Types of Proxies: Datacenter, ISP, Resi & Mobile

‘Proxy Type’ Is Four Questions Wearing One Name — This Page Answers the IP-Origin One

Every proxy has a value on four independent properties at once, and the word “type” is used for all four interchangeably. That is why a buyer can ask for one thing, be sold something that honestly matches the label they used, and still end up with an address that fails on the property they actually cared about.

The four questions, and where each one is settled:

• Origin — who the exit address is registered to: a hosting company, a consumer ISP, a household, or a mobile carrier. This page answers that one, and nothing else.
• Protocol — how your client talks to the proxy: HTTP/HTTPS, or SOCKS5 as defined in RFC 1928, which conventionally listens on TCP port 1080. Settled on HTTP vs SOCKS5 proxies.
• Header behaviour — transparent, anonymous or elite, meaning what the proxy does with the Forwarded header standardised in RFC 7239 and the older X-Forwarded-For it documents. That, and the forward-versus-reverse distinction, are settled on what a proxy server is.
• Connection behaviour — rotating, dedicated, static, shared or backconnect. Settled on the proxy types hub, which sorts mobile proxies by how the address behaves rather than where it came from.

Those are four axes, not four categories on one list — a single proxy sits on all of them simultaneously. A mobile IP can be delivered over SOCKS5 or HTTP, held dedicated or rotated on a timer, and it is a mobile IP throughout. Origin is the axis worth learning first for one reason only: it is the single property you can verify yourself, from outside, without trusting anything the seller wrote on the product page.

The Four Origins Side by Side — and How to Check Which One You Actually Bought

Most comparisons of these four sort them by price and by how much a target trusts them. Both are real, and both are somebody else’s judgement about the address. What follows sorts them by provenance instead — facts recorded in public registries that you can look up for any address in about a minute.

Deliberately absent from that table: cost, what each tier is best for, and where each one fails. Those are a different comparison with a different answer, and they are worked through on choosing a proxy tier for web scraping and, for the two tiers furthest apart, on datacenter vs mobile proxies. Read this table for what an address is; read those for what it is worth.

The Last Column Is Not an Opinion — It Is a Database Field

The reason this taxonomy is worth trusting is that it is not a marketing invention. Commercial IP-intelligence databases ship it as a typed field. MaxMind’s Enterprise database exposes a user_type whose documented values include business, cafe, cellular, college, consumer privacy network, content delivery network, government, hosting, library, military, residential, router, school, search engine spider and traveler — and a separate connection_type that is one of Cable/DSL, Cellular, Corporate or Satellite. Hosting, residential and cellular are three of the four origins on this page, sitting in a fixed enumeration that anti-bot vendors, ad platforms and fraud systems query millions of times a day.

That is what makes the origin axis checkable rather than rhetorical. When a target classifies your traffic, it is reading fields like these. So can you.

Four Lookups That Prove What You Were Sold

1. Get the exit address

   Send one request through the proxy to any IP-echo endpoint and record what comes back. Everything else keys off this address, not off the hostname you connect to — a backconnect gateway’s own address tells you nothing about the exit.

2. Read the announcing ASN and its organisation type

   Whoever announces the route into the global table is the network the traffic really leaves from. A hosting AS name here settles the question for lanes one and two of the diagram above, regardless of what the WHOIS record says.

3. Read the registry’s own organisation record

   Query the RIR for the block and look at the organisation and netname fields. This is the residential registration an ISP proxy is bought for — and where a datacenter address advertised as residential gives itself away immediately.

4. Read an IP-intelligence classification and the reverse DNS

   The classification field is what your target most likely reads. Reverse DNS is the free corroboration: hosting racks name themselves in structured, sequential patterns, while household and carrier addresses usually resolve to a generic operator pattern or to nothing at all.

# 1 — what address does the proxy actually exit from?
curl -s --proxy http://user:pass@proxy.example.net:8000 https://api.ipify.org
# 203.0.113.45

# 2 — which ASN announces it, and who owns that ASN?
whois -h whois.cymru.com " -v 203.0.113.45"
# AS     | IP            | AS Name
# 64500  | 203.0.113.45  | EXAMPLE-HOSTING, US

# 3 — what does the registry say the block is assigned to?
whois 203.0.113.45 | grep -Ei "^(org-name|OrgName|netname|descr)"

# 4 — reverse DNS: racks name themselves, homes rarely do
dig +short -x 203.0.113.45

If you would rather not run the commands, the PXM2 proxy checker returns the exit address and its ISP readout for a proxy you paste in, which covers steps 1 and 3 in a browser.

