In my last Industry Insight, I touched on sophisticated path selection. I will now go more in-depth into how sophisticated path selection capabilities work.
Next-generation SD-WANs offer more sophisticated path selection capabilities on the edge devices that continuously monitor links, transport paths and application performance on a per traffic-class basis, using real-time data traffic to calculate performance. The common metrics and criteria these SD-WANs use for path detection and switching are packet loss, latency, jitter, MOS, PESQ and hard-down.
Each SD-WAN edge device monitors round-trip delay for a service frame, which includes delay variation, loss ratio (the percentage of service frames that are not delivered), and the availability, as measured by the percentage of time the path was in a connected state. Active monitoring provides sub-second path failover and recovery.
These SD-WAN solutions can also conduct path selection monitoring for SaaS applications, using both active and passive probes. Dynamic traffic engineering and application-specific link selection can be based on:
- Local SD-WAN traffic steering policy configuration.
- Local application QoS configuration.
- Access circuit state and status.
- Information about latency, jitter and packet loss.
By utilising information about latency, jitter and packet loss for non-VPN sites, like SaaS, and other sites, such as YouTube and Netflix, over various access circuits, each branch device builds a database with key traffic engineering information.
Paths through which SLA responses are not received are considered to be path-down and are made non-available for SD-WAN forwarding. The edge network reacts in real-time, based on the defined SLA requirements of the applications.
An advanced SD-WAN will include voice and video codecs to analyse the real user experience of each voice and video session.
An advanced SD-WAN will also include voice and video codecs to analyse the real user experience of each voice and video session, and supports RTP- and SRTP-based voice and video applications; that class of information provides ongoing database updates of application identification signatures and codecs.
A composite path selection score is used, that takes into consideration TCP parameters, MOS-like scoring, round-trip-time, round-trip-delay, jitter, delay, loss and application performance metrics. Machine learning-based scoring for application policies can also be applied. In doing so, the SD-WAN will learn the network characterises and anomalies, and continuously optimise the path selection capabilities.
Progressive SD-WAN solutions:
- Have self-healing features, and architectures that virtualise edge networking and security functions within the enterprise WAN and multi-cloud networks.
- Are application-aware and user-experience-driven monitoring applications.
- Optimise traffic delivery, and provide a robust security posture.
Key capabilities of advanced SD-WAN solutions with sophisticated path selection include:
SLA monitoring: SD-WAN branch devices continuously monitor the performance of all paths. A branch-to-branch path is any valid transport tunnel between the two branches. For example, if two branches have two broadband links each, and the branches are in a single transport domain, there are four paths between those branches.
Adaptive monitoring: When SLA monitoring is configured on a WAN interface, the monitoring of paths to every neighbour link learned through multiprotocol BGP starts automatically. In the case of a full-mesh topology with numerous branches, such monitoring can result in a large amount of SLA traffic. To reduce the amount of SLA traffic monitoring on the network, adaptive monitoring will perform SLA monitoring only to neighbours that are actively passing traffic.
Data-driven SLA monitoring: Data-driven SLA monitoring is an extension of adaptive SLA monitoring that regulates the amount of traffic monitoring between branches. It accomplishes this by creating and deleting SLA-monitoring contexts, based on whether traffic is flowing toward a remote site.
Alternate path: When data-driven SLA monitoring is enabled, an alternate path to a destination branch is specified, accomplished by assigning one branch device as a hub that forwards traffic between branches. While a new SLA monitoring context is being created on the direct path between two branches, the alternate path is used to send the initial packets of a flow towards the destination branch.
Replication: Packet replication improves the quality of voice traffic, and other mission-critical application traffic. The SD-WAN nodes mirror packets among two or more paths. If a packet is lost on one link, the mirrored packet is delivered on secondary links. If the remote device receives more than one copy of the packet, it sends the first received packet toward the LAN, and drops subsequent packets.
Forward error correction: Forward error correction (FEC) controls errors in data transmission that occur over unreliable or noisy communication channels. The sender encodes the message using an error-correcting code and does so in a redundant manner. The redundancy allows the receiver to correct errors without having to request, over a reverse channel, that the sender retransmit the lost data. For FEC to work, the sender generates an FEC parity packet for every N packet it sends. On the sender, administrators configure the frequency at which FEC parity packets are generated. The receiver uses this parity packet to recover any lost packets. In this way, FEC minimises packet loss at the receiving end, improving the end-user’s quality of experience.
Leveraging SD-WAN solutions that use dynamic path selection policies to optimise how traffic moves between branches, data centres and clouds is critical to accelerating digital transformation strategies and migration to multi-cloud and public Internet transport mechanisms.
Rather than reacting to network problems, SD-WAN proactively monitors and automatically selects the best path based on business policies to maximise availability, reliability and performance.
Share