[tao-users] CORBA using oneway calls

[Johnny Willemsen]

Hi,

Thanks for using the PRF form. When you do regular two way request you
each time send data to the server, wait for user code, send back the
reply, and then the next request. With a oneway the client can just send
the data to the server without waiting for the user code and the
response, that is much faster as you noticed. You could even enable
buffering which could reduce the amount of tcpip packets which could
make it faster.

When you have a single threaded client and server you probably don't see
any out of order, but it is not guaranteed, when you enable multi
threading you will see out of orders, if there is an order at
application level I would recommend to add that into the user data.

Best regards,

Johnny Willemsen
Remedy IT
Postbus 81 | 6930 AB Westervoort | The Netherlands
http://www.remedy.nl

On 06/16/2016 10:28 AM, Quek Tai Wee wrote:
> Hi,
> 
>  
> 
> TAO VERSION: 1.6a_p13
> ACE VERSION: 5.6a_p13
> 
>  
> 
> HOST MACHINE and OPERATING SYSTEM:
> 
>      Red Hat Enterprise Linux Server release 6.6 (Santiago)
> 
> 
> COMPILER NAME AND VERSION (AND PATCHLEVEL):
>         g++ (GCC) 4.4.7 20120313 (Red Hat 4.4.7-11)
> 
>  
> 
> DOES THE PROBLEM AFFECT:
>     COMPILATION? No
>     LINKING? No
>     EXECUTION? Yes
> 
>  
> 
> SYNOPSIS:
> 
> Using TAO for a production equities feed publish/subscribe system.
> Would like advice on using oneway calls.
> 
> 
> DESCRIPTION:
> 
> We are doing performance tests on the throughput of the publisher
> sending 100K messages to subscriber, both running on different servers.
> 
> The average timings are around 25 secs for 100k messages.
> 
> The IDL method is something like this :
> 
>     void deliverMessage(in SL_Msg::Message msg);
> 
>  
> 
> Alternatively, when we run same tests using the oneway call , the
> timings are around 3 seconds for 100k messages, a vast improvement.
> 
>     oneway void deliverMessage2(in SL_Msg::Message msg);
> 
>  
> 
> Can anyone advised if it is ok to use oneway calls in the scenario that
> I have described ?
> Are there any cons to this approach , like possibility of receiving data
> out of band ? 
> 
>  
> 
> We have tried to send up to 500K messages with a running sequence number
> and so far there has been no out of sequence messages.
> 
> 
>  
> 
> Regards,
> 
>  
> 
> Quek Tai Wee
> 
> 
> 
> _______________________________________________
> tao-users mailing list
> tao-users at list.isis.vanderbilt.edu
> http://list.isis.vanderbilt.edu/cgi-bin/mailman/listinfo/tao-users
>