> On Mar 12, 2016, at 17:02, Tom Wiltzius <[email protected]> wrote:
> 
> Thank you both for the information!
> 
> It sounds like there isn't any significant downside the increasing the number 
> of waitress threads beyond the number of available CPU cores if we expect 
> them to be I/O bound rather than CPU bound. Is that true?
> 

This is correct.

> 
> I will investigate our nginx configuration; perhaps it's limiting the number 
> of requests per client to the upstream server. Thanks for that tip. We're 
> using SPDY  3.1 and I'm testing in Chrome, so I don't think the number of 
> requests should be throttled by the client or by nginx on the WAN side (it 
> should be one, persistent TCP connection).

Depending on where your app is blocked, you may be able to have NGINX buffer 
the request/response which may help clear out connections within waitress 
faster.

> 
> I haven't tried uWSGI, but I did try gunicorn and switched to using multiple 
> processes instead of multiple threads. That doesn't seem to have changed the 
> timings much, so I don't think we're blocking on the GIL.
> 
> The last option is the database or SqlAlchemy; I have not ruled that out yet 
> but I can write a script completely outside the context of the web server 
> that makes similar requests and see how it performs.
> 
> Thank you both again for the help.
> 
> On Friday, March 11, 2016 at 2:44:24 PM UTC-8, Jonathan Vanasco wrote:
> > My theory is that if the threads get tied up with a few slow requests, the 
> > server can no longer service the faster ones.
> 
> That's usually the issue.  It's compounded more when you don't pipe things 
> through something nginx, which can block resources on slow/dropped 
> connections.
> 
> A few ideas come to mind:
> 
> i'd take a look at your nginx config.  there are options to throttle the 
> number of connections per client. (upstream and WAN)
> your browser could also have a limit on requests as well, and the keepalive 
> implementation (if enabled on nginx) could be a factor.   are you sure 
> they're being sent in parallel and not serial?
> 
> it's possible that you're having issues with database blocking. 
> 
> it's also possible, though i doubt it, that you're running into issues with 
> the GIL. you could try using uwsgi to see if there is any difference.
> 
> 
> 
> -- 
> You received this message because you are subscribed to the Google Groups 
> "pylons-discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to [email protected].
> To post to this group, send email to [email protected].
> Visit this group at https://groups.google.com/group/pylons-discuss.
> For more options, visit https://groups.google.com/d/optout.

-- 
You received this message because you are subscribed to the Google Groups 
"pylons-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/pylons-discuss.
For more options, visit https://groups.google.com/d/optout.

Attachment: smime.p7s
Description: S/MIME cryptographic signature

Reply via email to