Network QoS

In a multi-site enterprise network, connection between the sites typically done by using WAN connection services either through a secure tunneling network via public network (the internet). There are many WAN technologies can be used for WAN link connection between the sites. Network Quality of Service (Network QoS) should be used on WAN links where congestion is a concern and/or preferential service of selected network application traffic is required.

Network QoS techniques can be used to prioritize mission critical traffic and provide a defined level of service for latency sensitive interactive traffic. This is a requirement for the deployment of real-time network applications such as Video Conferencing (VC) and Voice over IP (VoIP). Network QoS does not provide a solution for over-utilized WAN links that are continuously saturated.

In multi-site business networks, the use of network QoS is optional depends the requirements each of the sites, but is recommended to be used on slow WAN links that experience congestion. If network congestion is an ongoing problem then the bandwidth of the link should be upgraded.

If you need to use the network QoS for the WAN connection for each of site to the headquarter office, you can follow the recommended requirement with regards to network QoS below:

• If deploying interactive real-time applications such as VoIP then a suitable network QoS policy must be adopted. Network QoS can be done at the router or on a mission specific packet shaping appliance. For small business network or SOHO, there are many types of wireless routers that have the network QoS feature which are typically for voice or video streaming applications such as WRT610N wireless router, or DGL-4500 wireless router which is designed specifically for gamers in mind.
• The policy must classify traffic into distinct channels and guarantee a defined level of service for each channel. Any real-time interactive or latency sensitive traffic must be prioritized ahead of other non-critical traffic.
• The policy should take into account packet serialization delays and allow for packet interleaving for latency-sensitive application traffic.
• Network QoS deployment should not involve the changing of MTU sizes (application incompatibility problems) and should not starve lower priority services.

Classification of Network QoS

A successful network QoS policy requires classification and enforcement. Classification is the process of identifying traffic to be marked; this can be done based on packet type, size or address. This is done in order to separate “real-time critical” vs. “non-critical” traffic. Once network traffic has been classified it is then placed into a queue awaiting transmission. The enforcement of the network QoS policy defines which queues require priority service and how much data should be sent per process cycle of each queue. Enforcement of the network QoS policy only occurs when congestion is occurring on the network.

Note: If using payload encryption across the network, then the data must be pre-classified prior to encryption

Another important aspect of any network QoS strategy is to address packet serialization delays. Serializations delays occur because it takes time to transmit a packet across a slow WAN link (and waiting for large packets to complete transmission), which introduces latency delays into the network. Interleaving through the use of link fragmentation allows larger packets to be split into small fragments for transmission across the network QoS enabled link. This method of dynamic packet splitting and interleaving allows time-sensitive packets to access the link (via insertion) in a more timely fashion.

The aim of any network QoS policy should be to ensure that mission-critical or interactive traffic has a guaranteed minimum level of service (latency, jitter and bandwidth) in which to support real-time applications across the WAN. Network QoS methodology is described in RFC 2212 and 2290.

Without a network QoS strategy enabled on WAN links, service levels agreements (SLA) cannot be guaranteed for mission critical applications. Interactive applications will suffer extensively from jitter and latency delays, making their deployment unsuccessful.

Without network QoS a large file transfer or email attachment could starve network bandwidth from real-time mission critical applications and services. The network QoS prioritization must be carefully balanced in such a way to meet current and future SLA requirements and provide packet classification for all key network services.

Reference: Cisco.Com

See also:

• ISDN E1 and T1 definition
• WAN Definition
• PPP protocol