Technical Reference · Network Selection
VPN Protocols and Routes
Separate connection setup, transport, and exit routing, then use real-world tasks to decide whether to change protocols or routes.
If you haven't installed a client or imported a subscription yet, follow the Quick Start Guide to verify your connection, then return here to compare options. This page doesn't repeat the setup steps. Instead, it explains why the same app can behave differently across networks, protocols, and exits—and how to make comparisons you can verify.
Connection Layers and Setup Time
A client showing “connected” doesn't mean the app has received a response.
Map the full route, from your device to the app
When choosing a route, distinguish between the client, entry point, exit, and destination service. The client handles requests that match its proxy rules, then connects to an entry point through home broadband, an office network, or mobile data. The entry point accepts the session; the exit is the side of the connection that leaves the service network for the destination. The entry and exit may be the same node or separate nodes. The address your client connects to isn't necessarily the exit address seen by the destination service.
A protocol defines how the client and proxy establish a session, encapsulate data, and pass it along; a route describes where that data travels. The app's own communication method is another layer. For example, a website may use an encrypted connection even when the outer proxy uses a different transport. The exit's location, the distance from your device to the entry point, and the app server's actual location are all separate factors. The region shown in the client usually identifies a node; it doesn't reveal the full round trip or predict how quickly a particular site will respond.
Treat cold starts and existing connections separately
A first visit may involve resolving the proxy entry point, setting up the underlying transport, and completing encryption negotiation. The destination may then be resolved locally or remotely before the exit connects to the service. The webpage also has to establish its own session and wait for the server to process the request. These steps aren't always sequential: some results may be cached, some work can happen in parallel, and retransmissions can add delays. To compare setup speed, don't just count handshake arrows in a protocol diagram. Check whether a session already existed and where DNS resolution happened.
For a fair comparison, use the same device, network, and page. Record the first visit after reconnecting, then compare it with repeated visits over an established connection. Reopening a tab may not close the connection pool, and restarting the proxy may not clear the system's DNS cache. You don't need to clear every cache; just note what you reset and do the same in both tests. If the first visit is slow but later requests are smooth, check DNS, handshakes, or wake-up delays. If the page loads but a download gradually stalls, investigate sustained transfer instead.
Track resource use alongside end-to-end performance
Throughput is the amount of data delivered per unit of time; latency is how long a message takes to arrive or get a response. Large downloads depend more on sustained throughput, while interactive pages, remote terminals, and voice calls are more sensitive to sudden response delays. A client’s entry-point ping is only the result of a specific test to a specific address. It says nothing about the route from the exit to the destination, and excludes app rendering, queueing, and server response time.
Resource use isn't the same as protocol overhead. A device also encrypts and decrypts data, moves it between components, and schedules connections. Clients using a virtual network interface may handle extra routing and data processing too. A file may download faster while briefly using more processing power, but finishing sooner can also reduce total battery use. To compare resource use, keep the task consistent and note completion time, interruptions, and noticeable heat. Comparing processor usage at one moment can mislead if the amount of data transferred differs.
Shadowsocks and VMess
Shadowsocks focuses on lightweight proxy encapsulation; VMess needs to be evaluated together with its transport.
Shadowsocks: Check the encryption method and outer layer
In a common Shadowsocks setup, the client encapsulates the destination address and app data and sends them to the proxy, which then connects to the destination. With modern authenticated encryption, the receiver can detect tampering and protect data between the client and proxy. The basic setup doesn't require an additional TLS handshake, but it still needs a TCP connection when using TCP. Plugins or other outer layers add their own connection overhead.
To assess compatibility, don't just check whether a client lists the protocol. The encryption method, address format, and outer-layer options must match the subscription configuration, and support for the same fields can vary between clients. If a node imports successfully but fails on one device, first check the proxy configuration actually in use and the device's system rules. Don't keep weakening security settings in the hope of getting lucky. UDP forwarding also requires support from the client, server, and current network. A working webpage doesn't prove that an app's datagrams can get through.
VMess: Check the transport before comparing overhead
VMess defines its own request encapsulation and authentication. In practice, it can use different transports, so configurations with the same label may have different connection setup steps. For example, a WebSocket setup with outer TLS must first establish a secure session and complete the upgrade before sending proxy data. If an existing connection is reused, those steps may not be repeated for every visit. To understand setup time, consider both the proxy protocol and any preceding steps.
