/ 
Zing-0.20
River stage one • 7 direct dependents • 7 total dependents
/ Zing

NAME

Zing - Multi-Process Management System

ABSTRACT

Actor Toolkit and Multi-Process Management System

SYNOPSIS

use Zing;

my $zing = Zing->new(scheme => ['MyApp', [], 1]);

# $zing->execute;

DESCRIPTION

This distribution includes an actor-model architecture toolkit and multi-process management system which provides primatives for building resilient, reactive, concurrent, distributed message-driven applications in Perl 5. If you're unfamiliar with this architectural pattern, learn more about "the actor model".

INHERITS

This package inherits behaviors from:

Zing::Kernel

LIBRARIES

This package uses type constraints from:

Zing::Types

ATTRIBUTES

This package has the following attributes:

scheme

scheme(Scheme)

This attribute is read-only, accepts (Scheme) values, and is required.

METHODS

This package implements the following methods:

start

start() : Kernel

The start method prepares the Zing::Kernel and executes its event-loop.

start example #1
# given: synopsis

$zing->start;

PRIMATIVES

This distribution provides a collection of actor-model primitives which can be used to create sophisticated and powerful distributed systems, organized within intricate process topologies. The following is a directory of those primitives listed by classification:

messaging

These classes facilitate message-passing and communications:

processes

These base classes implement the underlying process (actor) logic:

stores

These classes handle data persistence for all messaging abstractions:

ready-made

These classes are ready-made process implementations using callbacks:

FEATURES

All features are implemented using classes and objects. The following is a list of features currently enabled by this toolkit:

actor-model

use Zing::Process;

my $p1 = Zing::Process->new;
my $p2 = Zing::Process->new;

$p1->send($p2, { action => 'greetings' });

say $p2->recv->{action}; # got it?

This distribution provides a toolkit for creating processes (actors) which can be run in isolation and which communicate with other processes through message-passing.

asynchronous

# in process (1)
use Zing::Domain;
use Zing::Process;

my $d1 = Zing::Domain->new(name => 'peers');
my $p1 = Zing::Process->new(name => 'p1');

$d1->push('mailboxes', $p1->mailbox->term);
$p1->execute;

# in process (2)
use Zing::Domain;
use Zing::Process;

my $d2 = Zing::Domain->new(name => 'peers');
my $p2 = Zing::Process->new(name => 'p2');

my $mailboxes = $d2->get('mailboxes');

for my $address (@$mailboxes) {
  # send each registered process a message
  $p2->send($address, { discovery => $p2->mailbox->term });
}

$p2->execute;

This distribution provides a multi-process management system which allows processes to be deployed and managed separately having the ability to communicate across threads of execution.

atomicity

# in process (1)
use Zing::KeyVal;

my $i = 0;
my $kv = Zing::KeyVal->new(name => 'stash');

while ($i < 1_000) {
  $kv->send('random', { value => 1 });

  # my $data = $kv->recv('random');
}

# in process (2)
use Zing::KeyVal;

my $i = 0;
my $kv = Zing::KeyVal->new(name => 'stash');

while ($i < 1_000) {
  $kv->send('random', { value => 2 });

  # my $data = $kv->recv('random');
}

This distribution provides data storage abstractions which perform atomic reads and write by leveraging Redis as the default data storage backend.

chainable

use Zing::Process;
use Zing::Ring;

my $p1 = Zing::Process->new(name => 'p1');
my $p2 = Zing::Process->new(name => 'p2');

my $ring = Zing::Ring->new(processes => [$p1, $p2]);

$ring->execute;

This distribution provides a mechanism for chaining (i.e. joining) two or more processes together and executing them in a turn-based manner. This ability allows you to design complex hierarchical process topologies.

channels

# in process (1)
use Zing::Channel;

my $chan = Zing::Channel->new(name => 'chat');

while (1) {
  if (my $data = $chan->recv) {
    # broadcast received
    warn $data->{text};
  }
}

# in process (2)
use Zing::Channel;

my $chan = Zing::Channel->new(name => 'chat');

while (1) {
  if (my $data = $chan->recv) {
    # broadcast received
    warn $data->{text};
  }
}

This distribution provides the means for braodcasting and communicating to multiple processes simulaneously through channels which are similar to FIFO queues.

clusterable

# in process (1) on cluster (1)
use Zing::Queue;

my $queue = Zing::Queue->new(name => 'tasks', target => 'global');

