An Engineer’s Guide to Proxy Types: From IP Fundamentals to Choosing the Right Architecture for Web Scraping and Automation
Imagine a familiar scenario: you’ve built a new data extraction module for an e-commerce platform or a protected online service. You run a quick test locally — everything works flawlessly. You deploy the scraper to a server using a datacenter proxy pool (after all, they are cheap and fast) — and within two minutes, the scraper starts returning “403 Forbidden” and “429 Too Many Requests” errors, or runs into Cloudflare Turnstile and DataDome challenges.
Switching to a residential proxy pool often reduces the number of blocks and CAPTCHA challenges, but does not solve the problem on its own: success still depends on IP reputation, request rate, TLS fingerprinting, cookies, and client behavior.
At this point, many engineers settle for a quick fix and return to their code, following the old adage: “If it works, don’t touch it.” However, in industrial automation, understanding why it works helps reduce infrastructure costs and debugging time. Residential proxies are not a collection of random techniques — they are a well-defined network architecture.
In this guide, we will break down the engineering mechanics of proxy networks: how target websites classify incoming traffic, why different protocols behave differently, how internal routing is organized, and which model — a targeted Premium plan, a Premium Mix plan for bulk traffic, or a thread-based Unlimited plan — best fits your specific task.
1. HOW A TARGET SERVER SEES YOUR REQUEST: ASN, WHOIS, AND IP CLASSIFICATION
Every time your script or browser establishes a TCP connection with a website, the remote server receives the source IP address. But an IP address is not just a random string of numbers — behind every address lies a well-defined public hierarchy.
IANA, the global Internet authority, allocates large blocks of IP addresses to Regional Internet Registries (RIRs: RIPE NCC, ARIN, APNIC, etc.), which in turn distribute subnets to end-user organizations: hosting providers, telecom operators, and enterprise networks.
This entire allocation chain is fully documented and publicly accessible via WHOIS databases and IP intelligence services (MaxMind, IP2Location, DB-IP). When analyzing an incoming request, anti-bot systems evaluate the ASN (Autonomous System Number) alongside IP reputation, request rate, TLS fingerprint, and session behavior:
- Datacenter IPs: Belong to server racks and cloud platforms. Their ASN is easier to recognize as server infrastructure; therefore, when automated requests are sent at a high rate, the risk of verification challenges or bans is higher.
- Residential IPs: Assigned by consumer ISPs to actual subscribers using home, apartment, and office connections. Such an IP makes traffic look like it comes from a genuine home user at the network level, but does not by itself neutralize other anti-bot signals.
2. UNDER THE HOOD: HTTP/HTTPS, SOCKS5, AND DNS LEAKS
When routing traffic through a proxy, the chosen protocol dictates how data is handled at the transport and application layers of the OSI model.
HTTP / HTTPS (OSI Layer 7 — Application Layer)
For plain, unencrypted HTTP traffic, the proxy server reads the request in full and forwards it on your behalf.
However, when working with HTTPS, the client sends a special HTTP request method: CONNECT target-domain.com:443. The proxy server opens a raw TCP tunnel to the destination server. Once established, the TLS handshake takes place directly inside this tunnel between your client and the target website.
SOCKS5: A General-Purpose Proxy Protocol
SOCKS5 operates between the application and transport layers and is completely agnostic to the application protocol. It can carry web traffic, WebSocket messages, and raw binary streams alike. Following a brief initial handshake, SOCKS5 simply shuttles raw bytes between the client and the destination.
Preventing DNS Leaks
One common cause of loss of anonymity for scrapers is a DNS leak. If a script first resolves the target domain's IP via a local DNS server, that query bypasses the proxy tunnel entirely: the local DNS resolver sees which domains the client is accessing and can correlate requests with your real network identity.
SOCKS5 allows the client to pass the domain name directly to the proxy and perform DNS resolution on the proxy node. However, this only works if the client is explicitly configured for remote DNS resolution: if local resolution occurs prior to connection, SOCKS5 cannot prevent a DNS leak.
