You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@etch.apache.org by Tony Butterfield <ta...@1060research.com> on 2010/05/18 20:23:27 UTC

Success with Etch

Hi Guys,

just a quick heads up and thank you for great work with the etch 
protocol. We've just
complete a first version of NKP (NetKernel Protocol) which replicates 
the NetKernel
resource oriented middleware that normally only talks to services on the 
same Java
virtual machine but now with NKP can talk across the network. Things 
have worked
out pretty well. A few little things took a bit of working out due to 
lack of documentation
but hey, we had the source!

Etch deserves to be more widely used. It was perfect for our needs 
because we could
focus on the messages and not worry about the low level stuff.

-- 
Tony Butterfield
CTO 1060 Research / NetKernel Architect

Re: Success with Etch

Posted by Tony Butterfield <ta...@1060research.com>.
Hi Scott,

the language reference stuff was really good so creating the .etch 
definition and coding to the stubs
worked really well.

Where we had more problem was with understanding the threading model and
configuring that - we ended up with a slightly customized TodoManager 
and Runner classes. So
that each server and/or client has it's own thread pool with named 
threads. We ended up going
with QUEUED handling of requests because all the heavy work at each end 
is done asynchronously on
our own NetKernel worker threads and having minimal hand off from tcp 
listener threads to
TodoManager threads, then finally to NetKernel already a lot. 
Interestingly we implemented
the protocol with decoupled request and response messages so that 
clients and servers could be
implemented asynchronously to handle a large number of concurrent 
threads without tying up threads.
This works well because of the reliability of tcp and the notification 
of lost connections etc.

The other thing we had to play around with was understanding the 
lifecycle of connections and what
options effect reconnects etc.

Cheers, Tony


scott comer wrote:
> On 5/18/2010 1:23 PM, Tony Butterfield wrote:
>> Hi Guys,
>>
>> just a quick heads up and thank you for great work with the etch 
>> protocol. We've just
>> complete a first version of NKP (NetKernel Protocol) which replicates 
>> the NetKernel
>> resource oriented middleware that normally only talks to services on 
>> the same Java
>> virtual machine but now with NKP can talk across the network. Things 
>> have worked
>> out pretty well. A few little things took a bit of working out due to 
>> lack of documentation
>> but hey, we had the source!
>>
>> Etch deserves to be more widely used. It was perfect for our needs 
>> because we could
>> focus on the messages and not worry about the low level stuff.
>>
> thanks, tony! glad you had success. could you please mention what 
> sorts of things you
> had trouble with?
>
> scott out
>
>

-- 
Tony Butterfield
CTO 1060 Research / NetKernel Architect
tab@1060research.com    +44-1454-311345


Re: Success with Etch

Posted by scott comer <we...@mac.com>.
On 5/18/2010 1:23 PM, Tony Butterfield wrote:
> Hi Guys,
>
> just a quick heads up and thank you for great work with the etch 
> protocol. We've just
> complete a first version of NKP (NetKernel Protocol) which replicates 
> the NetKernel
> resource oriented middleware that normally only talks to services on 
> the same Java
> virtual machine but now with NKP can talk across the network. Things 
> have worked
> out pretty well. A few little things took a bit of working out due to 
> lack of documentation
> but hey, we had the source!
>
> Etch deserves to be more widely used. It was perfect for our needs 
> because we could
> focus on the messages and not worry about the low level stuff.
>
thanks, tony! glad you had success. could you please mention what sorts 
of things you
had trouble with?

scott out