How Connections Work and What Metrics Mean
Start by mapping the full connection path
When you open a website, your device typically looks up the address for its domain, then connects to the destination. With a proxy client enabled, the process may also involve local traffic interception, rule checks, proxy authentication, and forwarding through a remote server. The client is not the website, and the proxy exit is not the website’s server. If a page keeps loading, the issue could be DNS resolution or the connection from the exit to the site. One symptom alone cannot tell you whether the entire route is down.
Think of the path as your device, local network, proxy entry, network exit, and destination service. With a direct topology, one service node may handle both entry and exit; a relay topology adds a forwarding step. Each segment has its own latency, bandwidth, and failure points. A client showing “connected” usually means the local tunnel or proxy service is active. It does not mean the remote site has returned usable content, or that every app uses the same exit.
Before testing, check how the app’s traffic enters the proxy. System proxy settings are generally used by apps that follow the system configuration. Virtual network interface mode can capture more traffic at the network layer, but is also affected by routing, exclusion rules, and system permissions. If one app works and another does not, check the capture scope and routing rules first. Switching protocols may change the symptoms temporarily while leaving the underlying configuration issue unresolved.
Latency, throughput, and time to first byte are different measures
Latency measures the round-trip wait for an interaction; throughput measures how much data can be transferred over time. Chat, page navigation, and remote control are more sensitive to latency, while downloads depend more on sustained throughput. Loading a page for the first time also involves DNS resolution, handshakes, server processing, and content delivery. A fast download test does not guarantee a fast page load. Treating a speed test as the whole experience misses these short-connection stages.
Connections can also be cold or reused. Starting a client or switching exits may require some setup again, while an existing session may avoid part of the wait. When comparing protocols, check both the first visit and subsequent visits; do not compare a warm connection on one setup with a cold connection on another. Browser caches, existing login sessions, and site-side caching can also make repeat tests look faster.
| What to observe | What it mainly reflects | What it cannot prove on its own |
|---|---|---|
| Connection setup | Wait time for DNS, handshakes, and authentication | Sustained download performance |
| Sustained throughput | Effective performance during long transfers | How quickly a page loads for the first time |
| Jitter and packet loss | Interaction consistency and retransmission pressure | That the proxy node is definitely at fault |
Set up a repeatable comparison
For a useful comparison, define the task before choosing a protocol. Are you trying to reduce the wait for the first page to load, stop video buffering, or prevent a long response from cutting out? Keep the device, access network, and destination service the same, then change only the exit or protocol. Note whether it is the first connection, peak usage time, or a network handoff. This gives you a meaningful comparison instead of an isolated speed figure with no context.
This guide does not rank protocols because client implementations, server configurations, and network conditions all affect results. The same protocol name can also refer to different transport encapsulation and connection reuse behavior. Identify the bottleneck first, then decide whether you need lower handshake overhead, different congestion control, or another route. That is more likely to produce a reliable result than repeatedly switching between similarly named nodes.
Shadowsocks and VMess: Trade-offs
Shadowsocks: a simple design does not mean every implementation is the same
Shadowsocks is a family of proxy protocols designed to carry protected proxy traffic between a client and server. It does not require every deployment to be wrapped as a web connection, so common setups are relatively straightforward. Protocol generations, encryption methods, and extended settings differ. Seeing Shadowsocks support in a client does not mean it can use every subscription configuration. If an imported profile contains unsupported fields, check the client’s core capabilities rather than deleting fields and hoping for the best.
Shadowsocks is often considered a lightweight option, but a simple design does not guarantee lower resource use on every device. Hardware acceleration for encryption, data copying, virtual network interface implementation, and logging detail can all affect CPU load. A client that uses little power while idle may heat up under many concurrent connections. Judge it using your own workload rather than relying on the word “lightweight” in a protocol description.
Shadowsocks can carry TCP traffic and, when both ends support it and are configured accordingly, forward UDP. A working web page does not prove that data for real-time voice or games is being forwarded correctly. Troubleshooting should check the transport used by the app, whether the client captures it, whether the server allows it, and whether the network can reach the destination. If any one part is missing, web browsing may work while real-time services fail.
