Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [jetty-users] Unix socket performance numbers

Hi,

On Fri, Nov 20, 2015 at 1:01 PM, Silvio Bierman
<sbierman@xxxxxxxxxxxxxxxxxx> wrote:
> Hi,
>
> Thanks. I would expect a speed difference between Java and native code but
> thought that SSL handling was part of the native JVM code.

It is not native C code, it's the JDK Java code, at least with the
default configuration.

> Therefore I
> suspected a slight difference if any at all. The results Greg shows us are
> very far apart.

As he said, take them with caution; they are experimental and we need
to confirm that the slow down is indeed caused by TLS.
Won't be the first time I *think* that something could be the reason,
just to be wrong later on.

Don't take these results as "The Truth"; Greg wanted to spark interest
on the Unix Pipe Connector feature.

> Perhaps I need to do a test setup. I would have leaned toward nginx before,
> is there a reason why I should prefer HAProxy?

HAProxy can talk HTTP/2 to the backend, nginx cannot (yet).

I'm guessing that because you are on 9.3 to leverage the SNI features,
you are also on HTTP/2 ?
If so, definitely HAProxy, see also:
https://webtide.com/http2-with-haproxy-and-jetty/

-- 
Simone Bordet
----
http://cometd.org
http://webtide.com
Developer advice, training, services and support
from the Jetty & CometD experts.


Back to the top