Performance Intelligence
How a route is selected, step by step.
Route quality is compared across the available paths, and performance sensitive traffic uses the option that best meets current conditions. This walkthrough shows what is measured, how the decision is made, and what changes when conditions upstream change.
The walkthrough
Three things decide which path a session uses.
Each step is stated separately so the decision can be followed without needing to read a routing table.
How route selection works
Step 1
What is measured
Round trip time, jitter, packet loss and route stability are observed continuously on each candidate path, rather than sampled once at turn up.
Step 2
How a route is selected
For performance sensitive traffic, the path that best meets current conditions is used. Standard Internet transit is always one of the candidates, and it is used when it is performing best.
Step 3
What happens when conditions change
Conditions upstream change through the day. When a candidate path degrades, preference moves to a better performing option instead of waiting for a support ticket to reveal the problem.
Interactive
Change the conditions and watch the selection move.
Standard Internet transit stays a candidate at all times. No managed path is forced when standard transit is performing better.
Route selection scenarios
Select a condition
How selection works: route quality is compared continuously across the candidate paths, including standard Internet transit. Traffic is not moved onto a managed path when standard transit is performing better.
Standard Internet transit
78 ms
loss 0.6% / jitter 14 ms
Measured path A / selected
44 ms
loss 0.1% / jitter 4 ms
Measured path B
52 ms
loss 0.2% / jitter 6 ms
Standard transit becomes less consistent at peak, so performance sensitive traffic uses the path that best meets current conditions. Customers notice fewer stalls in interactive sessions.
Illustrative path behaviour. The values are examples used to explain how a route is selected, not measured results on any network.
Why loss is expensive
A lost packet costs a round trip, not a few bytes.
In a synchronous exchange the sender waits, resends, and everything behind it waits with it. That delay is what the customer perceives.
What a lost packet costs
What a lost packet costs a synchronous exchange
Why the wait happens: the sender only learns a packet is missing when the response does not arrive, so it has to send that packet again. Everything queued behind it waits until the gap is filled.
What the customer feels: a pause in a voice reply, a clipped word, or a session that stalls part way through a task. In a synchronous exchange the cost of loss is the recovery delay, not the missing bytes.
Illustrative timeline, not to scale and not a customer performance claim.
How it is reported
Network conditions are reported with the effect they relate to.
A change in conditions can then be read against a change in experience rather than in isolation.
Reported measurement pairs
- Round trip timePerceived response time
- Packet lossSession stability
- JitterAudio and video intelligibility
- Route stabilityCompletion consistency
Next step
See the same decision applied to a route you care about.
An operator evaluation baselines current conditions on a nominated route and measures the customer visible effect before anything is offered to a customer.