We sit in your request path, so the honest questions have honest answers. Here they are before you have to ask them in a meeting.
The two you'd ask us. Answered before you have to.
We sit in your request path, so this is the right question. You choose what happens: fail open and traffic reaches your backend unfiltered, or fail closed and it stops at the edge. Per node, your call — because whether an outage should mean "let everything through" or "let nothing through" depends on what you're protecting.
Yes. TLS terminates at the edge, which is what lets us count and limit requests at all — so under GDPR we're your processor and you're the controller. That comes with a DPA, a named list of sub-processors, and Article 28 terms. Ask and we'll send them before you send us a single request.
We're not a CDN and we won't absorb a terabit DDoS. We rate-limit and meter your API at a European edge. If you need scrubbing at that scale, you need something else in front — and we'd rather say so here than in month three.
Every customer gets their own edge node and their own database file. Not a schema, not a row filter — a separate file. There is no query anywhere in the system that could return another customer's rows, because there is no shared table to query. It costs us a little more to run and it removes an entire class of mistake.
| What | Kept |
|---|---|
| Individual requests | A few days to three weeks, depending on your plan and traffic. Never less than one hour. |
| Hour-by-hour breakdown | 3 days — by endpoint, country and status code. |
| Daily totals | 90 days — requests, errors, bytes and latency percentiles. |
| Your configuration | Kept with full version history for as long as your account is open. |
Reports outlive the records behind them on purpose: a 90-day chart is served from daily totals, not from three months of individual requests sitting in a database. Less of your data kept, for longer usefulness.
DPA, sub-processor list and Article 28 terms are ready to send. Ask before you send us a single request.