AsyncEventingBasicConsumer slower than EventingBasicConsumer

[Noel McGrath]

I have 2 examples one using EventingBasicConsumer and the other using AsyncEventingBasicConsumer

When I run this the AsyncEventingBasicConsumer is slower to complete.

I would expect it to be faster as threads available to do work when IO going on.

Am I doing something wrong or is it a misunderstanding from me?

Using  EventingBasicConsumer:

class Program

{

static void Main(string[] args)

{

var factory = new ConnectionFactory()

{

//DispatchConsumersAsync = true,

AutomaticRecoveryEnabled = true,

Password = "n",

Port = 5672,

RequestedConnectionTimeout = 3000,

RequestedHeartbeat = 10,

TopologyRecoveryEnabled = true,

UseBackgroundThreadsForIO = false,

UserName = "n",

VirtualHost = "/"

};

var _list = new List<string>() { "trdevmq01b.ccs.local", "trdevmq01b.ccs.local" };

using (var connection = factory.CreateConnection(_list, "test"))

using (var channel = connection.CreateModel())

{

// create queue if not already there

channel.QueueDeclare("testqueue", true, false, false, null);

// set up consumer

var consumer = new EventingBasicConsumer(channel);

consumer.Received += Consumer_Received;

channel.BasicConsume("testqueue", true, consumer);

// publish

var props = channel.CreateBasicProperties();

for (int i = 0; i < 5000; i++)

{

Thread.Sleep(1);

var msgBytes = Encoding.UTF8.GetBytes("Message " + i);

channel.BasicPublish("", "testqueue", props, msgBytes);

Console.WriteLine("Publish " + i);

}

Console.ReadLine();

}

}

private static void Consumer_Received(object sender, BasicDeliverEventArgs e)

{

var body = e.Body;

var content = Encoding.UTF8.GetString(body);

Thread.Sleep(1);

Console.WriteLine("Finished handling " + content);

}

}

Using  AsyncEventingBasicConsumer:

class Program

{

static void Main(string[] args)

{

var factory = new ConnectionFactory()

{

DispatchConsumersAsync = true,

AutomaticRecoveryEnabled = true,

Password = "n",

Port = 5672,

RequestedConnectionTimeout = 3000,

RequestedHeartbeat = 10,

TopologyRecoveryEnabled = true,

UseBackgroundThreadsForIO = false,

UserName = "n",

VirtualHost = "/"

};

var _list = new List<string>() { "trdevmq01b.ccs.local", "trdevmq01b.ccs.local" };

using (var connection = factory.CreateConnection(_list, "test"))

using (var channel = connection.CreateModel())

{

channel.QueueDeclare("testqueue", true, false, false, null);

// set up consumer

var consumer = new AsyncEventingBasicConsumer(channel);

consumer.Received += Consumer_Received;

channel.BasicConsume("testqueue", true, consumer);

// publish

var props = channel.CreateBasicProperties();

for (int i = 0; i < 5000; i++)

{

Thread.Sleep(1);

var msgBytes = Encoding.UTF8.GetBytes("Message " + i);

channel.BasicPublish("", "testqueue", props, msgBytes);

Console.WriteLine("Publish " + i);

}

Console.ReadLine();

}

}

private static async Task Consumer_Received(object sender, BasicDeliverEventArgs e)

{

var body = e.Body;

var content = Encoding.UTF8.GetString(body);

await Task.Delay(1);

Console.WriteLine("Finished handling " + content);

}

}

[Karl Nilsson]

You haven't given us any idea of how much slower it is but generally async is not about raw speed and is in no way guaranteed to be faster than a synchronous implementation.

Cheers

Karl

--
You received this message because you are subscribed to the Google Groups "rabbitmq-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rabbitmq-user...@googlegroups.com.
To post to this group, send email to rabbitm...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/rabbitmq-users/5aa1f975-c1bc-428d-97b2-edd572b8ff33%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Karl Nilsson

[Noel McGrath]

Time taken in seconds for message to be completed in Consumer_Received

AsyncEventingBasicConsumer  90 secs

    EventingBasicConsumer           12 secs

[Noel McGrath]

I did not expect this difference and thought AsyncEventingBasicConsumer  would be faster as while it was waiting for Task.Delay to complete it would be able to do other work.

[Luke Bakken]

Hi Noel,

The .NET client pre-dates async / await and as such isn't "async all the way down". I suspect that if you looked at performance counters for your test application you would see a very high number of thread switches in your async version.

There may be improvements that could be made to speed up the Async* versions of the current codebase but that analysis and work hasn't been prioritized. The code is open-source so if you have the inclination and time to figure out why such a large difference in performance exists we would be very interested in receiving improvements via pull requests.

Thanks -

Luke

[Noel McGrath]

Thanks Luke,

Yes I do see a number of threads in use. I presume this switching is causing the delay, didn't realise it would cost the time difference though