What is Uptime?
Uptime is the share of a measured period during which a service is reachable and answering correctly, written as a percentage such as 99.9%. Providers report it monthly or quarterly from their own probes. The complement, downtime, is easier to reason about, since 95% uptime allows roughly 36 hours of outage across a 30-day month.
Availability measurement starts with a probe. An external monitor opens a connection to the endpoint on a schedule and records whether a known request came back inside the timeout. Probe interval sets the resolution, so a monitor checking every five minutes cannot see a 90-second outage, and figures from coarse monitoring run high. Proxy networks complicate the picture because there is no single thing to poll. A gateway listening on a typical HTTP proxy port such as 8080 can accept connections normally while the modem behind your assigned endpoint has dropped off its carrier.
The arithmetic behind the nines is short. Across a 30-day month of 720 hours, 99% permits about 7 hours of downtime, 99.9% permits roughly 43 minutes, and 99.99% permits a little over 4 minutes. Each extra nine costs an order of magnitude more engineering, which is why few proxy vendors claim more than three. Read the averaging period as carefully as the number. A percentage averaged over a year absorbs one catastrophic day that would have broken a monthly target outright.
Providers and customers often disagree about what counts as an outage. A target site returning 403 to your requests is not proxy downtime, even though the job fails either way; that belongs to success rate. A handful of dead exit IPs inside a large pool usually does not count either. Contracts tend to define downtime as the gateway refusing connections or failing to route traffic for a sustained window, measured by the provider's own monitoring, with announced maintenance excluded. Establish which clock is authoritative before you sign anything.
Run your own check instead of trusting a status page. A small job that fetches a neutral IP-echo endpoint through each configured proxy every minute and alerts after several consecutive failures will catch trouble long before a vendor dashboard changes colour. Store the failure class next to the count, because a timeout and a refused connection point at different problems. PROXIES.SX publishes 95% uptime, and the same discipline applies to any provider. During an incident the only numbers you can act on are the ones you collected yourself.
Where you meet it
Uptime comes up in two places. On paper, while comparing providers, it is a question about definitions and remedies. At 3am, when a scheduled crawl returns nothing, it becomes an operational question, answered by whether your monitoring can tell a dead endpoint from a destination that started blocking you. Build that distinction into retry logic early, because a client hammering a dead gateway drains the queue and teaches you nothing.
Common questions
Is 95% uptime good for a proxy network?
It depends on the workload and on how the provider defines an outage. In the worst case 95% allows about 36 hours of downtime in a month, though real networks usually degrade partially rather than failing whole. Batch scraping with retries and loose deadlines absorbs that; synchronous, user-facing traffic needs failover regardless of the headline figure.
How is uptime different from success rate?
Uptime measures whether the proxy service accepts your connection and routes it. Success rate measures whether the destination returned the content you asked for. A network can sit at full availability while your success rate against a hardened target runs at 40%, because blocks and CAPTCHAs happen after the proxy has done its job.
Does downtime include scheduled maintenance?
Usually not. Most agreements exclude announced maintenance windows, force majeure, upstream networks the provider does not operate, and failures caused by customer configuration. Check whether maintenance has a cap, such as a maximum number of hours per month, and whether the provider commits to notifying you in advance.
Related terms
Real 4G/5G mobile and residential IPs
PROXIES.SX runs carrier IPs in 100+ countries with HTTP and SOCKS5 on every endpoint. $4/GB down to $2.40/GB at volume, free endpoints and rotation, and your GB never expire.