If a fixed configuration works on one device but suddenly fails on another, check whether the system clock is set to update automatically, whether both devices use the same configuration source, and whether the client supports the configured transport. Some authentication mechanisms and outer-layer certificate checks can be affected by clock drift, but that doesn't mean every failure is caused by the time. If switching networks on the same device immediately restores the connection, first check whether that network can reach the entry point, then use the log stage to narrow down the problem.
Don't treat protocol names as a ranking of resource use
With similar routes and outer layers, fewer encapsulation steps can make setup and processing overhead easier to manage. But encryption implementations, hardware acceleration, and how the client moves data all affect the result. The processor on an older device may max out during a large download, while a newer device may be limited entirely by the network. Check whether processor use stays high during transfer and whether the device heats up, then pause the task to see if the load drops. If it doesn't, don't blame a particular layer based on heat alone.
When several apps are active, sharing an outer connection can avoid repeated handshakes, but it can also make tasks share queues and recovery processes along the same route. With many short requests, check whether repeated clicks still involve noticeable waits. During large transfers, see whether smaller concurrent requests slow down. Before comparing, make sure both tests use the same destination and proxy mode. If one proxies only selected apps and the other routes more system traffic, differences in resource use and response time can't be attributed to the protocol alone.
Trojan and VLESS
Different names don't necessarily mean different outer layers. First identify which layer actually provides security.
Trojan: Establish the secure outer session before sending proxy requests
Common Trojan deployments use TLS as the outer layer. The client first establishes a secure session with the server, then sends the proxy request and authentication details. When a connection fails, distinguish an unreachable entry point, a failed TLS check, and a rejected proxy request. If the logs show that the outer handshake completed, repeatedly changing certificate options won't help. Check that you imported the right subscription, the node is up to date, and the client is using the configuration for the selected node.
Outer certificate checks help confirm that the client has connected to the expected service. If verification fails, first check the system clock, then confirm that the server name in the configuration matches the service settings. Get updates from a trusted subscription source. On public Wi-Fi, complete any network access page first. Don't disable certificate checks as a routine fix: even if the connection appears to recover, you lose the identity assurance those checks provide.
VLESS: Lightweight proxy encapsulation; security depends on the configuration
VLESS typically handles lightweight proxy requests and identity checks. The protocol name alone doesn't guarantee encryption across the connection. Check which security layer and transport are actually enabled, and whether the client supports the configured options. Two VLESS nodes may use different transports, different entry points, or have different levels of compatibility with your client. Copying options from one node to another can break a configuration that was designed to work as a set.
When comparing with Trojan, if both use TLS, first check whether the route and session state are also the same. When supported and enabled by both client and server, session resumption can reduce negotiation overhead. Connection reuse avoids repeated setup, but it can't shorten the time data takes to travel through fiber or skip the exit's connection to the destination. If the first visit is slow only after a long idle period, note whether the connection was re-established rather than blaming all initial delay on the proxy protocol.
Rule out three common mix-ups by checking where the error occurs
First, check that both the client and operating system are online. Then see whether the problem affects every destination, a type of traffic, or just one website. If all destinations fail and the client can't establish a session, the problem is more likely with access, the entry point, or configuration. If the outer connection is established but one destination fails, check the route from the exit, destination DNS, app state, or the service itself. A browser's certificate error for the destination site is different from a certificate error on the proxy's outer layer; don't treat them as the same issue.
Next, distinguish a client node test from an app's actual route. Some apps follow the system proxy; others use a different connection mode. A working page in your browser doesn't prove a desktop app follows the same rules. Finally, make sure you didn't change the protocol, transport, and DNS at the same time. Restore a known working configuration through the official import process, then change one independently controllable setting at a time. If a subscription only lets you switch complete nodes, record the result as a comparison between two node configurations rather than claiming to isolate one setting.
Hysteria2 and TUIC
Understanding QUIC streams, datagrams, and congestion control is more useful than choosing by protocol label.
How QUIC can reduce blocking between streams
QUIC runs over UDP and provides encrypted sessions and multiple streams at the transport layer. If data is missing on one reliable stream, that stream still has to wait for the missing data, but independent streams don't all have to wait for the same ordered TCP byte sequence. That's different from putting many requests on a single TCP connection. However, all streams still share the link, congestion window, and sending schedule. If packet loss slows the whole connection, other streams are affected too.
