Quality of service takes a new meaning in today’s enterprise networks.
Not so long ago, a service level agreement (SLA) could hide a multitude of sins, a network running with a diversity of minor and/or intermittent problems could appear to be running at 100% efficiency - from the outside. With the move to next generation networks (NGN) and the integration of VoIP and video the 100% efficiency bubble is about to burst.
Traditional data networks are quite forgiving with data packets being re-transmitted in the event of network problems; however, realtime applications such as voice and video are not so forgiving, the 100% network efficiency target becomes more of a prerequisite. The need for more sophisticated monitoring and troubleshooting tools to ensure the high levels of network efficiency required to support NGN applications has therefore become essential. The Network Instruments Observer is an example of the type of tools corporations are having to turn to.
The Observer is able to provide comprehensive performance information on VoIP, video and data traversing the LAN, WAN and wireless networks. However, performance information alone is not sufficient to ensure quality. It must be possible to isolate and determine the cause of any network inefficiencies. Observers analysis capability automatically identifies network problems and can provide an explanation as to the cause and possible cure of such problems.
In providing QoS for VoIP networks the major factors taken into consideration are jitter, packet loss and delay, however, when considering the end user quality of experience (QoE) certain additional parameters must be considered such as hops through additional nodes and hybrid circuits, CODEC efficiencies and audio/analog impairments to mention but a few.
In general, a widely accepted metric for QoS is the MOS scoring system where a value of 5 is a really good voice call and a value of 1 is very poor. When it comes to verifying voice quality, Observer is able to provide the MOS scores for every voice call traversing the network and, in addition, is able to provide a wide variety of call metrics such as, jitter, packet loss, codecs used, signalling type, caller ID, Rfactor, call duration, burst and gap information as well highlighting the cause of any degraded voice call.
Often overlooked or assumed is the network QoS device settings or precedence levels. Once again, in a traditional data network, incorrectly set precedence levels can often be tolerated, but with real-time applications this is not the case. As a general rule, signalling should be set at precedence 3 or higher (a slight delay in establishing a call can be tolerated) while the RTP or the voice portion of the call should be set at precedence 5 or higher. The correct setting of these parameters can often be overlooked or changed during an equipment reconfiguration or swap-out. Observer is able to verify both QoS precedence levels as well as CoS (class of service) parameters if using MPLS.
Verifying QoS in an integrated multimedia network is essential as is being able to identify the cause of any drop in network performance. Is the reduced performance due to increased network delays, abnormal server responses, application problems or illegal/unauthorised activity on the network? It is important to have test equipment available to investigate and resolve any of these issues.
Finally, to ensure QoS, the impact of network growth must be taken into consideration. Capacity planning procedures must be initiated and, if possible, automatic notification of potential problems or catastrophes. In addressing these issues it becomes necessary to trend network performance and set critical alarm criteria at the central site as well as all key remote locations.
For more information contact Frank Barlow, managing director of Barlow Communications, +27 (0)11 886 4730, firstname.lastname@example.org