How it works
This check sends no requests of its own. Every 15 minutes, it looks at the response times the Uptime check recorded and asks four questions:
- Is the site over budget? The average of the recent responses is compared with your budget.
- Did it suddenly slow down? The recent average is compared with the 2 hours before. A jump by the slowdown ratio means something changed, even if you are still within budget.
- Are there spikes? The 95th percentile catches intermittent slow responses that a good average hides.
- Which phase is slow? Each phase is averaged separately: DNS lookup, TCP connect, TLS handshake, server response (time to first byte) and download.
A problem is only reported once several evaluations in a row agree (2 by default), so one slow window never alerts you.
Settings
Change these on the Settings tab of the uptime report, under Performance notifications.
| Setting | Default | What it does |
|---|---|---|
| Budget ms | 3500 |
Fail when the average response time is over this. |
| Slowdown ratio | 1.5 |
Warn when the site is suddenly this many times slower than the 2 hours before. |
| TTFB ms | 800 |
Warn when the server takes longer than this to send its first byte. |
| Slow checks | 2 |
How many evaluations in a row must agree before we report a problem. |
DNS resolution over 300 ms and TLS handshakes over 500 ms also warn.
Statuses
| Status | Meaning |
|---|---|
| OK | Response times are within budget and stable. |
| Warning | A sudden slowdown, spikes over the budget, or a slow server response, DNS lookup or TLS handshake. |
| Failed | The average response time is over your budget. |
By default you are notified of warnings and failures. Performance notifications can be snoozed together with uptime from the Snooze tab.
Fixing common problems
- Slow server response (TTFB): the application itself is slow. Profile your slowest pages, add database indexes, cache expensive work, and check that the server is not short of CPU or memory.
- Slow DNS: use a fast DNS provider with nameservers close to your visitors.
- Slow TLS handshake: enable TLS 1.3 and session resumption, and make sure the server sends a short certificate chain.
- Sudden slowdown after a deploy: compare the deploy time with the response time chart, then roll back or profile the change.