No Jitter is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Is QoS Becoming Irrelevant?: Page 2 of 2

Continued from Page 1

Where QoS Works

QoS works very well for enterprises that have reasonable control over the endpoints and infrastructure, such as in healthcare. An enterprise may also have mechanisms in place that allow UC endpoints and flows to be easily identified so that the classification and marking processes can work efficiently. It may also be possible to provision sufficient bandwidth inexpensively so that queueing rarely occurs and that any microbursts in data transmissions that arise are short enough so as not to impact UC&C applications.

Private wide-area communications over MPLS supports QoS. That is one of the big selling points of private WAN services. You purchase bandwidth from a WAN provider and then use QoS to prioritize the traffic that must transit the links.

Interestingly, a cloud-based UC controller might work well because the endpoints connect to the UC controller via TCP. Only the endpoint-to-endpoint communications benefit from QoS, and if the endpoints are within the enterprise network, QoS can be beneficial.

Where QoS Doesn't Work Well

QoS doesn't work across the public Internet. That's because QoS requires cooperation between priority levels. Everyone considers their own traffic to be the most important and getting agreement among the participants isn't possible. The result is that QoS markings are ignored in the Internet.

The lack of QoS over the Internet implies problems for cloud-based applications where QoS is desirable. Distributed UC endpoints that rely on Internet access for connectivity (e.g., telecommuters) also can't take advantage of QoS. Conference calling systems, another favorite cloud-based service, are handicapped as well.

Making QoS Useful

One way to make QoS useful is to keep all time-critical applications within the enterprise. But that's often not financially viable.

Most network equipment vendors now have software-defined WAN (SD-WAN) products that measure the transmission characteristics across multiple paths. SD-WAN devices can tell when the Internet path provides a level of service that is suitable for voice. Then, when the Internet latency and packet loss increases (typically on weekday afternoons), traffic can be switched to the more expensive MPLS paths. The neat thing about SD-WAN products is that you can specify that voice traffic only transit links that have certain characteristics (less than 1% packet loss, less than 100ms latency, etc.). Network administrators can define the criteria for prioritizing traffic and links for their mix of applications and link types.

Another approach to QoS is to have application servers communicate with the network's classification and marking system. For example, when a voice call is initiated, the call controller tells the QoS application to do ingress marking on a specific traffic flow. The ingress may be where the endpoint connects to the enterprise infrastructure or it may be where the flow enters the enterprise from and external connection. This type of interaction between the applications and network is rare and only a few networking vendors provide the required functionality.

Conclusion

QoS within the enterprise can be valuable for critical applications as well as voice and video. Cloud-based applications that require communications over the Internet cannot rely on QoS. The things we can do are limited and depend on the organization's infrastructure.

Adding bandwidth is good if links are oversubscribed and it can move the point at which problems occur even if links aren't oversubscribed. SD-WAN is a great solution for optimizing QoS over disparate paths and for inexpensively providing redundant paths. Application and network integration is still in the future so we'll need to rely on other mechanisms.

Time will tell if QoS is really becoming irrelevant.