An outer proxy using UDP doesn't make the traffic inside it unreliable. Byte streams that need reliable delivery, such as web traffic, still have ordering and recovery mechanisms. For apps that prioritize timeliness and can tolerate missing data, a specific proxy implementation may also provide datagram forwarding. Distinguish reliable streams from datagrams before deciding whether retransmission can resolve a particular stall. The destination service's own connection method doesn't change just because the client uses QUIC on the outer layer.
Hysteria2: Match sustained transfers to the actual link
Hysteria2 is based on QUIC and is often considered for sustained transfers over jittery or lossy connections. Its congestion control options depend on configuration and can include modes that respond to bandwidth estimates. Setting a higher bandwidth expectation won't increase the capacity of the physical link. Sending more data than the bottleneck can deliver may lengthen queues and slow small requests, control traffic, and other devices on the same network. A higher peak download speed may look impressive while making page interactions worse; choose based on the task.
If your subscription provides a ready-made route, start with the parameters supplied by the service instead of copying bandwidth values from another network. Run a repeatable transfer while checking how short requests on the same route respond and whether queues build up, then test again after stopping background transfers. If performance improves noticeably, contention on the link may be the main issue, not the protocol itself. If the connection won't establish at all, first test another access network rather than tuning transport parameters.
TUIC: Test short connections, datagrams, and network switching separately
TUIC also uses QUIC to carry proxy sessions and distinguishes reliable stream traffic from datagrams. For many short requests, check whether an existing session can carry new tasks. If you need datagrams, test with the real-time app itself; a regular webpage isn't a substitute. Client implementation, configuration, and server settings all affect these behaviors, so a TUIC label alone doesn't prove that a particular forwarding mode is enabled.
QUIC is designed to validate a new path and preserve a connection under certain conditions. But switching networks can also prompt the client to reconnect, change proxy mode, or invalidate an app session on the exit side. When testing a switch from Wi-Fi to mobile data, note whether the proxy reconnects, whether the app recovers on its own, and whether unfinished content can continue. One practical trade-off: if UDP is consistently unstable on your network, a TCP-based outer transport may hold the connection more reliably. This finding applies only to the access network tested; check again after switching networks.
Direct, Relayed, and Dedicated Routes
Route labels describe only part of a path. When choosing, clarify which segment each label covers.
First, identify how traffic is handed off between entry and exit
A route labeled “direct” usually means the client connects straight to the remote node serving the connection, without an explicit proxy relay. It doesn't mean the data crosses the network through just one router. Also distinguish this from “direct” in a client's rules, which usually means a request bypasses the proxy entirely. That's not the same as a proxy route labeled direct. If a rule bypasses the proxy, that visit can't be used to assess the proxy exit.
A relay separates the entry point from the exit. The entry may be closer to your network, with traffic handed off elsewhere within the service network. Whether this helps depends on the alternative route, not the number of hops. If the direct path to the exit crosses a congested peering point, a different entry and intermediate route may be more stable. Conversely, an added segment with insufficient capacity can make the entry-point test look better without improving real-world access.
| Route label | Key difference | What to verify |
|---|---|---|
| Direct route | Device connects to a remote node | Entry point reachability and end-to-end response at busy times |
| Relayed route | Entry and exit are separate handoff points | Whether changing the entry reduces long-connection drops |
| IEPL dedicated line | An Ethernet private line carries traffic over a specified segment | Whether cross-segment performance is steadier when access and exit are working normally |
A dedicated line covers a specified segment; it doesn't mean the entire route is exclusive
IEPL is an international Ethernet private line, typically providing Ethernet connectivity between sites. For end users, the key is to identify which segment uses the private line: the route from your device to the entry is still affected by local access, and the route from the exit to the destination may still depend on public peering. Even with more predictable transit in between, a congested home Wi-Fi upload or a slow destination server can extend end-to-end wait times. Assuming a dedicated line makes the whole network congestion-free can send troubleshooting in the wrong direction.
Distance mainly affects propagation latency, while handoffs, queues, and packet loss affect real-world performance. A good route can avoid unnecessary detours and queues, but it can't eliminate the round-trip cost of distance. The return path may also differ from the outbound path, so congestion on the way back can cause intermittent stalls even when outbound traffic looks smooth. A page loading instantly once only shows that request met the right conditions; it doesn't reveal the full route.
Coverage, exits, and comparable route options
1NVPN covers 90+ countries and offers 230+ routes. On the Servers page, find options by country or region, exit city, and route type, then test them against the same destination on the same device. If your task requires a specific exit region, hold that constant when comparing direct and relayed routes. If not, broaden your options. Don't skip end-to-end testing just because a route is geographically closer: destination services may assign different servers based on the exit network, and distance is only one factor.