3. RESIDENTIAL NETWORK ARCHITECTURE: FROM YOUR SCRIPT TO THE TARGET WEBSITE
A residential proxy network consists of three fundamental components:
1. Gateway: A data-center entry point. It accepts connections from your scraper, handles authentication, and maintains a fast connection to your client.
2. Routing Layer: The intelligent core. It evaluates your connection parameters — target country or city, whether to keep the current IP or assign a random one — and selects the optimal residential exit node.
3. Exit Node: A physical device with a real residential internet connection (a home router or PC) through which the request reaches the public internet on its way to the target website.
Where Exit Node IPs Come From
A residential proxy pool is not simply a list of addresses created by the provider. It is formed by IP addresses that consumer ISPs assign to home and mobile subscribers. In a typical network, end-user devices running client software maintain an outbound connection to the provider’s orchestration infrastructure, serving as exit nodes.
This model introduces practical operational constraints: a home device may change its IP address or go offline at any time. Consequently, a sticky session cannot last indefinitely — when a node drops, the orchestrator assigns a replacement. For the client, this process is completely transparent: you connect to a single gateway endpoint rather than directly to individual home devices.
4. THREE A-PARSER PROXY ARCHITECTURES FOR DIFFERENT USE CASES
Depending on your project requirements — precise city targeting, scraping millions of pages at a controlled cost, or 24/7 crawling across hundreds of threads — each use case requires a different approach to billing, authentication, and routing.
Architecture 1. Residential Premium Proxies: Country- and City-Level Targeting
Primary focus: Precise geolocation, good IP reputation, and extended session persistence.
If you manage ad campaigns (Google Ads, Facebook Ads, TikTok), operate anti-detect browser profiles (Dolphin Anty, AdsPower, Octo Browser, GoLogin), or scrape local search engine results with city-level accuracy, precise geographic targeting and session stability are often important.
Example: A request configured with parameters “country=DE”, “city=Berlin”, “session_id=abc123” will be routed by the Premium pool orchestrator to a residential IP in Berlin. The IP remains pinned to this session for up to 60 minutes.
Key Features
- Global Coverage: A pool of over 10,000,000 residential IPs with filtering options for 200+ countries and 5,000+ cities.
- Rotation Modes:
- Random IP on every request;
- Interval-based rotation — from 1 to 60 minutes;
- Sticky Session mode: the proxy retains the initial IP for the maximum possible duration; essential for logins, cookie persistence, and account warm-up.
- List Generation: Generate unlimited proxy lists in the Client Dashboard, each containing up to 10,000 IP addresses for different combinations of countries, cities, and timeouts.
- Instant Kill-Switch: Resetting your password in the dashboard instantly invalidates all previously generated proxy lists.
- Traffic-Based Billing: Pay only for consumed bandwidth — from $0.9 to $1.7 per GB. No restrictions on concurrent threads or connections per proxy.
Architecture 2. Residential Premium Mix
Primary focus: Low per-gigabyte cost for large-scale data collection.
When extracting 500,000 e-commerce product listings, monitoring competitor price feeds, or crawling terabytes of raw text for training LLMs and populating RAG databases — where specific country or city targeting is unnecessary — the global pool is a practical option. Before launching large batches, always verify with sample requests whether prices or inventory differ across regions.
How It Works: Premium Mix does not filter nodes by country or city and uses the entire global pool of 10 million IPs across 200+ countries. With per-request rotation, each connection receives a fresh random IP; bandwidth rates start at $0.30 per GB.
Why Premium Mix Can Cost Less
Premium Mix uses a shared global pool without being tied to a specific city or postal code: 10M IPs in 200+ countries.
- Bulk Pricing: From $0.30 to $0.80 per GB. For instance, a 500 GB package costs $250 ($0.50/GB), while a 3,000 GB package costs $900 ($0.30/GB).