VMess: distinguish protocol authentication from transport encapsulation
VMess has its own authentication and data protection mechanisms, while its behavior is also affected by the outer transport. Discussions often conflate VMess with encapsulation methods such as WebSocket, but VMess handles proxy communication and WebSocket provides a carrying channel. An extra layer may help with deployment or network compatibility, but it also adds processing steps. Two nodes labeled VMess do not necessarily have the same connection setup or transmission overhead.
Some VMess implementations require the device clock to be synchronized. If authentication fails despite a valid certificate, or a configuration works on one device but not another, check the system date, time zone, and automatic time setting. The point is to rule out an obvious clock mismatch; there is no need to guess the acceptable range. If logs clearly report an authentication error, refresh the subscription and check your account status rather than blaming insufficient bandwidth.
VMess may have mature import and routing support in long-established client environments, while a newer environment may offer a better option for the task at hand. Compatibility is a property of a specific client ecosystem, not a permanent feature of the protocol. Check whether your current client can fully parse the node, retain required transport parameters, and connect after the subscription is updated.
How to choose for your actual device
If your device has limited processing power, start with the subscription’s default settings, disable unnecessary verbose logs, and monitor load and temperature during sustained transfers. Avoid changing the encryption method, concurrency settings, and routing rules all at once; otherwise, it is hard to tell what helped. If the service offers different protocol endpoints in the same region, treat them as an approximate comparison only. They may still use different machines, routes, or rate limits.
For older devices, first check that the client is still maintained, system certificates are valid, and the subscription imports reliably. Staying on unmaintained software just to preserve an old configuration may create more problems than it solves. If a protocol is unsupported on the device, choose an option that the service actually provides and the client fully supports. Do not try to assemble a nonexistent node based on its protocol name.
Trojan and VLESS: Connection Boundaries
Trojan: understand where TLS comes in
A common Trojan design carries proxy traffic inside a TLS connection. TLS protects transmission between the client and proxy server; proxy authentication determines whether the connection is allowed. It uses related security technology to the one used when a browser visits an HTTPS site, but the connection endpoints are different: Trojan connects to the proxy server, while HTTPS connects to the website. Using Trojan does not replace the browser’s verification of the destination site’s certificate or change the site’s account rules.
Connection setup may involve waiting for DNS resolution, the network connection to the server, the TLS handshake, and proxy authentication. Whether an existing session can be reused depends on the client, server, and their settings. So it is too simplistic to say that TLS always makes a connection slow, or that one protocol always connects instantly. For frequent short tasks, session reuse and server response time can matter as much as the protocol name.
Treat certificate errors as certificate issues. Check the system time, server name in the subscription, and client logs, then make sure the subscription is valid. Do not disable certificate verification just to make the warning go away: that changes the identity verification boundary and may hide a configuration error. If the settings came from the panel, sync them again first. If the problem persists, submit the error category in a support ticket rather than sharing a full node link publicly.
VLESS: a lean proxy layer does not provide complete protection by itself
VLESS is designed as a relatively lean proxy authentication and forwarding layer; the protocol itself is not a complete transport encryption solution. Deployments typically pair it with an appropriate secure transport. A VLESS label identifies the proxy layer, but it does not tell you the connection’s protection, handshake process, or resource cost. The server and client must both support the required transport parameters; you cannot swap them arbitrarily on your device.
That is why copying a node requires more than keeping its address and authentication fields. The transport type, server name, and other associated parameters may all be necessary to establish a connection. Some fields may look optional but actually participate in routing or identity verification. If import fails, start with the complete subscription generated by the panel and check whether the client recognizes it. Do not delete unfamiliar fields and keep trying.
Choose between VLESS and Trojan based on the specific implementation and conditions. The combinations that import reliably on your device, how well the target network supports the transport, and whether connection reuse suits your apps all affect the experience. If the two options use different exits, you cannot attribute the whole performance difference to the protocol. To compare proxy-layer overhead, keep the outer transport and route as consistent as possible, while recognizing that some variables cannot be controlled completely.
Encryption scope, exit identity, and app privacy
A proxy tunnel protects a specific communication link; it does not automatically remove information that an app submits. Websites can still recognize your account, session state, and browser environment. The exit address seen by a destination service also does not mean your account’s region has changed. Payment region, past activity, and terms of service may each factor into a decision. An exit region is one access condition, not a guarantee of platform eligibility.
