[PDB-tech] proposed new attribute: max_allowed_peering_next_hop_latency

Job Snijders job at instituut.net
Mon Apr 25 11:19:24 PDT 2016


On Wed, Apr 20, 2016 at 05:29:27PM +0200, Arnold Nipper wrote:
> On 19.04.2016 17:19, Job Snijders wrote:
> > I propose a new attribute for network records called
> > "max_allowed_peering_next_hop_latency".
> > 
> > The purpose of this attribute is to signal to other interested
> > parties (human peering coordinators looking for new interconnection,
> > or computer scripts generating route server configs) what the
> > maximum unidirectional latency can be between the two (potentially)
            ^^^^^^^^^^^^^^^^^^^^^^ <- now "Round-trip delay time (RTT)"
> > peering routers.
> > 
> 
> Who of us really is able to do measurement of unidirectional latency?
> Hence I would propose to stick to some sort of RTT (if I understand
> your post in the other thread correctly it would be
> Min(RTT_of_a_series_of_pings)/2))
> 
> Otoh I do not really care as long as it is easy to measure the
> latency.  Doing it via ping *is* easy.

OK, let's change it to mean RTT. There is value in using the value as
produced by a ping utility without dividing by 2. Using RTT is
consistent with what the DE-CIX community expressed on a separate
mailinglist.

Kind regards,

Job


More information about the Pdb-tech mailing list