Datacenter Proxies: Addresses a Hosting Company Owns Outright

A datacenter proxy is not defined by being fast, and not by living on a server. It is defined by one thing: the organisation named on the registry record is a hosting or cloud company, and that same company announces the route. Registration and network agree. Nothing is hidden, because nothing is meant to be.

That agreement is why the tier is abundant. A hosting company can acquire contiguous blocks on the open address market and put them to work immediately, so supply scales with demand rather than with the number of real households or subscribers involved — because no households or subscribers are involved at all.

It is also why identification happens in bulk. Addresses arrive in contiguous blocks, so classification tends to be applied to the block rather than to the address: your address’s neighbours inside the same range shape how it is treated before you have sent a single request. That is a property of how the space is allocated, not a judgement about your traffic.

ISP Proxies: Residential Registration, Datacenter Wiring

An ISP proxy is a deliberate split between two facts that normally travel together. The registry record says a consumer ISP — the kind of organisation that sells broadband to households. The route is announced from a provider’s hosting rack. Registration and network disagree, on purpose, and that disagreement is the entire product.

The value proposition follows directly: when a target performs the lookup, it gets back the same classification it gets for that ISP’s genuine retail customers, with none of the hosting signal. The provider is not renting you a home connection. It is renting you a registration.

That framing explains the thing buyers find puzzling about the tier, which is that supply is tight and priced well above ordinary hosting space. Consumer-ISP-allocated IPv4 is scarce for a structural reason: the regional registries have nothing left to hand out. RIPE NCC made its final /22 allocation on 25 November 2019 and has run on a waiting list ever since. New ISP-registered space cannot be created — it can only change hands. A provider assembling an ISP pool is competing for a fixed stock of records, which is a very different supply problem from racking another server.

Note that “static” here is doing origin-adjacent work but is really a rotation claim, which is a separate axis: an address can be static and datacenter-registered, or static and carrier-issued, as with 静态移动代理. Do not read a stability promise as an origin claim.

Residential Proxies: Someone Else’s Home Connection, Borrowed by Consent

Residential is the only origin whose supply chain is a consent question rather than a procurement one. The other three are bought: rack space, a registry record, a SIM and a modem. A residential pool cannot be bought, because the addresses belong to households that were never selling anything.

The dominant sourcing model is peer-to-peer. A proxy SDK is embedded in a consumer application, and the device owner opts in — typically in exchange for payment, an ad-free tier, or unlocked premium features — to relay other people’s traffic while the device is idle and online. When you buy residential bandwidth, that is the arrangement you are buying into, one device at a time.

Three consequences follow, and none of them are marketing:

• Pool composition and uptime are outside the provider’s control, and therefore outside yours. Nobody schedules when a stranger’s laptop is switched on.
• An exit can vanish mid-session, because it is a phone that moved out of coverage or a laptop that was closed. Session length is a property of somebody’s day, not of your plan.
• The sourcing disclosure is a real document, and you should ask for it by name before committing volume. What you want to see is consent that is explicit, revocable, and not buried behind a dark pattern in an installer flow.

That last point is not a formality. Consent quality is the whole difference between a legitimate residential pool and one assembled from users who never understood what they agreed to — and it is a question about the provider, not about the technology. A vendor who cannot describe the sourcing model in writing has answered it anyway.

Mobile Proxies: Carrier-Issued Addresses Shared Behind CGNAT

On the origin axis a mobile proxy is two registry facts and nothing more. The announcing ASN belongs to a mobile network operator, and the address sits on the public side of a carrier NAT — the deployment that RFC 6598, published as BCP 153 in April 2012, reserved the 100.64.0.0/10 range to feed. The RFC is unusually blunt about why it exists: IPv4 address space is nearly exhausted, ISPs must keep supporting IPv4 growth until IPv6 is fully deployed, and no registry can hand every service provider unique space for its subscribers.

That is the origin definition, complete. The consequence people actually care about — that an address shared by hundreds of genuine subscribers behaves differently under scrutiny than one that is yours alone — is a trust argument rather than a provenance fact, and this site argues it in full elsewhere:

• What is a mobile proxy? — the trust case, and why carrier addresses are scored the way they are.
• How mobile proxies work — the request path end to end, and the mechanism behind address rotation.
• 代理类型 — and once mobile is the right origin, which behaviour to buy: rotating, dedicated, static, shared or backconnect. That is the connection axis, and it is a genuinely different question from this page’s.

The handoff is the point. Origin tells you which kind of address to buy. Behaviour tells you how that address should act once you have it. Answering the first question does not answer the second, and a buyer who stops here has done half the job.

Compare Carrier-Grade Mobile Proxies

Looking for the highest-trust tier? PXM2 provides dedicated 4G/5G mobile modems with real carrier IPs:

常见问题解答

Related Guides

Proxy fundamentals

核心移动代理指南