On 7/15/2010 9:58 AM, Yaakov Stein wrote: > Mikael > > As you have not been completely happy with the answers so far, > let me try to help. > > 1588v2 can (and usually does) run over UDP/IP (although it has other > transport options, > including a pure Ethernet one with Ethertype 88F7). > > So why do we need 1588 instead of NTP which has been around for a long time ? > 1588 has an advantage when there is support from the network elements it > traverses, > and this happens in Ethernet switches.
And NTP already has some support for 1588 where it is available. PTP in contrast which is strictly hierarchical and basically only uses 1588-enabled Ethernet. > To date, there aren't routers that provide > on-path 1588 support (your "new compliant hardware"), > and so there is little reason to prefer 1588. > The reality is that there is something like a 10-15 year rollout of any scheme in new hardware that will be able to deploy such PTP/1588 types of networks unless there is a clear and pressing reason to do so. In that case the protocol *must* be right or the whole thing will die under its own weight. What I haven't seen anywhere is any analysis of how all of the processing that is necessary for example in the MPLS proposal, will affect things like latency, jitter, precision, etc. of the resultant packets to the recipient client system. If it cannot deliver better results than what is currently available then there's no point to the whole exercise. Danny > In fact, there are multiple reasons to prefer NTP in such cases > (security, scalability, bandwidth consumption, ...). > > The 1588-2008 standard is available from the IEEE. > The Q13/15 work in the ITU has produced the G.826x series of Recommendations, > that are available from the ITU-T web site. > Contributions to Q13 are only visible to those with a TIES password. > > There IS a 1588 wikipedia entry - > http://en.wikipedia.org/wiki/Precision_Time_Protocol > and much more material available on-line (e.g., John Eidson's tutorial). > > Y(J)S > > -----Original Message----- > From: [email protected] [mailto:[email protected]] On Behalf Of > Mikael Abrahamsson > Sent: Wednesday, July 14, 2010 08:45 > To: [email protected] > Subject: [TICTOC] sync over IP > > > Hello. I'm a router guy. > > I work for a mobile phone operator, we've been in the mobile phone and IP > business for ~20 years. We're running GSM, UMTS and now deploying LTE. Our > network also carries a lot of IP traffic for commerical and residential > customers, mostly Internet traffic but we're also using MPLS L3 VPNs to > provide L2/L3 services to our customers, mostly L3 VPNs. > > We're going all-IP for all the mobile services, meaning we want to carry > everything over IP, including sync and circuit switched calls. > > This has made me a bit interested in sync over IP (since I've also been > using NTP for 15+ years). I read the charter and it seems to indicate that > this WG is trying to standardize sync over IP. > > When I look at the last years communication on this list I see a lot of > reference to 1588v2 and SyncE. When I look at these I keep seeing > references to "ethernet switch" in the diagrams, not IP routers. I also > see lots of references to new "compliant hardware" being announced. This > worries me. > > As an IP operator, we prefer to do routing as far out in the network as > possible, as we've had bad experience with large L2 switching domains. We > also prefer to keep our equipment for as long as possible, so any new > service we deploy should work over existing infrastructure with devices > that might be 10 years old. Deploying new IP/MPLS core/distribution > hardware to support the new services is something we do not want to do. > > I've been trying to understand the network requirements of 1588v2 but come > up short. There is no 1588v2 article on wikipedia at all. > > So I guess my question is if someone could please point me in the right > direction, where can I find a primer for "sync over IP for a router guy" > to make me better understand what's happening here? > _______________________________________________ TICTOC mailing list [email protected] https://www.ietf.org/mailman/listinfo/tictoc
