posted on Thursday, September 02, 2010 - 04:17 pm
I have been reading the RTPS spec from OMG but there's still some issues I need to resolve, mostly of them related to the real-time scheduling of a DDS system.
* I understand that one RTPS message can comprise several submessages. Therefore, when users try to publish data from a hard real-time stream, this message can include any other kind of info (i.e. heartbeats) which makes harder to bound the maximum size of the message in order to perform the schedulability analysis of your system. How do you resolve this? Obviously, if RTPS messages are always unbounded, one pessimistic approach is taking the largest data packet that can be sent through your particular transport.
* Could an RTPS message include different Data submessages if they share the same destination? (i.e. from different DWs: one from a hard real-time data stream and other from a general purpose data stream). If the latest is true, how is managed the Transport Priority QoS?
* Could DDS metatraffic be sent within the same RTPS message than a regular user data stream or this kind of traffic is always isolated from user traffic?