LeleVPN’s primary security messaging refers to quantum encryption. This does not mean every general-purpose protocol discussed here inherently supports post-quantum key exchange, nor does it prove that a particular node uses a specific cipher suite. Protocol names, security claims, and negotiated parameters are separate matters. Base technical conclusions on verifiable client information and service documentation; this page does not extend brand messaging into undisclosed claims about algorithms.
Protect local configuration as part of everyday use. Subscription links and node authentication details may grant access, so remove them before sharing logs or screenshots. For troubleshooting, it is usually enough to provide the client name, operating system, protocol type, and a redacted error category. If your account configuration needs further review, use a support ticket in the panel; do not upload a full subscription to a public forum.
Hysteria2 and TUIC: Transport Characteristics
What QUIC changes—and what it does not
Hysteria2 and TUIC are both related to UDP-based QUIC technology. QUIC handles secure connections, stream management, congestion control, and other tasks in user space, providing a way to carry multiple logical streams. It does not make UDP inherently reliable; it builds the necessary mechanisms on top of it. This distinction matters: using UDP underneath does not mean web data can be discarded at will, or that congestion on the route no longer causes delays.
Traditional TCP byte streams deliver data in order, so missing data can hold up later content in the same stream. QUIC’s independent streams can reduce some cross-stream delays, but each reliable stream still has to recover lost data, and streams on a connection may share a congestion budget. Multiple streams do not mean they are isolated from one another. If the underlying route keeps dropping packets or the wireless signal is poor, any reliable transport has to spend time and resources recovering.
The access network must allow UDP for these protocols to work well. Some networks restrict, drop, or handle UDP differently. The result may be a complete failure to establish a handshake, or a connection that starts normally and then stalls repeatedly. If a TCP-based option is stable in the same environment while QUIC-based options keep failing, check network compatibility rather than simply increasing bandwidth parameters. A protocol cannot recover data that the network keeps discarding.
Hysteria2 and TUIC are not interchangeable options
Hysteria2 provides a transport design for proxy use, while its congestion control behavior depends on the client and server implementations and negotiated settings. Some configurations specify expected bandwidth, but entering a higher value does not guarantee higher actual throughput. If it exceeds what the route can handle, queues may grow, packet loss may increase, and other apps on the same network may be affected. When using a subscription, follow the service’s settings rather than applying inflated values yourself.
TUIC also builds proxy capabilities around QUIC, but its authentication, forwarding, and connection management differ from Hysteria2. A client that supports QUIC does not necessarily support TUIC, and support for one TUIC configuration does not guarantee that it can parse every field in another implementation. Check that the subscription matches the client core, especially whether the import tool converts or drops required information. Similar names and shared foundations do not make configurations interchangeable.
In high-latency environments with packet loss, the two protocols may recover differently from common TCP-based options, but that potential depends on the route and configuration. If the server exit is already congested or the destination website responds slowly, changing the transport between your client and the proxy may not help. First determine whether the delay is in the proxy tunnel or on the rest of the route, then decide whether another option is worth testing.
The trade-offs of resources, battery, and network handoffs
User-space encryption, packet processing, timers, and keep-alive traffic all use resources. During sustained high-throughput transfers, processor load, wireless activity, and device temperature may rise together. When idle, overly frequent keep-alives can also prevent a device from entering a deeper power-saving state. Handshake speed alone cannot tell you how much battery a protocol uses, and QUIC is not inherently more or less power-efficient in every situation. Battery life depends on the complete implementation and how you use it.
QUIC has mechanisms for connection migration, but whether a proxy session continues smoothly when switching between Wi-Fi and mobile data depends on the implementation, network address changes, system background restrictions, and server policy. Existing web requests may also time out during a handoff. Test recovery after a network change separately and see whether reconnecting is necessary; do not treat a protocol-layer capability as a promise that apps will never disconnect.
Direct, Relay, and Dedicated Routes: Differences
Topology labels describe the route, not an overall performance ranking
A direct route usually means the client connects to the proxy exit node, which then accesses the destination service. There are fewer steps, and the configuration and failure boundaries are easier to understand, but the network between the client and node may still be complex. An exit that looks nearby on a map does not necessarily have a direct, smooth route. Carrier interconnections, actual routing detours, and congestion at different times can affect latency more than geographic distance.
