Modern telephony may have neatly combined voice and data systems into a seemingly seamless whole. But it's important to understand that not all "pa

The "packets" in this case are tiny pieces of voice and data that travel between sender and receiver in a network.

IP telephony (IPT) sends voice conversations over the same network that carries data throughout your company, whether it is a local-area network (LAN), a wide-area network (WAN) or even the public Internet. IPT technology breaks the sound into tiny digital units called "packets," then sends those packets over the network and reassembles them in the correct order on the receiving end.

But while data and voice can share the same network, they have different requirements. It is helpful to understand how voice differs from data, in order to ensure that they play well together on the same network.

Voice works in real time

Voice is described as a real-time application. Thus it's critical for the packets to flow smoothly; a few lost packets can noticeably degrade the quality of the communication. In order to ensure that a voice conversation maintains an acceptable quality level, voice must be guaranteed a certain amount of bandwidth on the network.

Voice doesn't necessarily have to preempt all the data traffic on the network, but it does need to be guaranteed a certain minimum amount of capacity. By incorporating a set of standards-based tools or mechanisms called Quality of Service, companies can make sure that both voice communications and mission-critical data applications perform well.

Voice is more predictable than data

Voice traffic is more orderly and predictable than data traffic. Data traffic is often "bursty," which means subject to spikes of activity that can consume all of the available bandwidth, while voice traffic takes a predictable amount of bandwidth. Voice traffic is also symmetrical; that is, the packet size tends to be the same in both directions. Data traffic, on the other hand, is generally characterized by smaller uploads (such as database queries) and larger downloads.

Voice is less tolerant of delay

Further complications arise because voice is an application that cannot tolerate much of a delay. If there is too much traffic on the line, or if a voice packet gets stuck behind a large data packet, the voice packet will be delayed to the point that the quality of the call is compromised.

Latency is the average travel time it takes for a packet to reach its destination. The maximum amount of latency that a voice call can tolerate one way is 150 milliseconds (100 milliseconds is optimum).

Voice doesn't like variability

In order for voice to be intelligible, consecutive voice packets must arrive at regular intervals. "Jitter" describes the degree of variability in packet arrivals, which can be caused by bursts of data traffic or just too much traffic on the line. Voice packets can tolerate about only about 75 milliseconds (40 milliseconds is optimum) of jitter delay.

Voice doesn't like packet loss

Packet loss is a common occurrence in data networks, but computers and applications are designed to simply request a retransmission of lost packets. Dropped voice packets, on the other hand, are discarded, not retransmitted. Voice traffic can tolerate less than a 3% loss of packets (1% is optimum) before callers begin experiencing disconcerting gaps in conversation.