- Unlimited Concurrency: No limits on parallel threads or concurrent socket connections.
- Authentication: Username and password.
- Operational Modes: Supports both per-request rotation and sticky sessions up to 60 minutes.
Architecture 3. Unlimited Residential Proxies (Thread-Based Billing)
Primary focus: High-volume 24/7 scraping with predictable traffic costs.
When operating headless browser clusters (Playwright, Puppeteer, Selenium), rendering heavy Single Page Applications (SPAs) with multiple asynchronous requests, or performing deep full-site crawls, bandwidth consumption can rise quickly. Under per-gigabyte pricing, budgeting becomes harder to predict: a single crawler bug or an unintended loop that keeps loading background video can consume a significant amount of traffic.
These workloads use a thread-based billing model.
For instance, for 500 threads running 24/7, you choose a fixed-rate thread package. All bandwidth is included with zero per-GB charges, and access is authenticated via your server's IP address.
How Unlimited Access and Authentication Work
Unlike gigabyte-metered plans, access here is authorized via IP Binding in the Client Dashboard. You specify your scraping server or desktop IP address once, then connect via standard
Code:
gateway-ip:portTwo dynamic proxy lists are available in the dashboard:
- Sticky Port Proxy List — perm_socks.txt: Each port is assigned a residential exit node with its own IP. The proxy remains pinned to its port while online. The list is automatically refreshed on the server every 30 seconds to ensure that only active IPs are listed. When the scraper requests the URL, it fetches the latest version of the list. Ideal for tasks that require keeping the same IP throughout a session.
- Random IP Proxy List — rand_socks.txt: The exit IP rotates on every new connection, selected randomly from a live pool of 100,000+ addresses. This list is static and does not need to be reloaded.
Unlimited Plan Economics
- Pay a flat subscription rate based on concurrent thread capacity (weekly or monthly) — from 20 threads at $10/week or $29/month up to 2,000 threads at $379/week or $999/month.
- All proxy traffic is included in the flat rate and is not billed by the gigabyte.
5. INDUSTRY OVERVIEW: COMPARING PROXY ALTERNATIVES
To make informed architectural choices, it is important to understand where other proxy types fit in the broader ecosystem:
- Datacenter Proxies: Fast and inexpensive, but their ASN is easier to recognize as server infrastructure. Suitable for open APIs and lightly protected targets; sensitive scraping requires conservative request rates and robust monitoring.
- Static Residential (ISP Proxies): Hosted in data centers but registered under ASNs assigned to consumer ISPs. They combine data center speed and stability with the trust associated with residential ASNs, but are expensive per static IP and gradually accumulate anti-bot flags over time.
- Mobile 4G/5G Proxies: Route traffic through cellular networks utilizing CGNAT, where many mobile subscribers may share a single public IP. Blocking such an IP may also affect legitimate users. Mobile proxies are usually more expensive and slower than residential proxies; choose them when the requirements of a particular site or use case justify the cost.
6. DECISION MATRIX: WHICH ARCHITECTURE TO CHOOSE?
| Use Case | Recommended Type | Authentication | Operating Mode | Why This Choice |
|---|---|---|---|---|
| Traffic Arbitrage, Google / Facebook Ads | Premium (Geo-Targeted) | Username / Password | Sticky IP (15–60 min) | Requires precise country/city geolocation and a persistent IP throughout the session. |
| Anti-Detect Browsers (Dolphin, AdsPower) | Premium (Geo-Targeted) | Username / Password | Sticky IP (30–60 min) | Useful when a session requires a stable IP and a specific location. |
| E-commerce & Marketplace Price Scraping | Premium Mix | Username / Password | Per-request Rotation | Low cost starting at $0.30/GB when precise city/country targeting is not required. |
| LLM / RAG / AI Dataset Crawling | Premium Mix | Username / Password | Per-request Rotation | Premium Mix rates start at $0.8/GB and decrease with volume, reaching $0.3/GB for the largest packages. |
| 24/7 Scraping in A-Parser | Unlimited | IP Binding | Sticky IP per port | Traffic is included in the plan, the budget is fixed, and the service scales up to 2,000 threads. |
| Regular SEO Position Tracking | Unlimited (20–100 threads) | IP Binding | Random IP on connection | A flat weekly or monthly fee with no separate per-gigabyte charge for heavy reports. |
7. PRACTICAL CONFIGURATION AND CONNECTION FORMATS
Proxy access follows two distinct models depending on the selected plan.