A relay adds a forwarding entry between the client and the final exit. The entry may be closer to the access network, then send traffic to the exit over another link. Its value is not that it somehow eliminates distance, but that it changes the route and may replace a poor segment of the public network with a better connection. A relay also adds equipment, queues, and maintenance points. If the entry is congested or forwarding fails, multiple nodes using the same exit may be affected together.
An IEPL dedicated line generally refers to the use of international Ethernet private-line resources along part of the route. What matters to users is which segment has more controlled transmission conditions, not just the dedicated-line label. The connection from your device to the access entry may still rely on your local network, and the route from the exit to the destination website may still use the public internet. A dedicated line cannot prevent home Wi-Fi interference, limited device performance, or a busy destination service.
| Type | Route characteristics | What to test | Common limitations |
|---|---|---|---|
| Direct | Client connects directly to the proxy exit | The actual route from the local carrier to the exit | Public network interconnections and peak-hour congestion |
| Relay | Entry forwards traffic to the final exit | Entry reachability and stability of onward forwarding | Additional forwarding steps and shared queues |
| IEPL dedicated line | Dedicated-line resources are used along the route | Sustained tasks and peak-hour performance | The local access network and destination site still introduce variables |
The entry and exit locations address different needs
The entry location mainly determines how your device connects to the service; the exit location mainly determines where the destination service sees the request coming from. A city in a node name may identify the exit, not the location of the entry data center. An exit in a particular city may have higher latency without the city label being wrong; the entry or internal route may simply differ. For an accurate assessment, refer to the service’s route information rather than guessing from the distance between cities.
For streaming and AI tools, the exit region can factor into a platform’s regional checks, but platforms may also consider accounts, payment regions, address reputation, and other information. An exit in the expected location cannot guarantee that a feature will always be available. For everyday browsing, there is often no need to fixate on a city name. A route that reliably completes the task and recovers well is usually more useful than one that looks closer on paper but keeps fluctuating.
Different route types may be available within the same exit region to offer alternative paths. That does not mean they share the same entry, resources, or protocol. When comparing them, first make sure the exit region suits the task, then check stability on your access network. For regular work, keep a backup route with different characteristics so you are not just switching names while relying on the same failure point.
How to read LeleVPN route information
LeleVPN covers 120+ countries / 190+ routes. This describes the range of available coverage; it does not mean every location suits your local network or that every route type is available in every region. When checking the network locations page, consider the city, route type, and intended use together. Availability in your subscription and how the client labels each option are determined by the current panel and subscription.
If several nodes slow down at once, check whether they share an entry, region, or route type. If only one destination website is affected, look at the route from the exit to that site and any platform-side restrictions. Topology helps narrow troubleshooting to a particular part of the route; it is not a basis for ranking direct, relay, or dedicated-line options for every network.
How to Diagnose Packet Loss, Jitter, and Congestion
Data can arrive late or fail to arrive for different reasons
Packet loss means data did not arrive as expected under the transport mechanism, but there are many possible causes. Wireless interference can trigger retries on the local link; full device or router queues may drop data; route changes can cause brief interruptions. Reliable transports try to recover lost data, so users may not see a clear error and may only notice a page stalling or download speeds fluctuating. Assuming a server is underpowered whenever something hangs can make you miss a problem closer to your device.
Jitter describes changes in waiting time. Even if average latency seems low, occasional long pauses can still affect voice calls, remote control, and real-time collaboration. Downloads can hide some variation with buffering, but interactive tasks need data to arrive on time. When comparing routes, look beyond whether the task eventually completes: note any frequent pauses and whether they coincide with uploads or downloads from other devices on your local network.
Basic diagnostic tools have limitations too. Routers along the path may assign lower priority to probe packets, making one hop in a route trace appear to have packet loss even when later hops and real apps work normally. Do not treat an intermediate hop’s probe results as proof of packet loss affecting your traffic. Consider the final destination, app behavior, and repeated observations together instead of drawing too much from a single route trace screenshot.
Why peak hours can make delays worse
During busy periods, more connections compete for the same link and data waits in queues; retransmissions use even more capacity. Queues can form on the home router, access network, relay entry, proxy exit, or destination service. When the upload link is saturated, control packets and interactive traffic may also be delayed. Downloads may continue while other web pages become sluggish. That does not necessarily mean the proxy protocol has failed; shared queues may be the problem.
