Quality of service (QoS) is a set of techniques that network devices use to classify traffic and decide which packets go first, which wait and which are dropped when a link is congested. It lets time-sensitive traffic such as VoIP calls and video meetings move ahead of bulk transfers, backups and updates. QoS doesn’t add bandwidth; it shares out the bandwidth you have according to the rules you set, and it only works on network segments where devices are configured to honor those rules.
At a glance
- QoS classifies traffic, marks it, and then queues, shapes or drops it according to policy.
- It matters most when a link is congested; on an uncongested link its effect is small.
- It works on networks you or your provider control, such as your LAN, Wi-Fi, SD-WAN edge and private WAN services; the public internet generally ignores your markings.
- Real-time voice and video benefit most, because they are sensitive to latency, jitter and packet loss.
- Policies need upkeep as applications change and more traffic is encrypted.
What problem it solves
Every link has a limit. When a large upload, a backup or a burst of software updates fills a link, every packet waits its turn, and real-time traffic suffers first: calls break up, meetings freeze. Bulk traffic can tolerate a short delay without anyone noticing; a voice call cannot.
QoS lets you tell the network which traffic matters most. During congestion, voice and video go first, business applications next, and bulk or recreational traffic gets what is left. The result is more consistent quality for the applications that matter, on the same links.
How it works
Classification. Devices identify traffic by address, port, application signature or existing markings. Encrypted traffic is harder to classify in detail, so many devices rely on addresses, destination service names or deep packet inspection signatures.
Marking. Packets are labeled with a priority class, most commonly a DSCP value in the IP header, or an 802.1p value on Ethernet and VLAN traffic. Wi-Fi has its own equivalent that maps to these markings.
Queuing and scheduling. Each outgoing interface keeps several queues. A priority queue for voice is serviced first, usually with a cap so it can’t starve everything else, and other classes share the remaining capacity by weight.
Shaping and policing. Shaping smooths traffic to a set rate so that your own equipment, not the provider’s, decides what waits. This is especially useful on broadband links where upload capacity is limited. Policing drops or re-marks traffic that exceeds a rate.
End-to-end consistency. QoS takes effect only on the devices that apply it, so it matters most where congestion actually happens, usually at the WAN edge, on private WAN links and on Wi-Fi. On a private WAN, the carrier maps your markings to its classes of service under the contract. On the internet, you control only your own edge; SD-WAN devices add value there by shaping traffic at each site and steering it over the link that is performing best.
When it matters for buyers
- Moving to VoIP, UCaaS or a cloud contact center, especially at sites with limited upload capacity.
- Branch networks on broadband, where shaping at the edge can do more for call quality than paying for a faster plan.
- Private WAN contracts, where classes of service, their sizes and what the carrier commits to for each class need to be spelled out.
- Wi-Fi-heavy offices, where wireless contention is often the real bottleneck.
- Recurring complaints about call quality despite apparently adequate bandwidth.
Questions to ask vendors
- Which classes of service do you offer, and what latency, jitter and loss do you commit to for each?
- How do you map our DSCP markings to your classes, and do you re-mark or strip them?
- On internet links, what can your equipment prioritize, and how does it shape upload traffic?
- How does your platform identify encrypted applications such as UCaaS and collaboration tools?
- Will you share per-class statistics so we can see drops and queuing?
- What QoS settings do you recommend for our voice and video platforms?
Our SD-WAN overview explains how application-aware routing and edge shaping work alongside QoS.
How it differs from QoE
QoS is the toolkit and the technical measures: priorities, queues and the network numbers that describe performance. Quality of experience (QoE) is how the result feels to users, often summarized for calls as a mean opinion score (MOS). Good QoS usually improves QoE, but it can’t fix a bad headset, a weak Wi-Fi signal or a slow application. QoS also differs from WAN optimization, which reduces the amount of data sent through compression, caching and protocol tuning rather than deciding which traffic goes first.
Problems this helps with
- Calls get choppy when someone uploads a large file
- Video meetings freeze at busy times of day
- Backups and updates slow down business applications
- Not sure whether our internet provider honors our traffic priorities
