For beginners who have imported a subscription but aren’t sure which server to choose. First rule out servers that won’t connect or don’t meet your exit-location requirements. Then compare response times, failure rates, and data usage for your actual tasks. Keep one everyday server and one backup on a separate route.
Latency: Check what’s being measured before comparing milliseconds
Latency in a server list isn’t a single, standardized metric. A TCP test typically measures how long it takes the client to establish a TCP connection to the server’s entry point. A reachable port only means the entry point accepts connections; it doesn’t confirm authentication or that the exit can reach your destination. A real-connection latency test sends a request through the server to a test website, so results also depend on the test site, connection setup, and current network conditions. Latency is not download speed.
Choose what to test
These menu steps use v2rayN 7.12.7 on Windows as an example. In the main window, switch to the subscription group you want to test and select two or three candidate servers. Keep your local network unchanged during testing; don’t switch from wired to Wi-Fi halfway through. Menu labels may vary slightly between versions.
Compare connection latency
Right-click the server list and select “Test server real connection latency.” Test each candidate three times and note the successful readings and timeouts separately. If all three tests succeed, compare the median. If a test times out, record it as a timeout—not as zero milliseconds in an average.
Switch servers and verify
Right-click a candidate and select “Set as active server.” In a browser configured to use the proxy, visit a site you actually need and confirm the request wasn’t routed directly by a routing rule. A highlighted row in the server list doesn’t prove the browser is using that server.
Retest with real tasks
Open the same page or download the same authorized test file through each candidate in turn. Avoid browser cache during downloads, and don’t run speed tests in parallel. If you usually connect in the evening, test again then; a single daytime result won’t reflect peak-hour conditions.
For example, server A records 82 and 86 ms, plus one timeout; server B records 110, 114, and 117 ms. These figures are illustrative only. For web browsing or remote terminals, consistent successful requests matter more, so try B first. For large downloads, compare sustained throughput separately—these few latency readings aren’t enough to pick a winner. If every candidate times out, check the test site, core logs, and local network before removing the subscription.
Traffic multipliers: Estimate how much of your plan they use
A traffic multiplier is a billing factor set by the provider, not a bandwidth multiplier. If the provider deducts traffic by multiplying actual usage, the same 10 GB transfer can consume different amounts of your allowance on different servers. The examples below use the same units and are not provider rates or measured results.
- Check what counts toward usage. If your plan adds uploads and downloads together, 8 GB downloaded plus 2 GB uploaded counts as 10 GB. Looking only at browser downloads can underestimate usage from sync and backup tasks.
- Check when the multiplier applies. A “low multiplier” in a server name may apply only at certain times or to specific plan tiers. Follow the provider’s billing rules. After switching, test with a small file and allow time for the usage dashboard to refresh.
- Choose a separate route for data-heavy tasks. At 30 GB of billable traffic per week, a 1× multiplier uses 30 GB of your allowance; 2× uses 60 GB. If both routes meet your needs, the lower multiplier is a better fit for regular use. There’s no need to pay for a slightly lower latency with extra data on every transfer.
Exit location: Verify server labels against the service you need
A location in a server name may refer to its entry point or its exit. The destination site usually sees the exit IP. With a relayed route, the entry and exit can be in different regions, and IP geolocation databases may show different locations for the same address. To verify a location, check the exit through an app using the proxy, then confirm the target service reports the region you need.
Choose a location and route for the task
Services that require a specific region
- For example, a site with a region-specific verification page. First choose servers whose exit location meets the requirement, then compare success rates among candidates in that region.
- Results may also depend on account region, browser language, and existing sessions. Keep your sign-in conditions consistent so you don’t mistake account differences for the effect of changing servers.
Everyday tasks with no location requirement
- For web browsing, documents, and remote terminals, focus on response-time consistency during peak hours. Nearby servers are worth testing first, but geographic proximity doesn’t guarantee fewer network hops.
- For sustained downloads and sync, prioritize consistent throughput and the traffic multiplier. If your requests use UDP traffic, such as voice calls, verify separately that both the app and route support it.
If a specific region is required, filter by exit location first. Otherwise, rank servers by task success rate, sustained performance, and traffic cost.
Protocol type: Check core compatibility before comparing routes
VMess, VLESS, and Trojan are protocol types, not speed tiers. A server also has transport, security, and server-side settings. Switching protocols often means switching servers or routes too, so differences between two server tests can’t be attributed to the protocol alone. Use a combination already in your subscription and supported by your local core. Changing one protocol to another in the editor won’t make it faster.
VMess with WebSocket and TLS
- User ID
- UUID assigned by the server
- Transport
- WebSocket
- Security
- TLS
- Path
- Must match the server configuration
This is a common combination. If the handshake or connection fails, also check that your device’s date, time, and time zone are correct.
VLESS with TCP and REALITY
- User ID
- UUID assigned by the server
- Transport
- TCP
- Security
- REALITY
- Flow control
- If the server uses Vision, your settings must match
This combination requires an Xray core that supports the relevant features. Being able to recognize a VLESS entry doesn’t mean the core can run the full configuration.
Trojan servers typically use password authentication with TLS. If the handshake fails, first check whether the subscription has expired, whether the server name matches the configuration, and whether your device’s clock is correct. Disabling certificate verification is not a fix. On desktop, check the core used for the protocol in v2rayN under “Settings” → “Parameter Settings” → “Core Type.” On Android, v2rayNG uses the Xray core, while v2flyNG uses the v2fly core. Don’t assume they support exactly the same subscription entries just because both can import the same subscription.
Final checklist: Keep one primary server and one backup
- Connects reliably: The server passes a real-connection test, and actual requests use the expected outbound route. Don’t make a server the primary choice if only its entry port is reachable.
- Works at peak times: Record the test time, local network, candidate servers, median latency, and failure count. For important tasks, also note sustained download speed or interruptions, so you can retest under the same conditions next week.
- Usage stays within budget: Note the actual multiplier and what counts toward usage, then estimate your weekly traffic. Move regular backups or sync tasks to a lower-multiplier route that meets your needs.
- Backup uses a different route: Where exit-location requirements allow, choose a server with a different entry point or route. Two servers with different names may still share an entry point and fail together if their upstream network has an issue.
After a subscription update, server addresses, multipliers, and supported protocols may change. Recheck your primary and backup servers on the day of the update; there’s no need to retest the entire list. Your notes can include locations and test results, but remove subscription links, passwords, and user IDs before sharing logs.