Effective congestion control adjusts the sending rate based on feedback from the route. More aggressive settings do not necessarily improve usable throughput; they may create longer queues instead. Bandwidth, concurrency, and buffer settings exposed by a client are not always better at higher values. For unfamiliar options, keep the subscription defaults. Adjust them only after identifying a specific bottleneck and understanding what the setting does, and keep a copy of the original configuration so you can restore it.
If performance is stable when idle but consistently worse when busy, compare different routes under similar conditions rather than switching protocols on the same route. If every exit stalls on the same home network but works on another access network, investigate the local connection. Cross-checking whether the issue lies with the route, device, or destination service is more useful than rerunning speed tests.
Make speed tests repeatable and useful
Before testing, pause background sync and large downloads, and keep the device location and network connection as consistent as possible. First check whether the local network works reliably without the proxy, then test the proxy task. Do not simply subtract the two speeds and call the difference protocol overhead: the destinations and routes may be completely different. Record whether the task completes, where it stalls, and the error category. Save redacted logs if needed, rather than recording only a peak speed.
To distinguish connection stages, use a terminal with curl installed to inspect the timing of an HTTPS request. The command below accesses a public example domain and contains no subscription or authentication details. Whether the request goes through a proxy depends on your terminal environment and how the client captures traffic. The output shows cumulative timings for that request, not a universal result for all apps. Caching or connection reuse can change the result.
curl -I -o /dev/null -sS -w 'lookup=%{time_namelookup}\nconnect=%{time_connect}\ntls=%{time_appconnect}\nfirst_byte=%{time_starttransfer}\ntotal=%{time_total}\n' https://example.com
If your terminal does not support the output redirection shown above, use the client logs and browser developer tools instead; there is no need to change system settings just to test. For more on keeping test conditions consistent, understanding tool differences, and recording comparisons, see our guide to VPN speed testing. The goal is to identify what can be improved, not to chase a best-case result that cannot be reproduced.
Platform Differences and Mobile Battery Use
The same subscription can behave differently across operating systems
LeleVPN supports Windows / macOS / iOS / Android / Linux, but the same platform name does not guarantee the same traffic capture behavior. Desktop systems may use system proxy settings, a virtual network interface, or an in-app proxy. Mobile systems generally capture traffic through system-provided network extensions or VPN permissions. The client core, permissions, and system policies all affect behavior, so it is not unusual for the same subscription to work differently across apps.
On Windows, check whether the system proxy and virtual network interface are set up as expected, especially for proxy settings left behind by an older client. On macOS, distinguish between system settings followed by the browser and environment settings used by terminal programs. On Linux, the desktop session, terminal environment, and background services often use different proxy configurations. If the browser works but command-line requests fail, first check whether those requests enter the client before investigating the exit or protocol.
On iOS and Android, network changes, background scheduling, and power-saving policies can affect connection recovery. On Android, also check for additional background restrictions from the device manufacturer. On iOS, check the status of the system network extension and whether other networking apps are enabled at the same time. A VPN indicator in the system status does not prove that an app’s traffic is following the expected routing rules; test it with the app you need.
| Platform | Check first | Commonly misread symptom |
|---|---|---|
| Windows / macOS | Traffic capture mode, permissions, and leftover proxy settings | A working browser does not mean every app is working |
| iOS / Android | Background restrictions and recovery after network changes | A disconnect after locking the screen does not necessarily mean the exit failed |
| Linux | Session environment, routing, and service configuration | Terminal proxy settings may not cover background services |
Battery life depends on wireless activity and background behavior
Mobile battery use cannot be blamed on encryption alone. Screen brightness, app content, wireless signal quality, background sync, frequent reconnects, and device temperature all affect power use. A weak signal may require more wireless activity to maintain a connection; frequent disconnects and reconnects also add handshakes and scheduling overhead. A protocol with low theoretical overhead may still deliver worse battery life if it keeps reconnecting on your current network.
Keep-alives can help maintain a session or network mapping, but sending them too often may trigger extra background activity. Connection reuse can reduce repeated handshakes, but it can also make one long-lived session carry more app traffic. Neither has a best setting that applies to every situation. For everyday use, keep the client defaults. If the connection fails to recover after being idle or drops when the screen locks, check system logs before adjusting anything.
