Asynchronous Awesome

Sometimes, our use of PHP grows beyond the typical request/response cycle of dynamic page generation. Unfortunately, the threaded nature of PHP - and the stateless nature of the server - betrays any efforts to expand our utilization of the server. Image processing, video rendering, APNS (Apple Push Notification Service) integration - any of these can easily take longer than is reasonable for a simple page request. Enter tools like message and job queues that empower daemonized PHP workers to handle data processing in the background. Yet further tools enable long-running event loops and asynchronous Promise-driven operations. PHP isn't multi-threaded, but that doesn't mean you're limited to a single-thread paradigm. I will demonstrate various use cases necessitating asynchronous operations, then delve into the code and the tools that make these systems work.

Every attendee will leave armed with new ways to think about the management of large data jobs in PHP and an understanding of the tools they can use to make it happen.

Transcript

2. 2 Hello. Eric Mann Director of Engineering, Vacasa
4. 4 In 2012, I had to build a way for the “Save PDF” button on this screen to work. It required building a PDF in the background and would sometimes take up to 5 minutes to work. Doing things asynchronously was hard!
5. 5 Report generation PubSub events Parallel I/O Parallel data access
6. 6 Pseudo-cron (WordPress native and tools like wp-async-task)
10. 10 Pseudo-cron (WordPress native and tools like wp-async-task) Mechanical Turk-like job systems Using another language entirely
12. 12 Lorem ipsum dolor sit amet, consectetur adipiscing elit. Fusce elit ex, consequat et tincidunt non, pharetra non risus. Quisque ut leo pretium, eleifend lectus in, ultrices diam. Quisque ac congue urna, non finibus orci. Concurrent Parallel (not covered today) Concurrent Parallel Image by Joe Armstrong - https://joearms.github.io/published/2013-04-05-concurrent-and-parallel-programming.html
13. 13 Lorem ipsum dolor sit amet, consectetur adipiscing elit. Fusce elit ex, consequat et tincidunt non, pharetra non risus. Quisque ut leo pretium, eleifend lectus in, ultrices diam. Quisque ac congue urna, non finibus orci. Concurrent Parallel (not covered today) Concurrent Parallel Image by Joe Armstrong - https://joearms.github.io/published/2013-04-05-concurrent-and-parallel-programming.html
16. 16 <?php // worker.php $importer = new PDFImporter(); $worker = new GearmanWorker(); $worker->addServer('localhost', 4730); $worker->addFunction('import', array($importer, 'import'); while($worker->work());
17. 17 <?php // app.php $client = new GearmanClient(); $client->addServer('localhost', 4730); $job = $client->doBackground('import', $file . '|:|' . $email); $db->record($job);
20. 20 <?php $connection = new AMQPStreamConnection('localhost', 5672, 'u', 'p'); $channel = $connection->channel(); $channel->queue_declare('demo', false, false, false, false); $msg = new AMQPMessage('Hello World!'); $channel->basic_publish($msg, '', 'demo'); $channel->close(); $connection->close();
21. 21 <?php $connection = new AMQPStreamConnection('localhost', 5672, 'u', 'p'); $channel = $connection->channel(); $channel->queue_declare('demo', false, false, false, false); $cb = function($msg) { // Send Tweet with $msg contents } $channel->basic_consume('demo', '', false, true, false, false, $cb); while(count($channel->callbacks)) { $channel->wait(); }
23. 23 <?php $channel->queue_declare('demo', false, false, false, false); $channel->queue_declare('demo-rc', false, false, false, false); $cb = function($msg) { // Send Tweet with $msg contents $result = new AMQPMessage( /* Tweet ID */ ); $channel->basic_publish($result, '', 'demo-rc'); } $channel->basic_consume('demo', '', false, true, false, false, $cb); while(count($channel->callbacks)) { $channel->wait(); }
25. 25 —Promises/A+
26. 26 <?php use Amp\Loop; function asyncMultiply($x, $y) { $deferred = new Amp\Deferred; Amp\Loop::delay(1000, function () use ($deferred, $x, $y) { $deferred->resolve($x * $y); }); return $deferred->promise(); } asyncMultiply(6, 7)->onResolve(function ($err, $result) { var_dump($result); // int(42) });
27. 27 <?php Loop::run(function () { $httpClient = new Amp\Artax\DefaultClient(); $uris = [ "google" => "http://www.google.com", "bing" => "http://www.bing.com", ]; $responses = yield array_map(function ($uri) use ($httpClient) { return $httpClient->request($uri); }, $uris); foreach ($responses as $key => $response) { printf("%s | %d\n", $key, $response->getStatus()); } Loop::stop(); });
29. 29 <?php class WorkerThread extends Thread { private $workerId; public function __construct($id) { $this->workerId = $id; } public function run() { // Do some heavy processing } }
30. 30 <?php $workers = []; foreach (range(0, 5) as $id) { $worker = new WorkerThread($id); $workers[] = $worker; $worker->start(); // Now they'll each run independently } foreach (range(0, 5) as $id) { $workers[$id]->join(); // Synchronously rejoin the main process }
32. 32 <?php Thread::run(function (Channel $channel) { printf("Received: %s\n", yield $channel->receive()); print "Sleeping for 3 seconds...\n"; sleep(3); // Blocking call in thread. yield $channel->send("Data sent from child."); print "Sleeping for 2 seconds...\n"; sleep(2); // Blocking call in thread. return 42; });
33. 33 <?php Loop::run(function () { $timer = Loop::repeat(1000, function () { static $i; $i = $i ? ++$i : 1; print "Demonstrating a live parent for the {$i}th time.\n"; }); try { // Do stuff } finally { Loop::cancel(); } });
34. 34 <?php Loop::run(function () { // Timer try { $context = yield Thread::run() // Child print "Waiting 2 seconds to send start data...\n"; yield new Delayed(2000); yield $context->send("Start data"); printf("Received: %s\n", yield $context->receive()); printf("Thread ended: %d\n", yield $context->join()); } finally { Loop::cancel(); } });
35. 35 Powers things like PDF conversion, video transpiling, APNS Powers cross-app messaging in load balanced environments Power concurrent data migrations for large data sources Powers task management in a performance, scalable, managed fashion
36. 36 Look at what the other communities are doing – we often have the same exact tools available, but a much larger community and lower barriers Challenge yourself to look at traditional programming problems from an asynchronous viewpoint – can things run faster in parallel?