Voice over IP is growing substantially and both Enterprise network managers and Service Provider network operations teams need to deploy management technology that can measure service quality and identify problems that are affecting the user experience. This task is made more complex both as IP problems can often occur in the last network segment or last mile and as IP problems are typically transient in nature.

 

The positive news for both Service Providers and Enterprise network managers is that the protocols and technology needed to give visibility to the customer premise, user desktop, or teleworker’s home is already available and widely deployed.

 

Problems affecting VoIP performance can range from congestion related packet loss and jitter to signal, noise and echo. Some problems, for example echo, result from combinations of impairments from both ends of the connection and hence can’t be adequately represented by reporting a simple “component” metric; for example an echo level of 40 dB may not cause an audible echo if the associated delay is small — however the combination of an echo level of 40dB at one end of a connection with an large jitter buffer setting at the other end could result in a very annoying echo problem. Impairments can also vary in time and hence the coincidence of problems can make a dramatic difference to the interpretation of the metrics.

 

The VoIP Performance Management Framework is based on the concept of combining embedded performance monitoring agent software in the VoIP endpoints at the network edges with the existing probe/analyzer test and measurement infrastructure. This allows data from both endpoints to be correlated in real time, and hence allows more accurate and meaningful statistics to be reported.

 

The Framework comprises a series of standardized protocols that share a common set of data.

 

 

This is a small software agent that is directly integrated into VoIP endpoints in order to provide real time performance measurements. Implementations such as Telchemy’s (News - Alert) VQmon/EP are extremely efficient, providing continuous performance monitoring and real-time alerts whilst using less than 0.001 MIPS of processing power to monitor a call.

 

RTCP XR payloads are exchanged every 10 seconds or so through the media path between IP endpoints. The RTCP XT VoIP Metrics [1] payload contains a set of metrics that provide high level MOS scores, detailed packet loss/discard statistics and signal related metrics. The functions of the RTCP XR protocol include:

 

A common misconception is that the RFC3611 VoIP Metrics report is to provide data from endpoints back to a collection system. This leads to the mistaken assumption that it is sufficient to implement the protocol at one endpoint only (for example customer premise equipment). In fact the protocol is primarily an information exchange between the VoIP endpoints and is designed to facilitate more accurate call quality estimation (and potentially sending end adaptation), hence it is highly desirable to implement at both endpoints (i.e. for carrier networks, in both CPE and trunking gateways).

 

An extended version of the RTCP XR VoIP Metrics report is under development in the IETF — “RTCP HR” — which is targeted for use in carrier backbone applications.

 

The SIP RTCP Summary [2] protocol has been developed within the IETF and is expected to become an RFC during 2007. The draft protocol has been fairly widely implemented and is used extensively for performance reporting. This protocol reports the same set of metrics as RTCP XR but can be sent directly to a central “collector” such as Telchemy’s SQmediator. This same concept has been applied to the other key signaling protocols, including H.323, H.248 and MGCP, resulting in protocols such as H.460.9 Annex B [3] and H.248.30 [4].

 

The RTCP XR MIB [5] (under development in the IETF’s AVT working group) provides access to the RTCP XR VoIP Metrics via SNMP, and also provides a means to report aggregate data (e.g. min/max/average for some group of calls). This is intended for use either in large endpoints (e.g. large gateways) or in server based systems that have collected data from either SIP RTCP Summary or RTCP XR VoIP Metrics reports.

 

 

Many VoIP equipment and technology providers have implemented components of the Performance Management Framework. Telchemy is the primary source of embedded agent technology within the industry, over 80 equipment vendors use Telchemy’s VQmon in over ten million devices. Texas Instruments (News - Alert), the leading provider of VoIP technology for IP phones and gateways, incorporated VQmon into all their VoIP solutions as part of their Piqua quality initiative and have been leading the industry in the development of adaptive algorithms for problem mitigation based on RTCP XR based call quality feedback.

 

The U.S. Cable Industry adopted the Framework as the basis for the PacketCable 1.5 VoIP [6] performance management architecture, reporting call quality data from MTA’s using RTCP XR and MGCP/NCS extensions.

 

Many of the major Internet Services/ IM/ VoIP service providers have started to use the Framework as the basis for their performance management systems. They have the added problem that they deliver services over broadband infrastructure owned by another, potentially competitive, service provider. The Framework provides multiple benefits – diagnosis of individual customer problems, identification of groups of customers that experience quality degradation (possibly using the same broadband service) and providing data to help determine if improvements to ATAs or softphones would help to solve quality problems. 

 

Traditional carriers have been incorporating requests for RTCP XR support into RFP’s for customer premise equipment, however slow to realize that the protocol should ideally be implemented at both ends of the connection (and hence in trunking gateways and session/border controllers). 

 

 

The concept behind the VoIP Performance Management Framework is simple — integrate a performance monitoring agent into each endpoint and report a set of metrics that capture the user experience and provide diagnostic data.   This is obviously applicable to IP video applications such as IPTV (News - Alert) and IP videoconferencing. There is work underway in developing a similar framework which is occurring within a wide range of standards committees. 

 

-----

 

 

[1] IETF RFC3611 – RTCP Extended Reports – VoIP Metrics Report Block

 

[2] Draft-ietf-sipping-rtcp-summary-03.txt

 

[3] ITU-T H.460.9, for use with H.323

 

[4] ITU-T H.248.30, for use with H.248

 

[5] Draft-ietf-avt-rtcpxr-mib-05.txt

 

[6] PacketCable 1.5 Codec specification