Tips for Successful VoIP Deployment —
Voice over Internet protocol (VoIP) telephone systems are the standard for telephony deployment in any commercial installation today. VoIP offers a myriad of advantages over traditional digital PBX phone systems, including increased flexibility, management, features, integration with computer softphones, and cost. A VoIP phone is essentially a digital phone with an Ethernet interface to packetize audio and control data for transmission over an Ethernet network. While VoIP can integrate into virtually any LAN without problems, issues can arise with network performance that affect the user experience.
Humans are very sensitive to delays and interruptions in speech, and delays of more than 200-250ms become noticeable and distracting. While we can accommodate small amounts of crackle or popping on the line, anything more can cause us to lose important parts of the conversation. These are just 2 examples of the types of quality metrics that are measured when testing VoIP quality of service (QoS) statistics. With these in mind, we can better understand how Ethernet networks operate differently when supporting VoIP traffic compared to standard data or video.
This article discusses the impact of packet size, delay, and jitter, which are important factors that affect VoIP performance.
Ethernet allows for a wide range of packet sizes for various applications. Packet sizes can be thought of as a vehicle that is carrying a number of passengers. A small packet is analogous to a four-passenger automobile while a large packet is analogous to a bus carrying dozens of passengers. Every packet, or vehicle, regardless of size, has an amount of overhead that takes away from its payload. The overhead in a vehicle is the engine compartment, boot, and other areas, that do not carry passengers. Using this analogy, it is easy to understand that a bus is much more efficient at carrying passengers than an automobile when considering the total amount of space the vehicle occupies on the roadway.
Coming back to Ethernet, every data packet has overhead in the form of the source and destination address information at the start and end, leaving the payload in the middle. Maximum efficiency is achieved when large packets are used in favor of small packets. A larger packet means fewer total packets are transmitted to carry a given amount of payload. Fewer packets on the network results in less workload on switches and routers which must process each packet by examining the source and destination address to determine where next to send them.
The standard size for Ethernet data packets is 1,518 byte which is used for virtually all data types from email to streaming audio and video. Most Ethernet protocols know when a network error occurs that results in a lost packet allowing the packet to be resent and the full message to be transmitted. When sending a large amount of data, a few out-of-order packets do not adversely affect the message, and the device on the receiving end will collect the late packets to complete the full message. This is fine when the information going across the network is not absolutely time-sensitive. Because humans are very sensitive to small delays when speaking to each other on telephones, VoIP protocols cannot use 1518 byte Ethernet packets.
Instead, VoIP protocols typically use 64 byte frames to maximize speech quality at the expense of network efficiency. When speech is broken up into small packets, the resulting data on the network is a rapidly flowing stream of small packet, each carrying just a fraction of a syllable of speech. Should one of these packets get dropped by the network, the person at the other end may hear a slight pop or crackle on the line. If the conversation was packed into large packets, one dropped packet may result in entire words being lost, making conversation impossible on a congested network.
Imagine transporting an entire football team and the staff on a highway using a single bus versus many automobiles. If the bus has a breakdown, the entire team is late. However, if using automobiles and one vehicle suffers a breakdown only 4 or 5 people are late, and the remainder will arrive on time. This is the same logic used by VoIP systems. It is better to get most of the conversation across at the expense of poor network efficiency than to have an efficient but unusable conversation.
VoIP Testing and Diagnostics
Understanding the key difference in how networks handle VoIP versus virtually all other forms of data, we can start to look at the key factors in measuring network performance for optimal VoIP quality, the 3 most critical being packet loss, delay, and jitter.
PACKET LOSS is a percentage of the total packets that are lost (discarded) by the network. Network switches and routers will discard packets if the incoming buffer is full due to congestion on the outbound side that prevents packets from being forwarded to the next hop on their way to the destination. Acceptable packet loss depends on many factors, though 3% or less is considered good.
A transmission tester is a tool that sends a stream of packets between 2 locations and measures the loss rate. When measuring VoIP, it is important to use a VoIP preset or manually configure the packet size to 64 byte to accurately demonstrate network performance.
DELAY is the time required for a packet to cross the network. Factors that increase delay are the number of switches and routers between the users on the call. LAN and WAN congestion leads to routers searching for alternate paths between locations often resulting in additional “hops”, with each additional hop adding
An acceptable delay time for VoIP is 200 ms (milliseconds) or less. A simple PING test can measure delay showing the minimum, maximum, and average delay. Be sure to change the default PING settings for packet size, count, and pause (time between tests). Optimal settings are 64 byte packet size, 1000 packets, 10 ms interval. The result is a 10-second test with 100 packets/second.
JITTER is the difference in delay time between packets. The less jitter, the more consistent the stream of packets is and the smoother speech sounds. Most VoIP equipment incorporates a jitter buffer to accommodate some amount of jitter, though less jitter always results in better sound quality.
Jitter of 30 ms or less between packets is best. Voice quality falls off dramatically as jitter exceeds 35 ms.
Jitter can be measured with dedicated network testers with dedicated hardware. Oftentimes computers and mobile devices introduce too much error to measure jitter accurately.
These 3 factors all play a role in the rated quality of a VoIP conversation. Each are affected by different network conditions. Therefore, it is vital that a network engineer troubleshooting VoIP systems has the proper tools to evaluate network conditions and aide in identifying sources of poor VoIP quality.