1. Configuring Premium and Premium Mix: Authentication & Parameters
Connections are established over SOCKS5 (or HTTP) using username and password authentication.
A core architectural feature: all targeting and rotation parameters are embedded directly within the username using suffixes. You do not need to change ports or gateway addresses — the gateway parses the parameters on the fly and routes traffic to the matching exit node.
Connection String Format:
Code:
[username]_[parameters]:[password]@[gateway-domain]:[port]Targeting parameters (country, city) and rotation settings (rotation interval, sticky mode) are passed directly in the connection string. In the Client Dashboard, the list generator automatically creates a ready-to-use list with the required parameters.
List Auto-Generation URL Format:
Code:
https://[list-generator-domain]/proxyList?login=[username]&password=[password]&[parameters]&format=login:password:host:portSample Generated Proxy List (up to 10,000 IPs):
Code:
[username]_[parameters]:[password]@[gateway-domain]:10000
[username]_[parameters]:[password]@[gateway-domain]:10001
[username]_[parameters]:[password]@[gateway-domain]:10002
...
[username]_[parameters]:[password]@[gateway-domain]:19999Key Engineering Recommendations for A-Parser
- In Proxy Checker settings, set proxy type to SOCKS5 and enable “Use proxy authorization”.
- For pre-filtered Premium and Premium Mix lists from A-Parser, we recommend enabling “No check proxy”: routine proxy checks consume prepaid bandwidth. When importing third-party or custom proxy lists, check them using the standard method first.
2. Configuring Unlimited Proxies (IP Binding)
Unlimited proxies do not require username/password authentication. Simply add your server or workstation's public IP address to the Client Dashboard.
Two dynamic list endpoints are available:
- http://list-server-domain/prx/perm_socks.txt — Sticky IP per port, refreshed every 30 seconds.
- http://list-server-domain/prx/rand_socks.txt — Random IP on each new connection.
Proxy List Format (IP
ORT):
Code:
198.51.100.10:21000
198.51.100.10:21001
198.51.100.10:21002
...
198.51.100.10:21047Key Engineering Recommendations for A-Parser
- For workflows requiring sticky sessions, use the perm_socks.txt list: enter its URL in A-Parser’s Proxy Checker and enable auto-refresh.
- Thread Budget Allocation: The sum of Proxy Checker threads (“Check threads”) and all active task threads (“Threads count”) must not exceed your plan's thread limit (e.g., 100, 500, or 2,000 threads). Exceeding this limit causes slower speeds and gateway timeout errors.
- In Proxy Checker for perm_socks.txt, enable “No check proxy” as well, since the list is already filtered on the backend.
SUMMARY
Effective data collection is not about finding a “one-size-fits-all proxy”, but about choosing the right setup for your specific workload:
- Need precise country, city, and session control for advertising and anti-detect browser workflows? → Residential Premium with username/password auth and per-GB billing.
- Need a lower per-GB cost for scraping millions of pages? → Residential Premium Mix with a global pool starting at $0.30 per GB.
- Need continuous 24/7 scraping across dozens or hundreds of threads? → Unlimited Residential Proxies with IP Binding, thread-based billing, and all bandwidth included.
Core conclusion: There is no single universal proxy for every workload. Determine whether your priority is granular geolocation, cost per gigabyte, or the number of concurrent threads — and select your setup accordingly.
Helpful Links
Last edited by a moderator: