Language Adoption


GitHub Adoption: Java


GitHub Adoption: JavaScript

GitHut Adoption


Modulecounts.com


StackOverFlow Survey


Tiobe Programming Community Index

● Ratings based on the number of skilled engineers, courses and third party vendors.


Indeed.com Job Trends


Language Adoption JavaScript has a large developer base

- #1 on GitHub with 48.8% more active repositories than Java- #1 on modulecounts.com with 10% more NPM modules than Maven - #1 used language by StackOverflow survey responders- #9 language on the Tiobe index

Java remains a hugely relevant language, particularly on the server - #2 on GitHub with xx% more than the next language- #3 on modulecounts.com with 73.8% more modules than the next language- #2 language on the Tiobe index- #1 on indeed.com for developer jobs


Deployment Modes


Usage in the Browser● JavaScript is ubiquitous in the browser

- Supported in every browser- Full integration with HTML and CSS

● Not affected by negative publicity....

Unless it is absolutely necessary to run Java in web browsers, disable it as describedbelow, even after updating to 7u11. This will help mitigate other Java vulnerabilities thatmay be discovered in the future.

This and previous Java vulnerabilities have been widely targeted by attackers, and new Java vulnerabilities are likely to be discovered. To defend against this and future Java vulnerabilities, consider disabling Java in web browsers…


Usage on the Server● Java has a long history on the server

- JPE launched in 1998

● Java has rich platform support:- Linux x86, Linux POWER, zLinux- Windows, Mac OS, Solaris, AIX, z/OS

JavaScript is a nascent language on the server- Limited platform support – although its growing- No API support to interact with the OS

Part of the browser security model- Frameworks like Node.js have changed that.


Introduction to Node.js Single Threaded Event based JavaScript framework

– Uses non-blocking asynchronous I/O

Wraps the Chrome V8 JavaScript engine with I/O interfaces– Libuv provides interaction with OS/system

● Designed to build scalable network applications– Suited for real time delivery of data to distributed client

Available on a growing set of platforms- Windows, Linux x86, Linux ARM, Mac OS, Solaris- Linux POWER, zLinux, AIX

libuvV8

Node Bindings

Node Standard Library

C

JavaScript


Async I/O Model


Typical approach to IO One thread (or process) per connection

- Each thread waits on a response- Scalability determined by the number of threads

Each thread:- consumes memory- is relatively idle

Number of concurrent customers determined by number of depot workers

Additional customers wait in a queue with no response


Asynchronous non-blocking IO One thread multiplexes for multiple requests

- No waiting for a response- Handles return from I/O when notified

Scalability determined by:- CPU usage- “Back end” responsiveness

Number of concurrent customers determined by how fast the food Server can work

Or until the kitchen gets slammed


Drawbacks of Asynchronous IO Tasks must execute quickly to avoid blocking the event queue

- Analogous to work done under a lock- Stick to the right jobs, eg, I/O- Delegate CPU bound tasks to back end processes

Easy to run out of memory- No direct bound on amount of parallel work- Holding state for each piece or work means unbounded memory usage


JavaScript and Asynchronous IO JavaScript is already event based in the browser

- eg. onClick and onMouseOver events

First class functions and closures fit well with events- Easy to create and pass function callbacks- Easy to execute callbacks in the context of the event

Node.js execution is based on an event loop- Asynchronous I/O built in from the ground up

Node.js execution uses a single thread- No need to worry about locking or shared data- Most machines are now multi-CPU, so cluster capabilities are provided


HTTP Server Examplevar cluster = require('cluster');var cpus = require('os').cpus().length;var http = require('http');

if (cluster.isMaster) { for (var i = 0; i < cpus; i++) { cluster.fork(); } cluster.on('death', function(worker) { console.log("Worker" + worker.pid + "died"); });} else {

http.createServer(function(request, response) { response.writeHead(200, {"Content-Type": "text/plain"}); response.write("Hello World!\n"); response.end();

}).listen(8080);}


HTTP Server Example with Clusteringvar cluster = require('cluster');var cpus = require('os').cpus().length;var http = require('http');

if (cluster.isMaster) { for (var i = 0; i < cpus; i++) { cluster.fork(); } cluster.on('death', function(worker) { console.log("Worker" + worker.pid + "died"); });} else {

http.createServer(function(request, response) { response.writeHead(200, {"Content-Type": "text/plain"}); response.write("Hello World!\n"); response.end();

}).listen(8080);}


Event Loop Example Hypothetical implementation of readFile():

function readFile(filename, cb) { nb_open(filename, function(err, fd) { if (err) cb(err, null); nb_read(fd, function(err, data) { cb(err, data); nb_close(fd); }); });}


JavaScript and Asynchronous I/O Very little time spent with events on the Event Loop

Provides good scalability, so should provide great performance for IO bound apps

Like WebApplications...


WebApp Performance


JSON Serialization JSON serialization of a newly

instantiated object

Maps- Key of message- Value of Hello, World!

Example response:

Results from TechEmpower.com Round 9 tests (2014-05-01)


JSON Serialization JSON serialization of a newly

instantiated object

Maps- Key of message- Value of Hello, World!

Example response:



Single Query Fetches single row from simple

database table

Row serialized as JSON

Example response:



Multiple Query Fetches multiple rows from a

simple database table

Rows serialized as JSON

Example response:



Multiple Query Fetches multiple rows from a


Rows serialized as JSON

Example response:



Data Updates Fetches multiple rows from a


Converts rows to objects and modifies one attribute of each object

Updates each associated row and serializes as JSON

Example Response:

Results taken from TechEmpower.com


Under the Hood


Polymorphism and Object Representation

Class pointer

Locks

Flags

int

Class pointer

Locks

Flags

int

Class pointer

Locks

Flags

int

Class pointer

Locks

Flags

int

Integer Object Name: Integer

Method Table

Super Class

Field Table

Constant Pool

Object Offsets

...

Integer Class

Java objects are fixed in size and shape- Determined by the object type- Type determined by class pointer

Class pointer gives access to Method Table- All methods executable from the object

Class pointer also gives access to Object Offsets- Fields in the object that are objects- Used for graph traversal during GC


Polymorphism and Object Representation

Field: offset

Field: offset

Field: offset

Map

1:

Length

2:

JSObject

JavaScript objects polymorphic and can change in size and shape

“Simple” objects:- Created with 32 fields- Contains a Map describing key : value

parings- Contains “extra properties” for overflow- Contains “elements” for array values- All fields are 64bits

Multiple maps can exist for “equal” objects- Dependent on order in which fields are added

to the object- ie. X then Y, or Y then X

Map

Elements

Extra Props

1:

2:

3:

...

...

32:

Map

1:

Length

2:

Fixed Array

Fixed Array

Map


Polymorphism and JIT Compilation Functions are stored in JavaScript objects as fields- No fixed set of methods for an object

● Objects are not typed, so data much be checked to determine how to handle iteg. the '+' operator:

- number + number → addition- string involved? → concatenation- objects involved? → convert to primitives then addition or concatenation

eg. property load:- Load prototype object- Load getter method- Load callback function

● Therefore not possible to determine what instructions to use just from the source code


Approach to JIT Compilation V8 JavaScript engine has two compilers

Full Compiler- Compiles JavaScript to native code on first execution- No bytecode stage, and no real optimization applied- Implements a Polymorphic Inline Cache (PIC)

Crankshaft- Execution counter based compilation strategy

Also supports on-stack replacement for code with hot loops- Compilation is done in the execution path- Optimized for types present in the PIC


Polymorphism and Garbage Collection

Implements Generational Garbage Collection, similar to the JVM

– New space, Old space, Large Object space, Remembered sets

– Additional JavaScript specific areas: map space, cell space, property cell space

– Implements concurrent mark/sweep capabilities

Root scan based on Global objects and local variables on the stack

– Reference tree built by following object references to further objects

– However... how do we know which variables are Objects rather than data?

• Solution is tagged pointers – use of 01 in the low order bits to denote a pointer


JavaScript on the JVM?


