Filed under: Industry Insights
ECMP is a routing technique that allows traffic sharing the same source and destination to be distributed through multiple network paths of equal cost. This involves leveraging routing metric calculations with hash algorithms to determine which available “next-hop” routers can be leveraged at the same price point to deliver traffic from one flow to its destination.
In simplest terms, ECMP is a load-balancing method that prevents bottlenecking at a single router while often increasing overall traffic speeds (without blowing up IT budgets). The ideal result is lower latency and increased resiliency at no cost to the network: If one service delivery component is degraded, the other elements of the same service are not impacted, allowing the service to remain operable.
Here’s how it looks in action.
Let’s say your business’ peak need for bandwidth is 500 Mbps for a single office. At the end of the day, employing five 100 Mbps circuits is a fundamentally stronger architecture than leveraging just one 500 Mbps circuit. This model makes it so that when one circuit is experiencing issues (which is inevitable), the other four circuits can operate normally. Essentially, by using five paths instead of one, end users may only experience a slight brownout rather than a total blackout when a single circuit goes down. (In fact, with proper planning, users may not even notice any interruption).
The challenge with managing and monitoring ECMP comes as a side effect of resiliency. By leveraging multiple, independent circuits, your standard legacy monitoring may only be looking at one network path/flow and not the complete picture. By taking multiple network circuits and pseudo-randomly distributing load across them, one user’s experience may be severely degraded while others are completely unaffected. One connection running well may mask another that’s having performance issues similar to how DNS clusters can often mask an underperforming server.
As with any approach to networking, the only way to guarantee success is by having as much visibility into the network as possible. In this case, teams need to have hop-by-hop visibility across all delivery paths at the same time.
How AppNeta tackles ECMP monitoring
As part of AppNeta’s active network monitoring, we perform network route discovery with TCP, ICMP and UDP to discover all possible network paths that a service or app may traverse. This approach can be applied to SD-WAN environments, where choices made by the SD-WAN device may otherwise obscure poor performance of the underlay networks.
Active, continuous testing of each component of ECMP or SD-WAN circuit ensures that your resilient architecture actually has the resources expected, and not a situation where your strong design is masking underperformance by your providers.
To learn about our unique, lightweight approach to monitoring network traffic from all sides, download our whitepaper, Four Dimensions of Network Performance Monitoring.