nginx performance test

yong xue ultraice at gmail.com
Mon Feb 1 12:05:05 MSK 2010


Yes, I think so too.
I reviewed the code, and this is problem about design.

the current design is much clear.
Maybe Sysoev have some simple approach.

2010/2/1 agentzh <agentzh at gmail.com>

> 2010/1/30 yong xue <ultraice at gmail.com>:
> > hi, sysoev,
> >     for proxy, can nginx give an new option, for example
> > client_max_body_size_in_buffer,  it will be served synchronously from
> > client
> > if client body size is greater than client_max_body_size_in_buffer ?
> >
>
> According to the current implementation, no easy way. ngx_proxy calls
> the ngx_http_read_client_request_body function to do the content body
> reading task for it, which always buffer the input request before
> creating the request for the remote upstream server.
>
> Even though technically speaking we *can* do that, but I'm afraid it
> will make things even worse if the backend server blocks a thread or a
> process on slow request processing (as in the Apache prefork mpm).
>
> So I don't think Igor Sysoev will do that *big* refactoring for
> something that often has little gain in real world ;)
>
> Cheers,
> -agentzh
>
> QQ 279005114 *grin*
>
> _______________________________________________
> nginx mailing list
> nginx at nginx.org
> http://nginx.org/mailman/listinfo/nginx
>



-- 
Best Regards,


QQ:312200

e-mail:ultraice at gmail.com <e-mail%3Aultraice at gmail.com>
MSN:it at easy-boarding.com <MSN%3Ait at easy-boarding.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://nginx.org/pipermail/nginx/attachments/20100201/9b34dfeb/attachment.html>


More information about the nginx mailing list