When choosing a backup route, prioritize whether it can isolate a failure rather than just collecting several similar-looking names. Different entry points may still share the same access or cross-region segments, and typical users can't see the full topology. What you can verify is whether switching restores service when the same problem recurs, and whether that result is repeatable. Record home broadband and mobile data separately; one successful visit on the latter doesn't establish what to expect from the former.
Peak-Hour Congestion and Packet Loss
Use timing, scope, and route comparisons to investigate; don't assume every stall has the same cause.
Congestion often shows up as longer waits first
When more data arrives than a link can deliver to the next segment, it queues up. Persistent queues can slow even tiny interactive requests without obvious packet loss; packet loss may increase only when the queue runs out of space. At peak times, heavy traffic can bottleneck access aggregation or cross-network peering. Long uploads can also slow downloads and page interactions because acknowledgments and DNS requests still need outbound capacity.
Reliable transport attempts to recover lost data. Recovery takes time, and the sender may reduce its rate, so users experience downloads that speed up and slow down, brief audio or video interruptions, or delayed responses in remote sessions. A connection that hasn't dropped isn't necessarily transferring smoothly, and a client that remains connected doesn't guarantee the app is receiving data. Conversely, a busy app server can take a long time to respond without any packet loss on the network. The same kind of stall doesn't necessarily have the same cause.
Build a troubleshooting sequence you can repeat
First, rule out contention on your home network. Temporarily pause cloud sync, background downloads, or livestream uploads; move closer to the Wi-Fi access point, or test over a wired connection if possible. Then compare an existing route with an alternative in the same app, noting the time, access method, destination, and specific symptoms. If the same task works smoothly during the day but fails again at night—and changing the access network or route fixes it—the comparison is more useful than an isolated speed-test screenshot.
When comparing entry-point tests, remember that intermediate devices may rate-limit or deprioritize the test packets. A hop that doesn't respond doesn't mean it dropped the traffic it forwards to the next node. A stable entry-point test also says nothing about the route beyond the exit. If only one website fails across several candidate routes, first compare it with another working destination rather than changing the entire client based on that site's status. It's more useful to note whether the timeout happens before the app starts or midway through a task than to record only “slow.”
Distinguish queueing, access instability, and packet-size issues
Some issues have little connection to the time of day but are closely tied to location or packet format. Moving around can change Wi-Fi signal strength; channel contention or switching mobile networks can cause brief instability. In other cases, small requests work but larger transfers keep stalling. On an encapsulated route, it may be worth checking the supported packet size and fragmentation handling. But server throttling, caching, and upload timeouts can cause similar symptoms, so one failed large download isn't enough to diagnose a packet-size issue.
For a suspected packet-size issue, first note the access environment and a repeatable task that triggers it. Then compare the client defaults with another supported transport or access network. Don't change system network parameters and proxy settings at the same time. If the issue occurs consistently on just one route, document the difference between the working and failing cases and what you've ruled out. That's more useful than a generic setting with no context about the client or route.
Mobile Battery Use and Session Recovery
Wake-ups during standby and work done during sustained transfers are two different issues.
Separate idle battery use from heavy transfers
During sustained transfers, a device sends and receives wireless data, handles encryption and encapsulation, and passes data to the app. Power use isn't determined by the protocol alone: screen brightness, video decoding, signal strength, and background tasks can easily change results. Keep the screen and app state consistent, transfer roughly the same amount of data, and note whether you're using Wi-Fi or mobile data. You can't compare total power use directly if one test transferred more data or ran longer.
When idle, look at keepalives and reconnections. Small but frequent packets can make it harder for the wireless radio to stay in a low-power state. With a weak signal, searching for a network, recovering from errors, and rebuilding sessions all add work. Overly frequent keepalives can trigger more wake-ups; keepalives that are too sparse may let intermediate devices clear idle mappings, forcing a new connection next time. With an off-the-shelf client, observe its default behavior first. Don't change keepalive settings based on a single battery reading.
