Determining request-response performance with long-lived connections

[Trustin Lee]

Hi Keith,

Sorry for the late reply..

On Thu, Feb 12, 2009 at 1:01 AM, kcounsell
<keith.counsell at billmelater.com> wrote:
>
> I am working on a Netty 3.0.2GA server application with long-lived client
> connections.  Our responses can be returned out of order to the client
> requests, and I am interested in tracking the full response time for a
> single request-response.
>
> My plan so far is to maintain state based on a message identifier and the
> channel.  The decoder would establish the start time and the encoder would
> determine the end time.
>
> Does this seem reasonable?  Is there a better way that I'm missing?

It sounds reasonable to me.  I guess you need to exchange a *lot* of
messages to get more statisticaly correct result.  Please let me know
if you have found a problem (perhaps a bug) with latency or throughput
during the test.  I'd like to hear from you about the result of the
test and improve Netty as much as I can.

> Thanks for the framework -- it's very helpful.

It would be great if you could tell me what property or feature of
Netty helped you writing your application or what features were
missing or inconvenient.  Your feed back will make Netty serve your
needs better in the long term. :)

Thanks,

— Trustin Lee, http://gleamynode.net/