Uptime
The share of time a proxy is actually reachable and working, which for residential pools is a live measurement rather than a guarantee.
Uptime is the share of time something actually answers, and the word means two genuinely different things depending on which family you are buying. Treating them as one number is the source of most disappointment with the metric.
On fixed infrastructure it behaves like an ordinary server statistic. A datacenter proxy or an ISP proxy is a machine in a rack with redundant power and network, so it is up almost always, the failures are infrastructural, and a published figure is a reasonable promise somebody can be held to.
On a residential pool it is not a promise at all, because the addresses belong to people. A household connection goes away when somebody reboots a router, unplugs for the night, changes provider or simply leaves. No provider can prevent that, and any provider claiming a fixed high uptime for individual residential addresses is describing something they do not control.
So residential availability is a distribution rather than a number, and the honest way to express it is how many addresses are confirmed alive right now in the country you need. A single headline percentage collapses a moving population into a figure that was true at some unstated moment.
Our own free-pool measurements show what that looks like at the extreme. As of 11 August 2026, of 589,918 free proxies discovered and tested, 4,023 answered. The median lifespan of the ones that ever worked was about 145 hours, roughly six days. Free supply is the worst case and paid residential is far steadier, but the underlying mechanic is the same: borrowed connections have lifespans, and a number printed once decays from the moment it is printed.
Which makes the useful question about process rather than about the figure. Not what uptime a provider advertises, but how recently the specific addresses you are about to use were confirmed working, and what happens when one dies mid-session. A pool that is re-checked continuously and one that quotes an impressive number annually can advertise the same percentage and behave nothing alike.
It is also worth separating uptime from success rate, since they are measured against different things. An address can be perfectly up, answering every connection, and still fail your job because the target rejects it. Uptime is about reachability; success rate is about outcomes, and only the second one is what you were actually buying.
How HProxy handles it
We re-check endpoints continuously precisely because residential uptime is a live fact rather than a printed one. The figure worth reporting is how many addresses are confirmed working right now, which is why we publish that and not a flattering annual percentage.
Frequently asked questions
What uptime should I expect from residential proxies?
Expect it as a pool property, not a per-address one. Individual residential addresses go offline constantly because they belong to households, so the meaningful figure is how many distinct addresses are live in your target country at the moment you run, and how quickly the pool replaces the ones that drop. A provider quoting a fixed high uptime for individual residential IPs is describing something outside their control.
Why did my proxy stop working mid-session?
On residential supply, most often because the underlying household connection ended: a router reboot, a device sleeping, somebody leaving the network. A sticky session is a request to hold an address, not a guarantee that the address survives, so code that assumes the exit persists for the full window will eventually be wrong. Handle a dropped exit as an expected event rather than an error.
Is uptime the same as success rate?
No, and confusing them hides real problems. Uptime asks whether the proxy answered you. Success rate asks whether the request achieved what you wanted at the destination. An address can be up on every attempt and fail every job because the target rejects it, which shows as perfect uptime and useless results.
How do I measure proxy uptime myself?
Check repeatedly against a target you control and record every outcome, rather than sampling once and generalising. Two things matter beyond the percentage: how the failures are distributed, since a handful of long outages is a very different problem from constant brief ones, and whether the address that failed is the same one each time.
Back to the full glossary.