multithreading not working for requests to the same url

multithreading not working for requests to the same url

[Mathijs]

Hi all,


In my application, a lot of time is spent on /inbox.
However, it looks as if rack does not support handling multiple
requests to the same url simultaneously.

I used this 'demo' app:
class HelloWorld
  def call(env)
    sleep 15
    [200, {'Content-Type' => 'text/plain'}, ["Hello world!"]]
  end
end
run Helloworld.new

run with rackup demo.ru
open 3 browser tabs, enter these urls:
- http://localhost:9292/path1
- http://localhost:9292/path1
- http://localhost:9292/path2

You'll see that if you start the requests roughly at the same time,
the first path1 and path2 will complete at the same time.
The second path1 comes 15sec behind.

Passing params (localhost:9292/path1?blah=3) will
trick the app to handle the request in parallel. Also, aliasing
localhost does the trick.
So it looks as if rack does some locking per url.
I couldn't find anything though.

I tried this under REE with mongrel and webrick and under jruby with
trinidad and webrick.
They all behave the same.

Is anyone aware of this behavior? What is the cause?
Thanks,
Mathijs

Re: multithreading not working for requests to the same url

[Konstantin Haase]

Concurrency is up to the concrete Rack handler (webserver), not rack.

Konstantin

On Jan 21, 2011, at 13:47 , Mathijs wrote:

> Hi all,
> 
> 
> In my application, a lot of time is spent on /inbox.
> However, it looks as if rack does not support handling multiple
> requests to the same url simultaneously.
> 
> I used this 'demo' app:
> class HelloWorld
>  def call(env)
>    sleep 15
>    [200, {'Content-Type' => 'text/plain'}, ["Hello world!"]]
>  end
> end
> run Helloworld.new
> 
> run with rackup demo.ru
> open 3 browser tabs, enter these urls:
> - http://localhost:9292/path1
> - http://localhost:9292/path1
> - http://localhost:9292/path2
> 
> You'll see that if you start the requests roughly at the same time,
> the first path1 and path2 will complete at the same time.
> The second path1 comes 15sec behind.
> 
> Passing params (localhost:9292/path1?blah=3) will
> trick the app to handle the request in parallel. Also, aliasing
> localhost does the trick.
> So it looks as if rack does some locking per url.
> I couldn't find anything though.
> 
> I tried this under REE with mongrel and webrick and under jruby with
> trinidad and webrick.
> They all behave the same.
> 
> Is anyone aware of this behavior? What is the cause?
> Thanks,
> Mathijs

Re: multithreading not working for requests to the same url

[Mathijs]

Thanks for the speedy reply.

I found my problem.
It turns out that it was my webbrowser queueing the requests.
Both firefox and chrome do this, but when hitting the same url from 2
different browsers, everything works fine.




On Jan 21, 1:59 pm, Konstantin Haase <k.ha...@finn.de> wrote:
> Concurrency is up to the concrete Rack handler (webserver), not rack.
>
> Konstantin
>
> On Jan 21, 2011, at 13:47 , Mathijs wrote:
>
> 144145
>
> 146
>
>
>
> > Hi all,
>
> > In my application, a lot of time is spent on /inbox.
> > However, it looks as if rack does not support handling multiple
> > requests to the same url simultaneously.
>
> > I used this 'demo' app:
> > class HelloWorld
> >  def call(env)
> >    sleep 15
> >    [200, {'Content-Type' => 'text/plain'}, ["Hello world!"]]
> >  end
> > end
> > run Helloworld.new
>
> > run with rackup demo.ru
> > open 3 browser tabs, enter these urls:
> > -147http://localhost:9292/path1
> > -148http://localhost:9292/path1
> > -149http://localhost:9292/path2
>
> > You'll see that if you start the requests roughly at the same time,
> > the first path1 and path2 will complete at the same time.
> > The second path1 comes 15sec behind.
>
> > Passing params (localhost:9292/path1?blah=3) will
> > trick the app to handle the request in parallel. Also, aliasing
> > localhost does the trick.
> > So it looks as if rack does some locking per url.
> > I couldn't find anything though.
>
> > I tried this under REE with mongrel and webrick and under jruby with
> > trinidad and webrick.
> > They all behave the same.
>
> > Is anyone aware of this behavior? What is the cause?
> > Thanks,
> > Mathijs