JavaScript on the JVM Rhino was bundled in JDK6

- Based on the Mozilla Rhino engine- Allowed JavaScript to be embedded in Java, and for JavaScript to call out to Java- Provided jrunscript command line utility

Nashorn replaces JavaScript in JDK8- Supports full ECMAScript 5.1 specification- Exploits invokedynamic to provide 2x to 10x better performance than Rhino- Provides jjs command line utility


JavaScript on the JVM

Rhino was bundled in JDK6ScriptEngineManager scriptEngineManager = new ScriptEngineManager();ScriptEngine nashorn = scriptEngineManager.getEngineByName("nashorn");

int sendVal = 7;nashorn.put("sendVal", sendVal);

nashorn.eval(" "+"var Thread = Java.type(\"java.lang.Thread\"); "+"var MyThread = Java.extend(Thread, { "+" run: function() { "+" print(\"Run in separate thread\"); "+" } "+"}); "+"var th = new MyThread(); "+ "th.start(); "+ "th.join(); "+"var resultVal = sendVal + 3;");

System.out.println(nashorn.get("resultVal");


ScriptEngineManager scriptEngineManager = new ScriptEngineManager();ScriptEngine nashorn = scriptEngineManager.getEngineByName("nashorn");






Introduction to Avatar.js Support for Node.js on Nashorn

Binary builds available from Maven– Avatar-js.jar– Avatar-js library (64bit)

Number of common NPM modules are supported

NPM required to dependencies

Issues for native NPMs– No native V8 APIs

From https://avatar-js.java.net/


Avatar.js vs Node.js Running Octane r33

– The “SpecJVM98” of the JavaScript world– Actually a JavaScript benchmark rather than a Node.js benchmark– So more comparing Nashorn to V8

Run on a 8 CPU Windows using:– Node.js v0.10.31– HotSpot 8u20

Settings are “out of the box” with no attempt to tune


Avatar.js running Octane Bencmark

Avatar.js

● Duration: 1m 56s

● Peak memory: 2830MB


Node.js running Octane Bencmark

Node.js

● Duration: 24s



Avatar.js vs Node.js Memory Usage

Avatar.js

● Duration: 1m 56s


Node.js

● Duration: 24s


Node.js is 4.8x faster

Avatar.js is >10x bigger


Avatar.js vs Node.js: Garbage Collection Avatar.js peak is:

– 920MB / 1.85GB

Node.js peak is:– 200MB / 220MB

Additional 700+MB of Object Heap

Additional 1GB of non-Object Heap


Avatar.js vs Node.js: Stack Traces

Interpretation of JavaScript at the Java layer means stacks are large and anonymous


Summary JavaScript has a large amount of interest and is growing

– Fits well for web applications and code sharing between server and browser– Async IO and event loop model makes it easy to write scalable applications

Weak typing and dynamic linkage makes programming easier, but introduces challenges– Errors found during compilation for strongly typed languages are found at runtime– JIT compilation loves certainty, which is removed

Nashorn and Avatar.js introduces JavaScript to the JVM– Adds interoperability– … but suffers from the same challenges


IBM Developer Kits for Javaibm.biz/javasdk

WebShere Liberty Profilewasdev.net

IBM Bluemixibm.com/bluemix

IBM Developer Kits for Node.jsibm.biz/nodesdk

http://ibm.biz/javasdk

http://wasdev.net/

http://ibm.com/bluemix

http://ibm.biz/nodesdk


Questions?


ibm.biz/javaone2014Visit Booth 5511 to learn about Cloud, DevOps and Mobile solutions

www.ibm.com/developer

Discovernew technical resources.

Developyour coding skills.

Connectwith developers.

http://ibm.biz/javaone2014


Copyright and Trademarks

© IBM Corporation 2013. All Rights Reserved.

IBM, the IBM logo, and ibm.com are trademarks or registered trademarks of International Business Machines Corp., and registered in many jurisdictions worldwide.

Other product and service names might be trademarks of IBM or other companies.

A current list of IBM trademarks is available on the Web – see the IBM “Copyright and trademark information” page at URL: www.ibm.com/legal/copytrade.shtml

http://www.ibm.com/legal/copytrade.shtml

JavaOne 2014: Java vs JavaScript

Software

Transcript of JavaOne 2014: Java vs JavaScript