To apply the necessary policy, the SD-WAN must identify incoming applications. Application identification is commonly based on the packet’s port number or protocols. Yet neither are effective in differentiating between Internet service and cloud applications, which often run over common port numbers and protocols. Therefore, the SD-WAN should implement full, seven-layer application identification (or true deep packet inspection [DPI]) but without impacting forwarding performance.

 

Applications network requirements vary. Voice and real-time applications are sensitive to jitter and packet loss. Bulk data transfer requires throughput, making latency and bandwidth important. The SD-WAN should allow you to set policies per application, identifying latency, packet loss, jitter, and bandwidth tolerances.

 

Those considerations must be mapped onto the real-time network conditions of available data services. The SD-WAN should detect the bandwidth as well as the real-time latency, loss, and jitter levels of each connection. While most SD-WANs detect whether a connection is active or has “blacked out,” many cannot detect general slowdowns in performance (“brownouts”). Both detection capabilities are critical.

 

At the same time, applications must be characterized in ways not necessarily related to network performance. General Internet browsing, recreational traffic, and guest traffic, for example, should be able to be blocked or treated as a lower priority. Industry regulations as well as security and privacy laws may require that certain data be carried on a private transport, such as MPLS. More generally, business priorities dictate the importance of a given application and whether it should be restricted to one network or another. The SD-WAN should allow the creation of policies reflecting this business logic.