# pull from global queue
$queue->recv;

# in process (1) on cluster (2)
use Zing::Queue;

my $queue = Zing::Queue->new(name => 'tasks', target => 'global');

# pull from global queue
$queue->recv;

This distribution provides support cross-cluster communication and thus operations. Using federated Redis as the data storage backend means you can scale your deployments without changing your implementations.

configuration

# configure the namespace
ZING_NS=app

# enable process debugging tracing
ZING_DEBUG=1

# configure the namespace, same as ZING_NS (defaults to "main")
ZING_HANDLE=app

# configure where the command-line tool finds catridges
ZING_HOME=/tmp

# configure the hostname used in process registration
ZING_HOST=0.0.0.0
ZING_HOST=68.80.90.100

# configure the resource target (e.g. when distributing across multiple hosts)
ZING_TARGET='global' # 'us-east', 'us-west', etc

# configure the datastore (defaults to 'Zing::Store::Redis')
ZING_STORE='Zing::Store::Redis'

# configure Redis driver without touching your source code
ZING_REDIS='server=127.0.0.1:6379'
ZING_REDIS='every=1_000_000,reconnect=60'
ZING_REDIS='sentinels=127.0.0.1:12345|127.0.0.1:23456,sentinels_cnx_timeout=0.1'
ZING_REDIS='server=192.168.0.1:6379,debug=0'

# configure the object encoder (defaults to 'Zing::Encoder::Json')
ZING_ENCODER='Zing::Encoder::Json'

# configure where the command-line tool finds catridges and PID files
ZING_APPDIR=./
ZING_PIDDIR=/tmp

This distribution provides environment variables that let you customize how Zing operates and behaves without the need to modify source code. These attributes are not required and fallback to sane defaults.

distributed

# in process (1..n) on cluster (1)
use Zing;

my $zing = Zing->new(scheme => ['MyApp', [], 16]);

$zing->execute;

# in process (1..n) on cluster (2)
use Zing;

my $zing = Zing->new(scheme => ['MyApp', [], 16]);

$zing->execute;

# in process (1..n) on cluster (3)
use Zing;

my $zing = Zing->new(scheme => ['MyApp', [], 16]);

$zing->execute;

This distribution provides a collection of actor-model primitives which can be used to create sophisticated and powerful distributed systems, organized within intricate process topologies.

event-driven

# in process (1)
package MyApp::FileUpoad;

use parent 'Zing::Process';

sub receive {
  my ($self, $from, $data) = @_;

  # react to file-upload events

  return;
}

my $p1 = MyApp::FileUpoad->new;

$p1->execute;

# in process (2)
package MyApp::TextTranslate;

use parent 'Zing::Process';

sub receive {
  my ($self, $from, $data) = @_;

  # react to text-translate events

  return;
}

my $p2 = MyApp::TextTranslate->new;

$p2->execute;

This distribution provides all the prerequisites needed to develop scalable reactive event-driven applications distributed across one or several servers.

fifo-queues

# in process (1)
use Zing::Queue;

my $queue = Zing::Queue->new(name => 'tasks');

# pull from queue
$queue->send({ command => { ... } });
$queue->send({ command => { ... } });
$queue->send({ command => { ... } });
$queue->send({ command => { ... } });

# in process (2)
use Zing::Queue;

my $queue = Zing::Queue->new(name => 'tasks');

# pull from FIFO queue
$queue->recv;

This distribution provides high-performance FIFO message queues which enhance messaging across processes when the order of operations and events is critical and where duplicates can't be tolerated.

hot-reloadable

use Zing;
use Zing::Daemon;

my $daemon = Zing::Daemon->new(
  name => 'myapp-sleep',
  app => Zing->new(scheme => ['MyApp::Sleep', [], 4])
);

$daemon->execute; # pid 12345

# $ kill -USR2 12345

This distribution provides zero-downtime through hot-reloading which is where the process watchers (supervisors) keep the app running and gracefully reload child processes at runtime.

log-shipping

use Zing::Zang;

my $zang = Zing::Zang->new(on_perform => sub {
  my ($self) = @_;

  $self->log->fatal('something went wrong');

  return;
});

$zang->execute;

# tap the logs using the command-line tool

# $ zing logs

This distribution provides the ability to ship the logs of individual processes to a specific centralized channel which can be tapped asynchronously, especially by the command-line tool.

