IPBX

Successfully adding VoIP to a data network requires careful considering of a number of factors. Without advance planning, you risk falling into a common pitfall: sketchy voice quality and problems connecting calls or keeping them connected. A network may be sturdy enough to handle all kinds of data applications, but without proper planning will fail to adequately deliver voice.

 

A whitepaper by Clark Brown at FacetCorp on this very topic can help in the planning process for VoIP by pointing out common pitfalls and how to avoid them. The full whitepaper is available online, and is summarized here.

 

Brown begins by explaining that voice applications require certain features from networks that are not important for data applications.

 

“File downloads and database programs require every byte to be delivered correctly, but they are flexible with regard to how long it takes to get the bytes from one location to another,” Brown says in the whitepaper. “Voice, on the other hand, requires the bytes to arrive in a very timely manner, although it is more flexible about losing a few bytes here and there.”

 

Ensuring that features vital for voice are added to the network requires looking at three main topics: network quality, available bandwidth, and packet competition.

 

 

First up is network quality, or the ability of a network to deliver most packets on time. Three main difficulties tend to crop up for voice services if network quality isn’t designed properly.

 

Packet Loss – This refers to the percentage of packets that do not reach their destination. For VoIP, packet loss of more than 3 percent is problematic. Traffic overloads are the primary cause of packet loss.

 

Jitter – This refers to individual voice packets taking different amounts of time to travel from one end of the network to the other. VoIP equipment at the receiving end puts packets into a buffer so they can be played back in an unbroken stream of audio. To ensure audio of acceptable quality, the jitter buffer depth (length of time)should be 50 milliseconds or less.

 

Latency – This refers to the amount of time it takes packets to get from source to destination. Latency of longer than 200 milliseconds is problematic for VoIP. If latency is too long, echo can occur and conversations become frustrating at best.

 

 

In order to successfully deploy VoIP, there must be adequate bandwidth available. Brown notes that “available bandwidth” is not the same thing as “total” or “raw” bandwidth. Rather, it “is a measurable amount of voice traffic that can be transported by the chain

of routers and switches that make up the data network.”

 

Available bandwidth is measured by comparing the amount of voice data that needs to be sent with the amount of voice data the network is capable of carrying. Several factors, summarized below, are involved in this process.

 

Raw Bandwidth – This refers to the overall bandwidth capability provided by each connection. For example, a T1 might provide 1.5Mbits of raw bandwidth. But, connections lose bandwidth because of header information included with each packet. Thus is can be unclear at first how much bandwidth is actually available.

 

Bottlenecks – Bottlenecks in network traffic occur when the available bandwidth on one end is unequal to the bandwidth on the other end. For example, a VoIP device at one end may use a T1 connection, but at the other end there is a smaller, 53Kbps dialup connection. This means the total available bandwidth will be less than 53Kpbs.

 

Streaming UDP – Because voice is sent in many, small UDP (user diagram protocol) packets, it is important to consider the per-packet price imposed by the network technology. An example of this is a 768Kpbs point-to-point connection where files—which are sent in large TCP packets—can be downloaded at 600Kpbs but due to the per-packet price only 180Kpbs of voice traffic can be sent over the connection.

 

Codecs – Determining whether or not the existing network can handle voice traffic requires calculating how much voice traffic there will be—and this depends on the codec used. The G.711 codec provides toll-quality voice, but uses about 80Kpbs of available bandwidth per call. A better choice may be G.729 which provides near-toll-quality voice and uses only 26Kpbs per call.

 

Packet loss, jitter, latency, and sufficient bandwidth are all important to VoIP networks. But a more complicated issue is the competition between data and voice packets.

 

“When there are a lot of data packets to send (as in a file download), they will be put onto the network ahead of voice packets and they will interfere with voice conversations,” Brown says in the whitepaper.

 

This problem requires some sort of queuing scheme to prioritize voice packets over data packets. Three common methods are summarized below.

 

Congestion and FIFO Queuing – One way to handle traffic is the use of routers connecting fast networks and slow ones. The router’s job is to keep the slow side as busy as possible while maximizing the data traveling through the network. All packets from the fast network are queued up and handled as fast as possible by the slow network. But this type of queuing can cause problems with VoIP because it doesn’t distinguish between data and voice packets.

 

Weighted Fair Queuing – A better way to ensure the timely delivery of voice packets is to feed them into a separate queue and be given higher priority. While this method is better than FIFO queuing, it’s still not perfect.

 

Other Queuing Schemes – More sophisticated routers use other schemes to queue packets, and some of those features can be combined with the router’s QoS (Quality of Service) features to ensure voice traffic arrives in a timely manner.

 

Quality of Service, or QoS, is a special configuration the ensures VoIP conversations work well even when there is a lot of data traffic also traveling the network, Brown explains in the whitepaper.

 

QoS configurations fall into four broad categories, outlined below.

 

Best-Effort QOS – Most routers are configured with this type of QoS by default. It uses a queuing scheme that doesn’t make allowances for the special needs of voice traffic, and thus is not appropriate for VoIP in most situations.

 

Differentiated Service – This type of QoS takes the first step toward solving conflicts by using traffic classification to give differential treatment to different types of packets. In most cases, low latency queuing (LLQ) can then be used for VoIP traffic, ensuring that it gets to its destination in time. Older Cisco routers used a special version of LLQ called IP RTP Priority. (The company adopted LLQ in newer versions of its router software.)

 

Dedicated Service – This type of QoS is manifested in routers configured so that a certain amount of bandwidth is permanently assigned for voice traffic.

 

Guaranteed Service – Also referred to a Integrated Services Architecture, or IntSer, this is the most complex method of QoS. It makes “reservations” for bandwidth needed for each call using the RSVP protocol. The reservation is then honored by each router in the path between the source and destination devices. Unlike with dedicated service, the reservation is temporary, lasting only until the call is finished.

 

 

To learn more about data networks and VoIP, check out FacetCorp’s (News - Alert) PBX channel on TMCnet.com.

Mae Kowalke previously wrote for Cleveland Magazine in Ohio and The Burlington Free Press in Vermont. To see more of her articles, please visit Mae Kowalke’s columnist page.

 

 

IPBX