Hi Ward, I think there musr be some confusion on this. Yes, what you're saying is correct, if only because switches are not infinitely fast. So, yes, after the switch processes one message, the next message will take some amount of time to process.
It's not clear if this thread is asking "what is a typical time to process a flow_mod?"[1] or "is there a way to control the delay of when a flow_mod gets processed?". The answer to the first question is: it depends on the implementation. An optimized software implementation like OVS will likely be sub-millisecond where as I have seen other hardware switches take ~5 ms. As I mentioned before, I created a tool oflops to measure these things. The answer to the second question is: if you want a delay between messages, then the controller should delay sending the messages. I really feel like I'm missing something in this thread -- apologies if I'm being slow. - Rob . [1] or packet_out, as the thread originally stared On Wed, May 11, 2011 at 12:43 AM, Ward hussen <[email protected]> wrote: > Dear Ben Pfaff, > > It means that if OpenFlow switch finishes processing of first FLOW_MOD at > time t, then second FLOW_MOD packet would be processed after t + t0. t0 may > vary depending on load to the switch and t0 will always be greater than 0. > Am I right? > Do you know minimum value of this t0 time? > > Best Regards, > > Ward > > > > >> > Is this possible to change the time gap between processing of two >> > FLOW_MOD?packets in a buffer of switch. >> >> No. > > >
_______________________________________________ openflow-discuss mailing list [email protected] https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
