Listen to post:
|
As I discussed in my previous post, real-time traffic has two characteristics that are challenging for the Internet. First, as the packets have a limited time value and cannot be re-transmitted, any significant change in the transport and packet delivery has the potential of being audible (or visible in the case of video) to the participants. And, as most real-time conversations last orders of magnitude longer than most other types of internet interactions, the probability of a network incident impacting the packet transmission is dramatically higher. The result is that real-time traffic needs a deterministic transport with minimal latency.
The challenge is that in the network world getting Service Level Agreement (SLA) determinism generally has a steep price. Whether a dedicated wire or MPLS, the cost of traditional WAN technology increases in direct relationship to determinism. SD-WAN solves this very problem by utilizing route and component diversity made feasible by the improvements in technology and the affordable costs of Internet bandwidth.
The basic concept of SD-WAN is the concept of route diversification. The two SD-WAN edge points (the point between the Enterprise LAN and the carrier WAN) create multiple route paths between them. For example, in the diagram, each of the red paths represents a different route between the SD-WAN node on the left and right. When the actual traffic arrives at the SD-WAN node, it can decide, based on a factor such as traffic type or current route performance, which route to put the actual data packets into. All of this can be controlled by the SD-WAN controller that oversees the operation.
While the diagram shows a simple premises SD-WAN, the addition of Points of Presence (POP) in a core cloud SD-WAN enables management of the paths between the POPs. This can enable enhanced determinism as much of the variation in Internet traffic delivery happens in the core that is bypassed by having a cloud core. We will discuss this specific topic in a future post.
The benefits to real-time traffic are clear. In the traditional network, if the path that is currently being used for the real-team session flow is impacted, whether through failures or peering issues that limit capacity, the user traffic will stay in that route and the quality of the real-time interaction traffic may be compromised. In the SD-WAN, the traffic can be dynamically moved from the impacted route to the best route available at that time. Through this mechanism, an SD-WAN has the potential of using the best possible route at any point in time between two locations on the Internet, all of the while using the lower cost service of the open Internet, assuming there are sufficient paths for route diversity.
The result is that SD-WAN changes the determinism and cost model of the modern WAN. because of route diversity and path management, SD-WAN enables the MPLS equivalent determinism required by real-time traffic at close to the open Internet cost model. At the core, the concept is simple, but there are many layers of complexity and value that must be considered as part of a well-engineered SD-WAN solution. For example, the routes must be monitored for their current transport characteristics, the traffic type of flows must be determined, the different flows and their relative policies must be included, and more. All of these are critical for VoIP and other real-time traffic.
In considering an SD-WAN solution, there are a number of factors that should be evaluated if optimizing real-time traffic. Whether the SD-WAN is implemented as a premise or cloud solution is a consideration. If backhaul is required and the use of Points of Presence can also have an impact. Also, how the SD-WAN classifies the traffic — this too can have a major impact on real-time determinism. Other considerations like cloud Software as a Service (SaaS) access and security are important. Over the next few months, we will both discuss how to use SD-WAN, but also some of those key characteristics and capabilities that an SD-WAN solution must have to maximize value to real-time traffic.