QoS parameters

Organizations can measure QoS quantitatively by using several parameters, including the following:

• Packet loss happens when network links become congested and routers and switches start dropping packets. When packets are dropped during real-time communication, such as a voice or video calls, these sessions can experience jitter and gaps in speech.
• Jitter is the result of network congestion, timing drift and route changes. Too much jitter can degrade the quality of voice and video communication.
• Latency is the time it takes a packet to travel from its source to its destination. Latency should be as close to zero as possible. If a voice over IP call has a high amount of latency, it can experience echo and overlapping audio.
• Bandwidth is the capacity of a network communications link to transmit the maximum amount of data from one point to another in a given amount of time. QoS optimizes the network by managing bandwidth and setting priorities for applications that require more resources than others.
• Mean opinion score (MOS) is a metric to rate voice quality that uses a five-point scale, with a five indicating the highest quality.

Implementing QoS

Three models exist to implement QoS: Best Effort, Integrated Services and Differentiated Services.

Best Effort is a QoS model where all the packets receive the same priority and there is no guaranteed delivery of packets. Best Effort is applied when networks have not configured QoS policies or when the infrastructure does not support QoS.

Integrated Services (IntServ) is a QoS model that reserves bandwidth along a specific path on the network. Applications ask the network for resource reservation, and network devices monitor the flow of packets to make sure network resources can accept the packets.

Implementing IntServ requires IntServ-capable routers and uses the Resource Reservation Protocol (RSVP) for network resource reservation. IntServ has limited scalability and high consumption of network resources.

Differentiated Services (DiffServ) is a QoS model where network elements, such as routers and switches, are configured to service multiple classes of traffic with different priorities. Network traffic must be divided into classes based on a company’s requirements.

For example, voice traffic can be assigned a higher priority than other types of traffic. Packets are assigned priorities using Differentiated Services Code Point (DSCP) for classification. DiffServ also uses per-hop behavior to apply QoS techniques, such as queuing and prioritization, to packets.

Network architecture also affects how an organization implements QoS. A Multiprotocol Label Switching (MPLS) network includes a private link that offers end-to-end QoS along a single path. SLAs for MPLS specify bandwidth, QoS, latency and uptime. However, an MPLS can be expensive for organizations.

Software-defined WAN (SD-WAN) uses multiple connectivity types, including MPLS and broadband. SD-WAN monitors the state of current network connections for performance issues and uses its multiple connectivity types to fail over based on state. For example, if packet loss exceeds a certain level on one connection, SD-WAN capabilities will look for an alternative connection.

QoS mechanisms

Certain QoS mechanisms can manage data traffic quality and maintain the QoS requirements specified in SLAs. QoS mechanisms fall under specific categories depending on the roles they play in managing the network.

• Classification and marking tools differentiate between applications and sort packets into different traffic types. Marking will mark each packet as a member of a network class, which allows devices on the network to recognize the packet’s class. Classification and marking are implemented on network devices such as routers, switches and access points.
• Congestion management tools use packet classification and marking to determine which queue to place the packets in. Congestion management tools include priority queuing; first-in, first-out; and low-latency queuing.
• Congestion avoidance tools monitor network traffic for congestion and will drop low-priority packets when congestion occurs. Congestion avoidance tools include weighted random early detection and random early detection.
• Shaping tools manipulate traffic entering the network and prioritize real-time applications over less time-sensitive applications such as email and messaging. Traffic shaping tools include buffers, Generic Traffic Shaping and Frame Relay Traffic Shaping. Similar to shaping, traffic policing tools focus on throttling excess traffic and dropping packets.
• Link efficiency tools maximize bandwidth use and reduce delay for packets accessing the network. While not exclusively for QoS, link efficiency tools are used in conjunction with other QoS mechanisms. Link efficiency tools include Real-Time Transport Protocol header compression, Transmission Control Protocol header compression and link compression.