Your users are physically close to your edge nodes — maybe 20-50ms away. Your origin servers are 150-300ms away. The whole point of edge infrastructure is to bridge that gap. But how do you know if your edge layer is actually helping? That's where latency budgets come in.
What is a Latency Budget?
A latency budget is the maximum acceptable time for a request to complete from the user's perspective. If your target is "under 200ms," that's your budget. Every millisecond your edge layer adds is a millisecond subtracted from your application's budget.
Breaking It Down
A typical edge request involves: DNS resolution, TLS handshake, edge processing (cache lookup, routing), origin fetch (if cache miss), and response delivery. Each of these stages consumes part of your budget. Monitoring each independently is the only way to know where your latency is actually going.
Setting Realistic Targets
Don't set a single global latency target. Your users in Tokyo and your users in London have different network paths and different expectations. Set per-region targets based on actual network conditions, and measure against those — not against some idealized global number.
At EdgePulse, we built our monitoring around this principle: every check measures the full path, and every alert is evaluated against the appropriate regional baseline. It's the only way to get signal without noise.