Tuesday, November 3, 2020

Rate Limiting and QoS

TL;DR: Conceptually, how is QoS and rate-liming handled in carrier networks? Is there a way I can improve reliability for VoIP and video conferencing when actual bandwidth isn't the issue?

Context: We're a small reseller that provides internet service to some business and residential customers.

Our business customers have strands of fiber that interconnect directly with our head end, and terminate at a switch. We don't have a PON OLT (yet). Basically our fiber service is one big LAN.

Previously each business customer would get a 100Mb/s loop that would simply operate at line speed. We used to have a bunch of crappy 100Mb/s media converters jumbled into a switch but I've since replaced that with an SFP switch. Customers wanted an option to save cost when they didn't need 100Mb or 1Gb service. My solution was to configure rate limiting on the switch ports.

The problem I have now is that when demand is high, packets that exceed the rate limit are dropped. Customers experience some stutter and occasional drops during voice and video calls. This is particularly vexing given the current remote work and school situation.

I've been trying to wrap my head around QoS but when I do a packet capture on the inbound traffic, it all comes in with the same DSCP value. Is that typical?



No comments:

Post a Comment