VOIP Traffic Segregation Should Be App Developers' Work, Phone-To-Phone

To achieve highest possible VOIP quality, it seems obvious that phones should generate IP UDP packets with the proper markings to segregate the traffic. The problem is that IP network is a cloud of many competing routing protocols, each with its own forwarding algorithm, each is bent on dominating the market for monetary gains. Civilizations come and go; protocols pop and burst throughout history. How can the Internet work worldwide is a mystery, let alone ensuring the highest quality, right?


People's will makes the roads, VOIP is no exception. The competing protocols need to come together to make ways for the greater good. The evidence is in an experiment with this 2-leg setup, . And the netflow collection export shows that, indeed, legitimate hardware developers faithfully copy DSCP field from different vendors into MPLS labels and out onto phones.

I first checked with freeware Kibana Elk collector, which omits MPLS flow exports, but I hold out the hope that the actual packet have faithful replications.

I used the common flow exporting configuration,

, and I see tell-tale signs of MPLS packet markings in the "experimental bits" fields with the VPN label as well as the transport label,

, then I turn to commercial grade Plixer netflow collector, and bang, CS3 marking for SCCP control packets (TCP port 2000) is in the rainbow graphing,

, and further inspection of all the flows between phones and CUCM shows TCP non-SCCP-control packets that have common DSCP 0 in the MPLS labels.