As enterprise networks get bogged down under the weight of streaming video – still having World Cup Soccer flashbacks? – it’s important to have a network monitoring and analysis solution that can track and analyze the most common problems associated with video and voice over IP. Remember, both video and voice are real-time protocols with unique network performance requirements, and the only way to have a problem-free video and VoIP experience for your end-users is to cure the common video and voice over IP ailments: latency, jitter, and packet loss. We typically refer to these as the three-headed beast.
or go to
Ever watched the news and listened to a story being broadcast from the other side of the world? As the signals get bounced from satellite to satellite and, ultimately, from one reporter to another, the delay causes the news anchor and correspondent to be out of sync. The reporter asks a question. After waiting a few seconds without getting a response from the correspondent, the reporter starts to ask again, just as the answer from the first question starts to come in. This interruption causes the correspondent to pause after a few seconds, causing a very awkward, out of sync exchange, all due to latency. This can happen on your network as well. These delays cause conversations to come off as unnatural and sometimes uncomfortable, and that’s assuming the voice or video even makes it to the receiver at all.
Latency is simply the overall delay for data to get from one end-user to the other end-user over the network. Many factors come into play in generating latency, including network propagation delay (electrons still move at some fraction of the speed of light), processing delay within network devices (like queuing and decision latency), and processing delay within VoIP devices (like encoding/decoding, compression/decompression and buffering).
Packet-based analysis can measure latency, letting you know whether or not excessive latency is an issue on your network, and which users might be experiencing latency-based issues. Network administrators can identify the problem in real-time and solve it before latency gets worse. How long is too long? It depends on each situation, but the ITU recommends a maximum one-way delay of 150 milliseconds (msec) for VoIP.
Jitter is the variation in packet delivery intervals at the receiver. Regular delivery of IP packets is required for the final digital-to-analog conversion at the receiver to work correctly. A typical receiver expects packets to be delivered every 20 msec, no more and no less. When the packets start to deviate from this expected delivery sequence, jitter occurs. Most VoIP receivers employ dynamic jitter buffering to reduce the effects of jitter, but when packet delivery cannot be adequately regulated through buffering the resulting jitter can cause static and other audio anomalies, like apparent stuttering, uneven audio, and abnormal speech rhythm. Keep in mind that buffering adds to the overall latency in transmitting VoIP packets, and the 150 msec one-way delay recommended by the ITU takes all sources of latency into account; so increased jitter effectively decreases the budget for other sources of latency on the network.
Jitter can be even more detrimental in multimedia systems. With jitter, videos become jerky or irregular and very difficult to watch. If jitter levels become too high, packet loss will result, with a resulting loss of data.
There are times when packets, especially critical multimedia packets, never make it to the receiver. Packet loss causes missing sounds, syllables, words, or phrases. DSP algorithms may compensate for up to approximately 30 msec of missing data, but anything more and the algorithms typically can’t compensate for the data loss and adverse effects occur.
Real-time protocols like video and voice are much more susceptible to packet loss than traditional network data, since there is very little cushion to wait for missing data or to put out-of-order data back into the right sequence. After about 150 msec or so (remember our ITU latency budget?), any data that is missing or out of order is essentially lost forever, since there is no way to properly reconstruct and maintain the real-time data stream, and this creates gaps (packet loss) in the data.
Gaps of more than 30 msec are noticeable to listeners, and for G.711 that’s about 2 consecutive packets, so it doesn’t take much packet loss for call quality to degrade. A typical word spans approximately15 consecutive RTP (Real-Time Protocol) packets, and at that level of packet loss the complaints will come rolling in. Generally, look to keep packet loss rates below 2% overall for RTP data. Losses of 5 - 10% make calls all but intolerable, and packet loss that comes in bursts is much worse than more dispersed loss.
As more companies move their communication systems onto digital networks, and as the volume of this traffic increases, problems are bound to happen more frequently. Thus it’s important to have a network monitoring and troubleshooting solution that provides full visibility into all traffic on your network, simultaneously, so you can pinpoint the exact cause when problems with voice and video occur.
Authors Bio: Jay Botelho is the Director of Product Management at WildPackets, Inc., a network analysis solutions provider for networks of all sizes and topologies. Jay holds an MSEE, and is an industry veteran with over 25 years of experience in product management, product marketing, program management and complex analysis. From the first mobile computers developed by GRiD Systems to modern day network infrastructure systems, Jay has been instrumental in setting corporate direction, specifying requirements for industry-leading hardware and software products, and growing product sales through targeted product marketing.