Load-Balancing at connection time only

in Help Center
Hi,
In the current version of ICE, load-balancing is operated at connection time. Are there any plans to make this mechanism available at a request level? Or at least for stateless methods. For example, we could imaging having a keyword in the slice definition file that would indicate that a method is stateless and can therefore be load-balanced for every client call.
Regards,
Cyrille
In the current version of ICE, load-balancing is operated at connection time. Are there any plans to make this mechanism available at a request level? Or at least for stateless methods. For example, we could imaging having a keyword in the slice definition file that would indicate that a method is stateless and can therefore be load-balanced for every client call.
Regards,
Cyrille
0
Comments
You can already achieve per-request load balancing by modifying the proxy configuration programmatically and disabling the caching of the connection with the proxy. When proxy connection caching is disabled, each invocation on the proxy will obtain a random connection matching one of the endpoints of the proxy.
For more information, I recommend reading the connection establishment chapter in the Ice manual.
Cheers,
Benoit.
Hello Benoit and many thanks for your answer.
Still I have a question regarding performances. Since the connection will be recreated for every call, I suppose there is going to be an overhead in the processing time of each request. I beleive that the Ice layer does not keep a connection pool on the client side. Correct?
If these assumptions are correct, would removing the cache would be something that you would recommend? Or is this only a woraround to allow load-balancing at request level?
Regards,
Cyrille
Cheers,
Benoit.
Benoit,
Thank you for your answers.
Cyrille