Myth-busting SD-WAN: Part 2 — Does path selection contribute to QoS?

Ricardo Belmar
May. 31 2018

Welcome to Part 2 of our “Myth-busting SD-WAN” series. Last time, we explored QoS in general and discussed the limitations of FEC in the context of delivering the true end goal for QoS, that is, enhanced QoE for users and key applications. In this edition, we turn towards the issue of path selection, which is one of the three key techniques employed by SD-WAN solutions to provide QoS and control application performance.

Performance Myth #2 — Path Selection

So, what is path selection? Path selection is the ability to route traffic between two or more links, making a choice about which to choose.  That sounds a useful and practical idea - and it is a great solution - but is it QoS? In short “no”.  In fact, it is really leveraging the underlying transport to deliver the performance and hoping that one of the links at the branch site can properly support the application in question.  Hope is neither a tactic nor is it QoS.

Deeper investigation makes it clear that not all path selection solutions are created equal. Do they merely offer failover, in the event of link loss, or does the application switch links due to a ‘brownout' condition (i.e. a link is still up, but has deteriorated)?  If the former, then this is just a high availability solution, but it is not a solution that will proactively improve performance (let's be clear - performance and availability are two different things, a topic we'll discuss in a future post).  If it's the latter, then we need very precise criteria to allow us to understand applications at a very detailed level and the performance characteristics that are needed.

Most solutions make path selection decisions based on two factors - packet loss and latency. As a result, they understand applications at a high level (for example, VoIP vs. a specific codec; or Skype for Business vs. messaging, audio, video, screen sharing, etc., within the Skype for Business application flow), but that is inadequate and doesn't take account of the granularity that users require to really differentiate between the performance demands of individual applications that are competing for the same bandwidth resources.

This means that many solutions inevitably over-allocate resources (unnecessarily push traffic to a high-performing link) or under-allocate resources (route traffic on a low-performing link), or both. Neither is an optimal outcome.  In addition, most solutions route traffic between multiple links at a port level or by bundling together many sessions of a specific application, without discriminating between them. This is inefficient and does not optimally allocate bandwidth between links or consider what outcomes are really desired.

A better approach…

A better, more efficient approach that can deliver a more predictable outcome would be to use a broader range of metrics for overall path selection decision criteria, while also understanding application performance needs at the most granular level (what does the application really need?).  With this model, individual applications can be identified and differentiation according to the needs of each enterprise can be applied, according to the policies they choose.

This is the model employed by InfoVista. Our solution has the ability to route different sessions from the same application to different paths, for optimal efficiency. This provides the ability not only to deliver more traffic in total, but also via the optimum path, ensuring higher performance levels.

As a result, InfoVista makes detailed and accurate decisions to route traffic on the right WAN path so that the application is neither given more performance allowances than it requires, nor less than it needs.

However, even though the InfoVista solution offers superior performance, based on policies and active path selection according to the needs of different applications, this still isn't true QoS control, as it depends on the underlying network to deliver the performance.

Stay tuned for more…

Read Part 3 to learn about the myth around prioritization and how that fits into delivering QoS! Plus, we'll discuss how Infovista uniquely delivers true QoS control!

Looking for more SD-WAN myths to bust? Watch our on-demand webinar, where Zeus Kerravala, founder of ZK Research, joined us to bust the Top 7 SD-WAN Myths!

Written By