mailboxes

use Zing::Process;

my $p1 = Zing::Process->new(name => 'p1');
my $p2 = Zing::Process->new(name => 'p2');

$p1->mailbox->send($p2->mailbox->term, { say => 'ehlo' });

my $message = $p2->mailbox->recv;

$p2->mailbox->reply($message, { say => 'helo' });

This distribution provides every process with its own unique mailbox which can receive messages from other processes as a means of cooperating through message passing.

management

use Zing::Kernel;

my $kernel = Zing::Kernel->new(scheme => ['MyApp::Logger', [], 2]);

$kernel->execute;

This distribution provides a kernel process that can be used to manage the deployment of child processes and is used to wrap cartridges by the command-line tool when daemonizing process.

multitasking

use Zing::Zang;

my $zang = Zing::Zang->new(
  on_perform => sub {
    # do something
  },
  on_receive => sub {
    # handle something
  },
);

$zang->execute;

This distribution provides all processes with an event-loop that allows them to perform multiple operations in sequence and supports operating in a non-blocking manner.

non-blocking

use Zing::Zang;

my $zang = Zing::Zang->new(
  on_perform => sub {
    my ($self) = @_;

    $self->defer({ command => {...} }) and return;

    return;
  },
  on_receive => sub {
    my ($self, $from, $data) = @_;

    return unless $self->term eq $from; # from myself

    # do something

    return;
  }
);

$zang->exercise;

This distribution provides features that allow processes to operate in a non-blocking manner, yielding and deferring operations, and chunking workloads.

parallelism

use Zing::Process;

my $p1 = Zing::Process->new;

my $f1 = $p1->spawn(['MyApp', [id => 12345] 1]);
my $f2 = $p1->spawn(['MyApp', [id => 12346] 1]);

This distribution provides multiple ways of executing operations in parallel, including spawning processes via forking with the guarantee of not creating zombie processes.

supervisors

use Zing::Zang::Watcher;

my $zang = Zing::Zang::Watcher->new(
  scheme => ['MyApp', [] 8]
);

$zang->exercise;

This distribution provides watcher processes which to supervise child processes but also are capable of multitasking and performing other operations while monitoring supervised processes.

virtual-actors

use Zing::Zang::Spawner;

my $zang = Zing::Zang::Spawner->new(
  queues => ['schemes']
);

$zang->exercise;

This distribution provides the ability to use virtual actors, which are processes (actors) created on-demand as a result of some system event. This feature is enabled by the Zing::Launcher and Zing::Spawner superclasses.

COMMANDS

Given the following process (actor):

# in lib/MyApp.pm

package MyApp;

use parent 'Zing::Single';

sub perform {
  # do something (once)
}

1;

With an application cartridge specifying 4 forks:

# in app/myapp

['MyApp', [], 4]

The zing command-line application lets you manage Zing applications from the command-line using these commands:

start

$ zing start app/myapp

The start command loads an application cartridge which returns a "scheme" and runs it as a daemon.

stop

$ zing stop app/myapp

The stop command finds a running application by its PID (process ID) and terminates the process.

logs

$ zing logs --level fatal

The logs command taps the centralized log source and outputs new events to STDOUT (standard output).

SEE ALSO

The Actor Model

Concurrent Computation

Concurrency in Go/Erlang

The Akka Project

The Actorkit Project

The Orleans Project

The Pyakka Project

The Reactive Manifesto

DISCLOSURES

The following is a list of all the known ways Zing is not like a traditional actor-model system:

  • In Zing, actors act independently and aren't beholden to a system manager.

  • In Zing, actors are always active (each runs its own infinite event-loop).

  • In Zing, actors can communicate unrestricted (no approved communicators).

  • In Zing, actors can block using poll but do not block by default.

  • In Zing, the default datastore/backend is Redis which means the system is (by default) subject to the guarantees and limitations of that system. Data is serialized as JSON and stored in plain-text.

AUTHOR

Al Newkirk, awncorp@cpan.org

LICENSE

Copyright (C) 2011-2019, Al Newkirk, et al.

This is free software; you can redistribute it and/or modify it under the terms of the The Apache License, Version 2.0, as elucidated in the "license file".

PROJECT

Wiki

Project

Initiatives

Milestones

Contributing

Issues