Three ways to rotate#
| Method | Where it happens | Typical use |
|---|---|---|
| Per connection | At the gateway: every new TCP connection gets a new exit | Scraping, SERP, price checks |
| Timed (sticky) | At the gateway: the same exit is kept for N minutes per session id, then replaced | Multi-step flows, carts, logins |
| Client-side | In your code, cycling through a list of static addresses | Datacenter or ISP lists |
Rotation and rate limits#
A target that allows 60 requests a minute per IP allows 6,000 a minute across 100 exits. Rotation converts a per-IP ceiling into a per-target budget you set yourself. It does not remove the target's global defences: unusual patterns from many IPs (identical fingerprints, perfect timing) are still detectable.
Common mistakes#
- Rotating in the middle of a session that the site tracks by IP: log in on one exit, get logged out on the next. Use a sticky session for those steps.
- Expecting a new IP per request while the HTTP client reuses connections. Disable keep-alive or close the connection.
- Rotating across countries by accident: pin -cc- (and -city- when relevant) so that every exit stays in the right market.
How it works at ProxShift#
On ProxShift residential and mobile gateways, rotation is per connection by default; -sid- with -ttl- (1 minute to 24 hours) turns it into a timed sticky session. Dedicated mobile devices rotate on demand or on a timer you set.
Questions people ask#
Is rotation free?
On pool networks it costs nothing extra: you pay for gigabytes, not for the number of exits used. On static lists you pay per address, so the rotation budget is the number of IPs you rent.