This technology primer aids wide area network engineers in understanding CoS/QoS basics. Learn what quality of service is and isn't, along with how it compares with class of service. To transform your basic IP network to a robust one that can traffic voice and video applications, read this introduction to reliable packet delivery.
When your packet absolutely, positively has to be there…
A basic IP network—like the Internet or many corporate wide area networks (WANs)—is a “best effort” network. Routers take all the traffic we users generate and try to get it across the bandwidth they have. If too much traffic tries to get through at once, resulting in congestion, routers drop packets to ease the congestion. Applications ignore the packet loss, or request retransmission of the dropped data. Well-behaved applications throttle back their transmissions to try to avoid congestion, but many (especially media-sharing applications) don’t.
For most of the older applications, such as email and Web browsing, and some of the new ones, such as instant messaging, packet loss and variable delivery speed are not much of an issue.
For some applications, though, packet loss and big shifts in throughput are not acceptable. Newer applications, including video streaming, virtual desktops and real-time communications tools such as voice over IP (VoIP) or video conferencing, can suffer severe performance degradation over a congested WAN or network link.
While there are many tools for application delivery optimization (ADO), including application delivery controllers and WAN optimizers, one key type of technology is already embedded in many enterprise networks: QoS (Quality of Service).
What is QoS?
In a nutshell, QoS is the differential treatment of network packets based on some classification scheme. You identify to the switches or router which traffic streams are to be given preferential treatment, and you specify the treatment particular packets are given. For example, a WAN or network engineer might tell a device to deliver all VoIP packets before delivering any non-VoIP traffic; or to reserve a certain amount of bandwidth for each VoIP conversation; or to cap bandwidth for some application at a specific level.
The key mechanism for enabling QoS is the Differentiated Services Code Point (DSCP), a 6-bit piece of the IP packet header, with 3 bits reserved for defining traffic classes and 3 more to define other service markers, so that traffic is to be assured delivery, or given priority on delivery, for example.
What QoS isn’t
It is extremely important to know that QoS isn’t a single thing or a specific technology. It can refer to any technology built into routers or switches that implements flow control, traffic shaping, or queuing in support of differential service on the network.
What is CoS?
You’ll hear about CoS—Class of Service—in the same breath as QoS in many conversations, and for good reason. It is an implementation at the network link layer—Ethernet, for almost everyone—of packet tagging in support of QoS on the LAN. CoS as defined by the IEEE 802.1p standard uses 3 bits in the Ethernet frame header to define a traffic class, allowing eight classes to be defined but with “0” reserved to represent “best effort” default priority. Consequently, switches or other devices can tag up to seven kinds of traffic for special treatment.
WAN gear can pick up CoS tagging and act on it at higher levels (IP, that is) of the network. It can, for example, translate CoS class tags into DSCP class tags.
What WAN engineers need to know about CoS/QoS basics
Although CoS tagging on a LAN is pretty straightforward, with the seven tag values representing a simple hierarchy of priorities from lowest to highest, the 64 classifications available via DSCP tags provide for very complex traffic management schemes—and very little is controlled by standards. So all equipment vendors and carriers can do their own thing! This requires careful attention to detail, either in engineering or in a service-level agreement (SLA) negotiation.
More information on CoS and QoS basics:
- In this tip, learn how to use QoS classification mapping to improve WAN application traffic.
- Understand application delivery optimization and the end user.
- Learn QoS congestion avoidance techniques for VoIP traffic.
- Changes drive need for application SLAs, app-aware WANs and agile QoS.
- What you need to know about meeting application performance and SLAs in the cloud.
This was first published in September 2010