| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Hello Steinar, As you correctly state the use of 1588 transparent (or boundary) clocks would allow, at least in principle, a lower priority to be
set for a 1588 stream, between that of RoE and “other” traffic. However transparent clock functionality is by no means universal and is still considered by some as controversial. It therefore may be the case that it might be necessary to use the same priority
as RoE packets. Deterministic insertion of PTP is in principle possible (at least in some scenarios where the PTP devices are aware of, or are the source of the RoE stream) but would constrain the synchronisation topologies possible. Although the required bandwidth for an individual 1588 stream is small compared to the RoE flow it is associated with, and the 1588
packets themselves are relatively small, there will be multiple 1588 streams,
1 per 1588 master / slave pair and their impact on packet jitter of the RoE stream (or conversely that of the RoE on PTP performance) and “other” traffic blocking behaviour may perhaps become significant in some deployment scenarios. Of course all of this pre-supposes the use of 1588 as at least part of the synchronisation solution. Some of the issues could be addressed
by specifying adopting a particular 1588 profile mandating 1588 encapsulation, message rates, use of on path support (Boundary or Transparent clocks), etc.. but perhaps that is then staying too far from the intent of this group? Regards Peter From: Steinar Bjørnstad [mailto:Steinar.Bjornstad@xxxxxxxxxxxxxxx]
Hello, As I know 1588 it is a packet-stream with a much lower packet-rate than a ROE stream. If 1588 is tunneled transparently through the network, it will be very sensitive to packet-jitter and must receive highest priority. In the case of a
CBR ROE It should however be possible to deterministically insert 1588 packets in between the ROE packets which will be generated in a deterministic way. In the case of transparent clock processing of 1588 packets at intermediate nodes packet-jitter is tolerated
and the packets will not need the same priority as ROE packets. This may be required in the case of VBR ROE streams? I.e. I am not sure if I understand how e.g. 1588 should impact packet jitter and packet/frame size. Perhaps you can help me out with a concrete example? Best regards Steinar 26. feb. 2015 kl. 12:25 skrev Peter Turnbull <pturnbull@xxxxxxxxxxxxxxx>:
|