Dr. Dobb's is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Channels ▼


Actor Virtual Machine

Rex Young can be reached at [email protected] His blog is at http://www.java.net/blogs/rexyoung/.

Object-oriented programming has been at the center of mainstream software development since the early 1990s. Although OOP languages have evolved, the OO model foundation hasn't changed much. OOP is still deeply rooted in sequential programming within a single address space in the following ways:

  • Objects are related and linked by references. This tightly coupled relationship is based on the assumption that all objects are in the same address space.
  • Objects interact with each other by synchronous method invocations. The caller's thread of control always steps inside the callee. Although the OO model says that an object encapsulates data and behaviors, the object does not have any control over who would execute a behavior and when. Thus objects actually encapsulate the definition of the behaviors rather than behaviors themselves. This makes an object just a stop for a thread to reach into, do something, then get out of in a sequential route through which the thread executes.
  • OOP uses threads as its concurrent model. Multiple threads are used to execute the different parts of an OO program concurrently. Such a program may be abstracted as multiple sequential thread routes over collections of objects. This concurrent model adds extra semantics for a programmer to handle. On the one hand, at design level this model adds a new dependence of the design of thread routes to objects besides their own data and behaviors. On the other hand, at programming level it adds handling interleavings to the programming of the objects when they sit on a route where multiple threads cross. This kind of programming in non-trivial programs is incomprehensible to humans and cannot and should not be trusted, as Edward Lee concluded [Lee 2006].
  • The OO model suggests modeling live, real-world entities using active objects [Booch 2007]. In this approach an active object is defined as an object with its own thread of control. Unfortunately, active objects do not change the abstraction that programs may be viewed as multiple, sequential thread routes over collections of objects. Non-active objects remain passive, and even active objects are forced to be passive when they are on the thread routes of other active objects.

Currently, several forces are pushing OOP and its foundations to overcome their limitations and be ready for highly concurrent and distributed systems. These forces include evolving hardware (e.g., multicore processors), increasing complexity of programming languages, and increasing level of expectation of new and growable software [Grogono 2008].

Despite the limitation, the OO model has proven to be a unifying concept in computer science [Booch 2007]. For this reason, I decided to leave the OO model untouched for computations (at which it is good) and add a component-based coordination model called the "Actor Virtual Machine" (AVM) as a new extension. A coordination model is the glue that binds separate computations into an ensemble [Gelernter 1992].

The AVM model brings to the OO model message-passing concurrency, which is the basic framework for multi-agent systems, the natural style for a distributed systems, and lends itself well to building reliable systems [Van Roy 2004]. Moreover, AVM customizes the message passing with a dynamic message-binding mechanism that encourages human interventions (e.g., hot code swapping, adjustable workflow) via visual syntax and tools at run time.

The AVM Model

The core idea behind the AVM model is to introduce an object, called the "AVM object", that forwards messages for objects. Sender objects synchronously invoke methods on the AVM object. The AVM object also synchronously invokes methods on receiver objects, but using its own threads. As a result, sender objects' thread of control never step into the inside of receiver objects. Thus asynchronous message passing is achieved between senders and receivers.

Sender and receiver objects are called actors for the sole reason that they are aware of an AVM object and use it for sending and/or receiving messages. Objects that work with an AVM object other than sending and/or receiving messages are not called actors. The name of actor is borrowed from the Actor Model, but AVM actors do not comply with the Actor Model. For example, AVM actors may carry out computations without receiving an incoming message, are not required to support become command for change of behaviors, etc. [Agha 1986].

The AVM model employs several techniques and elements, from basic to advanced. Code fragments use Java-like syntax for demonstration only. These techniques and elements are designed to be used individually as a utility toolbox or combined as a coordination language, which is based on message-passing concurrency, up to user's choices, where applicable.

Asynchronous Message Passing

Actors interact with each other by asynchronous message passing, which effectively breaks the abstraction that OO programs may be viewed as multiple sequential-thread routes over collections of objects, as asynchronous interaction separates the thread of control among actors.

class AVMObject
  void send(AVMCallback receiver, Object message);
interface AVMCallback
  void onMessage(Object message);

The AVM model introduces an AVM object between a sender and a receiver. First, the sender invokes the AVMObject.send() method, and AVMObject immediately lets the sender's thread return. Then the AVMObject uses its own thread to invoke the receiver. onMessage() method. As a result, the sender successfully sends an asynchronous message to the receiver.

The AVM model requires that messages between a certain pair of sender and receiver arrive at the receiver's callback method in the order they are received by the AVMObject. A sender is required to know the references of the AVMObject and a receiver. The receiver is required to implement the AVMCallback.onMessage().

Related Reading

More Insights

Currently we allow the following HTML tags in comments:

Single tags

These tags can be used alone and don't need an ending tag.

<br> Defines a single line break

<hr> Defines a horizontal line

Matching tags

These require an ending tag - e.g. <i>italic text</i>

<a> Defines an anchor

<b> Defines bold text

<big> Defines big text

<blockquote> Defines a long quotation

<caption> Defines a table caption

<cite> Defines a citation

<code> Defines computer code text

<em> Defines emphasized text

<fieldset> Defines a border around elements in a form

<h1> This is heading 1

<h2> This is heading 2

<h3> This is heading 3

<h4> This is heading 4

<h5> This is heading 5

<h6> This is heading 6

<i> Defines italic text

<p> Defines a paragraph

<pre> Defines preformatted text

<q> Defines a short quotation

<samp> Defines sample computer code text

<small> Defines small text

<span> Defines a section in a document

<s> Defines strikethrough text

<strike> Defines strikethrough text

<strong> Defines strong text

<sub> Defines subscripted text

<sup> Defines superscripted text

<u> Defines underlined text

Dr. Dobb's encourages readers to engage in spirited, healthy debate, including taking us to task. However, Dr. Dobb's moderates all comments posted to our site, and reserves the right to modify or remove any content that it determines to be derogatory, offensive, inflammatory, vulgar, irrelevant/off-topic, racist or obvious marketing or spam. Dr. Dobb's further reserves the right to disable the profile of any commenter participating in said activities.

Disqus Tips To upload an avatar photo, first complete your Disqus profile. | View the list of supported HTML tags you can use to style comments. | Please read our commenting policy.