Overview
A slow experience may originate in DNS, connection setup, transfer or server processing. Identify the phase before choosing a diagnostic method.
Different metrics answer different questions
Bandwidth describes capacity, latency describes travel time, and throughput is the achieved transfer rate. Loss, congestion, connection limits and server performance can reduce throughput below nominal capacity.
Physical proximity is not the whole story
Interconnection, routing policy and load also influence a path. Compare the same regions, time windows and content when evaluating options, rather than treating one speed test as a lasting result.
Troubleshoot in layers
Verify DNS first, then connectivity and TLS, followed by HTTP and application behavior. A hop that does not answer a probe is not by itself proof that it blocks application traffic.
- Record the test region, provider and timestamp
- Compare origin and delivery-entry observations
- Keep successful and failed samples rather than averages alone
Match an improvement to its bottleneck
Caching can reduce repeated origin work, connection reuse can reduce setup, and appropriate asset sizes can reduce transfer. Remeasure after each change to confirm that it improves the intended phase.