[DLNEX] passive probe measuring network latency in combination of network exposing Latency for special services

[Linda Dunbar <[e-mail]>]

At NANOG this week, there are many companies showing off their network monitoring capability, from BGP route monitoring to Latency Monitoring. Companies like Teridion, ThousandEyes, Internap, etc. all use passive probe packets to measure the latency through a network. Some of those companies can provide shorter latency path for upper layer.

If network elements or segments support DETNET, OIF's FlexEthernet, or other latency optimized mechanisms within a node, and can allocate resource for a percentage of traffic (latency sensitive flows), is it is more valuable for those nodes or segments to expose the upper bound latency to upper layers on those flows than the passive monitoring with probes? Possible for upper layer to request those resources (via some kind of control plane)? Like what CCAMP has done?

Facebook presentation at NANOG pointed out (https://www.nanog.org/sites/default/files/20160922_Quinn_Being_Open_How_v1.pdf ):
BGP can only do so much
* BGP offers little data about the forwarding plane
* BGP "best" paths are oblivious to:
* Capacity
* Latency
* Packet Loss
What if the network elements can expose those attributes?

Just some points for discussion.

Linda Dunbar