Re: [tsvwg] FW: New Version Notification for draft-white-tsvwg-l4sops-00.txt

[Jonathan Morton <[e-mail]>]

> [SM] This description ignores the birthday paradox issue that Pete Heist demonstrated. A misbehaving L4S will cause less havok there but it is not without side-effects on stochastic flow queueing systems, we might as well acknowledge that, no?
> 
> [GW] IIRC that issue can occur in fq_codel, but is much less likely in Cobalt (due to set-associative hashing).  Is that correct?

The set-associative hash (presently in Cake, not actually part of COBALT) does significantly push back the onset of flows colliding in the same queue.  It does not 100% eliminate the possibility, so you still have to manage the risk (product of likelihood and consequence) appropriately.  Two flows with mutually compatible congestion responses colliding is a mild enough consequence, but that is not nearly so clear for when the responses are incompatible, especially when it is the newcomer that is dominant.

What no fancy hashing algorithm can compensate for is when the distinction between flows is hidden by a tunnel encapsulation.  Since the L4S team has been adamant in other conversations that tunnels are a common feature on the internet, and I believe we should expect a mixture of conventional and high-fidelity flows within these tunnels, that should definitely factor into your analysis.