tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Christopher Schultz <ch...@christopherschultz.net>
Subject Re: Using NIO connector with Apache Httpd 2.4
Date Tue, 06 Aug 2013 15:40:26 GMT
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Mark,

On 8/6/13 10:50 AM, Mark Thomas wrote:
> On 06/08/2013 16:09, Christopher Schultz wrote:
>> Mark,
>> 
>> On 8/5/13 10:12 AM, Mark Thomas wrote:
>>> On 05/08/2013 15:50, Abhijith Prabhakar wrote:
>>>> Hi All,
>>>> 
>>>> We are currently using HTTP connector in tomcat 7.42 and
>>>> planning to switch to AJP NIO connector.  When I was reading
>>>> through the docs I found "WARNING: The NIO connector for AJP
>>>> is experimental."
>>>> 
>>>> This made me think that NIO connector might not be mature at 
>>>> this point.  Can somebody who had experience with NIO
>>>> connector let me know if it is stable and any pitfalls I
>>>> should be aware of?
>>>> 
>>>> Stable enough to be used on enterprise apps which is expected
>>>> to get thousands of requests in a hour?
>>>> 
>>>> Any information would be helpful.
>> 
>>> That warning dates form when we added NIO support to AJP. That
>>> was some time ago now and I don't think there is any need for
>>> that any more. I'll remove it.
>> 
>> IMO, using NIO with APR doesn't really give you an advantage:
>> one should have a (somewhat) fixed number of connections between
>> httpd and Tomcat, so the httpd configuration is far more
>> important than the connector you choose on the Tomcat side, isn't
>> it?
> 
> The big advantage is that you can use persistent connections from
> httpd without having to worry about making sure Tomcat has enough
> threads to handle them all or disabling connection re-use. It
> really helps when you have multiple httpd instances load-balancing
> to multiple tomcat instances.

Aah, that does make sense.

Still, having (say) 200 threads on the Tomcat side isn't a big deal
with a single-fronting-server. The real problem is on the httpd side
where if you have 10 back-end servers you'd need 2000 threads if
everything was blocking.

Things get more interesting when you have -- as you say -- an N * M
setup where the number of blocking-connector threads on the Tomcat
side starts to rise geometrically.

Thanks,
- -chris
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.14 (Darwin)
Comment: GPGTools - http://gpgtools.org
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
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=y1hm
-----END PGP SIGNATURE-----

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org


Mime
View raw message