When comparing battery use, keep screen activity, task type, access network, and device temperature as similar as possible. Do not compare continuous video playback on one setup with idle standby on another. The client’s share in system battery statistics may include network activity carried through it, not just extra power consumption. More useful things to watch are unusual heat during the same task, repeated reconnects, and sustained background activity.
Troubleshoot background issues without disabling every power-saving option
If Android has trouble reconnecting after the screen locks, check the client’s background activity and battery management permissions in system settings, one at a time. There is no need to remove restrictions for every app. See our Android setup guide for the full installation and permissions process. If the client is still connected but the destination app has dropped, check that app’s own background settings rather than treating the two issues as one.
LeleVPN allows unlimited simultaneously connected devices, but devices may still share a home internet connection and subscription traffic. A computer using all the available access bandwidth for downloads can affect mobile devices on the same network; this is not a device-count limit. See our guide to multiple devices and family sharing for more details. Include shared tasks in your observations when evaluating battery life and connection stability.
Choose and Troubleshoot by Use Case
Browsing, video, and AI tools have different requirements
For everyday browsing, focus on first-page load time, navigation delays, and reliable repeat visits. If downloads are fast but pages start slowly, check DNS, handshakes, and destination response time rather than looking for more throughput right away. For work and long sessions, reliable reconnection and a consistent exit matter more. A route with occasional high peaks but frequent reconnects is usually less useful than one that completes tasks steadily.
Video depends more on sustained throughput and consistent buffering, and is also affected by regional libraries, account conditions, device capabilities, and platform rules. The exit can affect a platform’s regional checks, but cannot guarantee permanent access to content. First confirm that the content is available for your account and region, then check whether playback remains stable. If only a specific platform has an issue, see our streaming access guide rather than assuming every failure is caused by low bandwidth.
Long AI responses, real-time interactions, and file uploads have different session requirements. If a response stops partway through, the cause could be a connection drop, platform-side service issues, usage limits, or an account status change. First check whether other websites are affected, then review the platform’s error message. For connection issues involving Midjourney and Discord, see our guide to choosing a route for AI image generation and troubleshoot the relevant stage of your workflow.
Keep a primary and backup option, and troubleshoot step by step
Start with the subscription’s default route and complete a real task before comparing backup options with different route characteristics. Test the backup in advance rather than waiting until something goes wrong. Avoid frequent automatic exit switching in pursuit of a theoretical best result: some apps treat a change of exit as a change in session environment and may require verification or terminate requests, interrupting your work.
When something goes wrong, first check that the local network is working, the client is running, and it has the required permissions. Then confirm that the subscription is synced and the account is valid. Next check whether the app is using the proxy, whether the right routing rule is selected, and whether DNS resolution succeeds. Only then narrow the issue down to the entry, transport, or exit. If just one protocol in the subscription cannot connect, focus on compatibility and network conditions. If only one website fails, look at the destination service and exit conditions.
After switching routes, start a new request. Old connections may still be tied to the previous session, and a browser may retain an error state or cached response, leading to an unfair assessment of the new route. Close and reopen the affected app if needed, but do not make clearing all account data your default step. Keep a brief record of each change so you can identify what helped. If there is no improvement, restore the previous settings rather than making the configuration more complicated.
Choose the technical setup separately from the plan
A protocol or dedicated-line label does not tell you how much data you need. Choose a plan based on your actual tasks and usage: monthly plans include ¥9.9/month for 60GB, ¥18/month for 250GB, and ¥28/month for 500GB. Data resets monthly on the activation date; mid-cycle upgrades are prorated based on the remaining days. Do not estimate long-term usage from a single speed test or mistake unlimited devices for unlimited data. See the full comparison on the plans and pricing page.
For occasional use, compare the data packages: ¥158/300GB, ¥358/1000GB, and ¥658/3000GB. Data is available until used and never expires. LeleVPN offers a 14-day no-questions-asked refund and supports Alipay / WeChat Pay / USDT. No email address is required; create an account with a username and password. These service details are separate from how a protocol performs on a specific network and do not imply technical or promotional terms not stated here.
For help, use a support ticket in the panel and include your operating system, client, route type, destination app, when the issue occurs, and what happened after changing networks or routes. Do not share your password, full subscription, or unredacted authentication fields. See the help center for common questions. If you have not connected yet, return to the quick start guide and work through the steps in order.