One of the most important things that must be configured in concert with available VoIP solutions is Quality of Service (QoS). Without QoS options properly configured, the quality of voice (and video) could, and probably will be, sacrificed along with the overall demands of general traffic. These options provide a priority channel that is used by the voice traffic so that quality can be maintained while also allowing general traffic flow. This article reviews QoS basics and briefly discusses available QoS options and how they operate to provide quality for voice traffic.
Many of these QoS concepts are integral when studying for a Cisco voice certification. QoS concepts are covered on all of the following exams:
• 640-461 ICOMMv8.0 – CCNA Voice
• 642-437 CVOICE v8.0 – CCNP Voice
• 350-030 CCIE Voice Written – CCIE Voice
There are a number of QoS factors to consider when configuring a modern QoS implementation on Cisco, or any other vendor’s equipment. However, the most basic of these concepts revolves around what QoS is attempting to accomplish. There are four major factors that need to be controlled in order to have a quality VoIP phone call; these include:
• Bandwidth – The amount of end-to-end available bandwidth dictates whether a call will work correctly or not. With unlimited constant bandwidth, a voice call can work from end-to-end without much issue; however, bandwidth is rarely unlimited. The codec selected for use over a specific line is dictated by the amount of available bandwidth and the number of active calls required.
• Delay – Unlike with data communications, too much delay on a voice call can make the quality of the call unbearable. Of course, all voice communications have some amount of delay which must be kept to a number that is as small as possible. Typically, with VoIP, optimum call quality includes an end-to-end delay of less than 150ms.
• Jitter – Jitter is the amount of delay variation in call traffic. If traffic over a connection is constantly delayed at 100 ms, no issue occurs. However, if for the first portion of the call there is short delay (e.g., below 5ms), followed by a period of long delay (e.g., over 300ms), and then another short delay, the receiving voice device may have trouble synchronizing all of the incoming traffic as it is received in an inconsistent manner.
• Loss – Obviously, the loss of voice packets results in the loss of audio on the connection. Small amounts of loss (< 1%) over the course of a connection will probably not be noticed, but if this loss becomes a large problem then significant loss in voice quality occurs.
There are a number of different methods that can be used to control the QoS of a voice connection; these include:
• Classification and Marking
• Link Efficiency
• Congestion Management
• Congestion Avoidance
Classification and Marking
The most commonly used method of QoS classification and marking is Differentiated Services (DiffServ). The general concept of DiffServ is to monitor the traffic coming through a device; all traffic is then classified into a specific traffic classification (for example, Voice Traffic or Data Traffic). Once this traffic is classified, it is marked with this classification using one of a number of methods. Commonly with IP traffic, the ToS field is used in the IP header and is classified with a Differentiated Service Codepoint (DSCP). This marking is then used by successive devices in prioritizing which traffic to process first.
See related article on QoS Marking and Classification
There are a number of different link efficiency mechanisms. The most commonly known mechanisms include IP header and payload compression. Other mechanisms include Link Fragmentation and Interleaving (LFI). These are typically used on slower speed serial links to improve delay by fragmenting larger packets into smaller ones, thus allowing other smaller packets to be processed. Obviously, the more efficient the link, the less delay is subject to a VoIP connection.
The concept of congestion on a connection is rather simple to explain; the more congested a link, the less likely a packet will be able to get through in a timely manner required by VoIP (think, rush hour in NYC or LA). Congestion management mechanisms attempt to control the amount of congestion faced by traffic by processing the traffic in a variety of different ways, some more complex than others. Many of these methods are used in conjunction with markings given to traffic (e.g., DSCP). The most common methods include:
• Priority Queuing (PQ)
• Custom Queuing (CQ)
• Weighted Fair Queuing (WFQ)
• Class Based – Weighted Fair Queuing (CBWFQ)
• Low Latency Queuing (LLQ)
See related article on Queue Configuration and Congestion Management.
Congestion avoidance is another method of QoS; the most common of the techniques used is called Weighted Random Early Detection (WRED). Basically, WRED attempts to predict that congestion will be forthcoming, and when this happens packets are selectively dropped to avoid congestion.
There are a number of different QoS concepts that must be understood in order to properly implement a VoIP network or pass the Cisco voice certification tests. The concepts covered in this article are a simple overview of the high level QoS options available. Hopefully, this article will help the student understand these high level concepts before digging into the depths required for true understanding.