by Robert Massad
Previous Guest Columns

Policy-based networks: Why not further along?
by Steve Pettit
July 2004

Solve the bandwidth dilemma
by Teejay Riedl
June 2004


Identify your storage options
by Paul Mayer
May 2004

Visualize the virtual network
by James Leach
April 2004

Maximize the power of fax
by Tom Linhard
March 2004

Who will dominate Web conferencing?
by Ian Widger
February 2004

NAS gains traction
by
Joe Disher
January 2004

Focus on data context, not content
by D. Keith Denton

December 2003

Are you ready for Web-age collaboration
by Robert Moore

November 2003

DNS growth has just begun
by Paul V. Mockapetris

October 2003

Has convergence innovation been stifled?
by Iain Milnes

September 2003

Manage VoIP quality and performance
by Robert Massad

August 2003

Is "wireless security" an oxymoron?
by Michael Sutton

July 2003

Pick a provider in 10 easy steps
by Dave McCandless

May 2003

A necessary evolution
by Tom Harper

March 2003

Seek certification of outside partners
by Lindell Wilson

February 2003

Choose a systems integrator
by Judy Matthys
December 2002

 

Robert MassadManage VoIP quality and performance

Market research indicates that VoIP use is becoming more prevalent in the enterprise market. Important concerns about service or call-quality measurement and overall real-time manageability persist, however, as enterprise network managers seek how best to manage their VoIP services.

These managers are concerned that there is often a difference between service-quality levels reported by traditional network-management tools and the opinions expressed by service users. The main reason for this difference is that traditional management tools were not designed to account for network characteristics that are crucial to assessing VoIP call quality.

Network behaviors that can significantly impact VoIP performance include jitter buffer discards and bursts (i.e., varying periods of high and low packet loss). Current management tools, designed for data applications, simply do not address those behaviors.

VoIP requires packet processing beyond the network interface, where the jitter buffer resides. Its role is to add a small amount of delay to each packet to compensate for congestion-induced variations in packet arrival times or jitter at the receiver. Packets suffering from excessive jitter (i.e., packets arriving outside the jitter buffer’s delay window) will be discarded. The discards can cause significant artifacts in the audio heard by the listener. Thus, accounting for discards is crucial to gaining insight into call quality.

Jitter buffer operational details, however, are not visible to traditional on-the-wire or network interface-oriented management approaches. When traditional tools report packet loss statistics, they do not include jitter buffer discards. Yet, these discards are generally a far greater source of quality degradation than lost packets. Management tools lacking insight into jitter buffer discards are fundamentally deficient for VoIP.

The distribution of packet loss and jitter buffer discards is also critical. Packet loss concealment (PLC) capabilities in modern vocoders (voice coder/decoder) can conceal packet loss and discards. Traditional management tools report packet loss statistics only as an average or percentage. They do not consider their distribution, masking potentially significant problems or indicating problems where none exist.

VoIP call quality is usually measured against PSTN “toll quality” voice, focusing on intelligibility to the listener and conversational smoothness. To score “listener” quality, many parameters must be taken into account, including the vocoder type, whether PLC is enabled, whether packets were lost in burst or gap states, the number, density and duration of bursts, and at what point in the call losses and discards occurred. To score “conversational” quality, end-to-end delay must be measured and included in the calculations to understand the potential effects of “double talk.”

Current management tools simply report performance information as independent data items, such as the packet loss percentage or the amount of delay. The network manager is then left with the complex task of correlating the different factors to determine if the call was toll quality.

IP Centrex presents an even more difficult call-quality management scenario for network managers. The IP Centrex service provider provides IP transport and call manager/gatekeeper functionality for IP phones located on customer LANs. Traffic traverses not only the service provider’s network, but also the customer’s corporate LAN and potentially a remote LAN. The subscriber’s LAN would likely be outside the control of the service provider but could be a source of service degradation.

To address these problems, the Internet Engineering Task Force is in the process of extending the current RTP/RTCP standard (RFC 1889/1890) with a VoIP-specific profile that includes the necessary data to fully manage the service. The draft, RTP Control Protocol Extended Reports (RTCP XR), can be found at www.ietf.org/internet-drafts/draft-ietf-avt-rtcp-report
-extns-06.txt
.

Network managers need real-time management tools that can measure VoIP call quality in a clear and simple way. Those tools must identify, measure and correlate the key impairments, their variability in time, and the contextual and human factors driving VoIP quality-level perception. Such tools will alert them to real problems and lead them quickly toward problem resolutions.

For more information from Telchemy:
www.rsleads.com/308cn-256

Massad is vice president of Telchemy Inc., Suwanee, Ga. Send comments for publication to guest@comnews.com.