VoIP services operate differently than other applications. These differences make it trickier to host on a data network and require that special steps be taken. Part one of this two-part story at Network Performance Management does a good job of describing the distinctions (part two of the report, available here, continues the discussion).
There are many ways in which the treatment of interactive and static Internet protocol (IP) traffic differs. Much traditional data is sent over IP networks using the transport control protocol (TCP). TCP is stable because packet transmissions are acknowledged and those that don't arrive in tip-top shape (or at all) are resent. This is a time-consuming task, however, best suited for time-insensitive applications such as e-mail.
Conversely, VoIP and other highly interactive protocols use the Real-Time Transport Protocol (RTP). RTP is fast enough to support interactivity. The cost for this speed, however, is the inability to resend packets that are lost or too far out of kilter to be useful. The packets used in VoIP, the story says, also are smaller (and therefore more numerous) than e-mail or packets used for static applications.
The bottom line of all this is that since the transmission protocol is less hardy, the underlying network architecture must be more robust in order to keep problems to a minimum. That means that quality of service (QoS) on VoIP networks must be high and strictly enforced.
This piece at Enterprise VoIP Planet also is part of a series. This is the first installment, and it lays the groundwork for a deeper understanding of what is required to create a well-functioning VoIP service. This installment (the others can be seen here, here and here) deals with the Open Systems Interconnection Reference Model, a seven-layer model that sets the ground rules enabling unrelated computers to communicate over a network. The writer says a lesser-known part of the standard -- which was promulgated by the International Organization for Standardization -- is a management framework for managing computers. The elements addressed in this part of the OSI standard are fault management, accounting management, configuration management, performance management and security management.
This is a very nicely written outline of what VoIP is and why it is important. The VoIP Magazine writer makes a very good point: VoIP is a term that has more than one meaning. To customers, it means good service at the lowest cost, while to service providers it refers to a platform delivering the minimally acceptable quality -- and at the highest possible price. The piece also provides a service by defining QoS in terms of the telecommunications network and the enterprise. The conclusion suggests that QoS is more a goal to continually strive toward than a particular "service condition or status" that can be attained.
The reality is that VoIP is just the tip of the iceberg in terms of highly interactive Internet protocol (IP)-based services. It stands to reason that adding video to the mix will ratchet up the pressure on the network. The volume of data will be greater, as will be timeliness and synching requirements. This EE Times piece doesn't focus on QoS. Instead, it discusses Quality of Experience (QoE) which, as the name implies, emphasizes what the end user perceives, not the underlying metrics of the delivery network. The two factors that are important in QoE, the writer says, are the overall availability of bandwidth and how traffic is managed at the edge of the network.
Unlike sinks and bathtubs that drain in the opposite direction from ours, network infrastructure requirements are much the same in the Southern Hemisphere as in the north. IT managers should pay attention to industry guidelines released by The Communications Alliance on QoS for VoIP and IP networks. Indeed, the standards -- written by Australian industry and government engineers and reported on in VoIP News -- are meant to be international. The IP network guidelines cover best effort and two levels of managed classes of traffic. The VoIP guideline covers delay, echo, codec choice and loss, the story says.