The same protocol can behave differently across platforms
1NVPN supports Windows / macOS / iOS / Android / Linux. Platform support means you can find a compatible connection method; it doesn't mean every client handles background tasks, sleep, or system routing in the same way. After a desktop wakes from sleep, its previous session may have expired. On mobile, system policies can restrict background activity when the screen is locked. Include the platform, connection mode, and test conditions in your comparison.
| Platform | Scenario to reproduce | What to check |
|---|---|---|
| iOS / Android | Lock the screen, unlock, or switch networks | Whether the proxy reconnects and the app reconnects on its own |
| Windows / macOS | Resume work after sleep | Proxy settings used by the app and whether the task can resume |
| Linux | Different results in terminal and desktop apps | Proxy environment and system routes used by the process |
If updates stop arriving when the screen is locked, first work out whether the issue is background app activity or connection recovery. If messages arrive as soon as you unlock the device and the proxy logs show no reconnection, check how the system schedules background apps before blaming a dropped route. If the proxy has reconnected several times and the app only recovers when reopened, check both stages. See the Getting Started Guide for installation, import, and system permissions. For initial iOS setup, see the iPhone VPN Setup Guide.
Establish a battery baseline on your usual networks
Choose a route that connects reliably where you usually use your device, then record battery use and session state during typical standby and active use. Compare another transport on the same device at a similar time. System battery reports may not attribute all network-related power use to the proxy process, so use them to spot trends rather than calculate the exact cost of each layer. If network quality differs between tests, repeat them instead of treating the result as a general protocol ranking.
For real-time sessions you use every day, prioritize wake-ups and whether connections survive network changes. For short tasks mostly done with the screen on, focus on how quickly the connection starts and whether unnecessary background activity continues afterward. If your device keeps reconnecting in a weak-signal area, test from a more stable access point before judging the configuration. Battery life and reliability aren't always at odds: avoiding needless reconnects is usually more useful than repeatedly retrying the same unstable link.
Choose Protocols and Routes by Use Case
Let your most common tasks determine your primary setup, and keep an alternative ready for when something goes wrong.
Interactive work: find where the delay occurs
With AI coding tools, web-based chat, and remote development environments, the first response, ongoing output, and recovery after idle time are different parts of the experience. If the response starts slowly but then stays steady, compare it with a regular visit to the same service to see whether the delay comes from the handshake, server-side generation, or the workspace itself. If output often stops halfway through, check whether the task coincides with network switching, heavy traffic contention, or a particular time of day. Traffic from an IDE and command-line tools may also follow different routing rules, so a browser-only test isn't enough.
For a consistent comparison, repeat the same request in the same project and note the first output, ongoing interaction, and first action after idle time. Don't use actions that automatically submit, publish, or incur charges as repeatable test cases. For everyday use, a route that keeps your workflow going and recovers from problems is more useful than one that's fast once but forces frequent disconnects and reconnections. See How to Choose a Route for AI Coding Tools for tool-specific guidance.
Streaming, downloads, and real-time apps: different metrics matter
For streaming, first check whether the service works through the selected exit, then watch for buffering and whether playback resumes smoothly after seeking. Access to the service's webpage doesn't mean every video asset takes the same route, and peak download speed isn't a substitute for uninterrupted playback. Keep the content and resolution consistent when comparing routes so automatic quality changes don't hide congestion. For downloads, track completion time and retries rather than relying on a momentary peak.
Voice and gaming are more sensitive to uneven delivery times; data that arrives late may no longer be useful. First confirm that the app's traffic actually uses the route, then look for sudden interruptions during interaction. If problems occur only while downloading at the same time, reduce contention on the shared network first. If a connection fails only on a particular access network, compare it with a tested alternative transport. Game-specific routes are optimized for different purposes; see Game Boosters vs. VPNs: How to Choose for details.
Keep a primary route, a backup, and clear retest criteria
Start by finding a route that handles your main tasks during the times you usually use it. Save the node name, protocol combination, exit region, and access network it works with. Then decide when to switch: for example, don't change routes because of one test result; switch to the backup only after the same app stage repeatedly fails and you've ruled out local network contention. Test the backup at least once in the conditions where problems are most likely, and confirm it can handle the key task. Opening a webpage once while idle isn't enough to rely on it for work.
If your subscription offers only complete node configurations, you can say, “On this access network, this node configuration completed the task more reliably.” There's no need to declare an overall protocol winner. Retest with the same kind of task after changing clients or network conditions. If you use multiple devices, test each one separately: support for unlimited simultaneous connections describes the service's device policy, not a guarantee that your local broadband or Wi-Fi has no shared bottlenecks. Once you've chosen a setup, compare monthly subscriptions and data plans on the Plans page based on how often you'll use them. Data plans last until used and never expire; billing options aren't a measure of route performance.