For connections that work but have slow downloads, evening slowdowns, or sluggish page loads. Use the same destination to distinguish initial response time from sustained throughput, then test different nodes and network connections separately. Check local proxy settings, simplify custom routing, and investigate DNS only when the results point to your device. Identify which change affected performance rather than settling for a node that works temporarily.
Set a baseline: distinguish initial response time from download speed
A website that feels “slow” can mean two different things: it takes a long time to respond after you open it, or it responds quickly but downloads remain slow. Time to first byte includes connection setup, handshakes, and time spent by the destination server; whether DNS lookup time is included depends on where the lookup happens. It is not simply the node's round-trip latency. Sustained throughput measures how much data arrives per second once transfer begins. Track both so a slow initial page load isn't mistaken for a bandwidth problem.
- Choose a fixed download that you’re authorized to access and that’s large enough for testing. Use the same URL and download method in every round. Make sure the data isn’t coming from the browser cache or resuming a previous download. Pause cloud sync, other downloads, and system updates during testing.
- Test the same destination directly first to establish a baseline, then access it through the proxy. If the destination isn’t reachable directly, choose a separate destination that is, and use it only to check for problems with your local connection. Don’t compare its direct speed with proxy speeds to a different destination.
- Record the time of day, network connection, client and core versions, selected node, actual outbound route, time to first byte, and sustained throughput. Use the median for each set and note the range. If your download tool reports MB/s, multiply by 8 to convert to Mbps; for example, 4 MB/s is 32 Mbps. If a small file finishes before the stable-transfer window, use a larger test file rather than treating peak speed as the average.
Layer 1: Node performance and exit routes
To compare nodes, keep the device, network, and destination fixed, then test two similar nodes—for example, servers in the same region using the same transport. This reduces variables related to location and transport. Confirm that each round actually uses the node you selected manually. If the configuration includes automatic routing or fallback, check the actual outbound route; the interface may show node A while the request is handled by a different route.
Example: alternating between nodes for speed tests
| Outbound | Median time to first byte | Download speeds across three rounds |
|---|---|---|
| Node A | 380 ms | 8、9、10 Mbps |
| Node B | 420 ms | 30、32、33 Mbps |
Next variable to test: network connection
In this example, node A has a faster initial response but consistently lower throughput. Keep A selected and retest on a different network. If A improves while B stays about the same, investigate the route from the original network to A before changing local DNS settings.
If A is slow on both networks, try a second download destination. If several destinations are slow, check for rate limits, high load, or limited exit capacity on the server hosting A. If only one destination is slow, consider the route from the node to that destination and any rate limits imposed by the destination server. Without server-side metrics, report that “access through A is consistently slow” rather than concluding that the server is overloaded.
Layer 2: Network connection, time of day, and transport
Keep the node fixed and switch the same device to a different network—for example, from home broadband to a mobile hotspot. Hotspot testing uses mobile data, so if you need to limit usage, use the same small-sample method on both networks. If possible, retest your home broadband over Ethernet. If Ethernet restores speed but Wi-Fi remains slow, investigate the local network. If both are slow but the hotspot works, check the broadband connection or the route beyond it.
- Compare times of day. Run one set of tests during peak evening hours and another off-peak. If several nodes slow down only in the evening, investigate time-based congestion or a shared exit first. The time of day alone can’t tell you whether the bottleneck is at the ISP or on the node side.
- Look for signs of packet loss. If the node responds to ICMP, record the round-trip times and packet loss across 30 probes during both fast and slow periods. If an intermediate hop doesn’t respond but later hops are stable, don’t count that hop’s timeouts as packet loss affecting your traffic. If the node doesn’t respond to probes, focus on connection timeouts and download variability instead.
- Compare supported transports. Only compare transports, such as TCP (labeled RAW in some versions) and WebSocket, if the server already provides two working endpoints. Use the published port and the corresponding TLS, SNI, path, and other settings for each. If the WebSocket endpoint also passes through a reverse proxy while the other connects directly to the server, record the result as a difference in the entire route—not as an effect of the transport alone.
Don’t compare VMess, VLESS, or Trojan directly with TCP or WebSocket: the first group consists of connection protocols, while the second consists of transports. Changing WebSocket to TCP in the client alone won’t switch the server configuration. A test in that case measures a failed connection, not slower speeds.
Layer 3: Proxy entry point, routing, and DNS
If the same node works normally on another device on the same network, or only one app is slow, focus on the local device. First confirm that requests are reaching the client, then simplify custom routing, and finally use the logs to investigate DNS. Keep the original settings and change one thing at a time. If the problem can’t be reproduced, restore the original configuration and keep monitoring.
- Check the local listener and app proxy settings. In the desktop version of v2rayN, check the core’s local listener under “Settings” → “Parameter Settings”; the exact fields depend on your version. If the connection details show
127.0.0.1:10808, also make sure that endpoint supports the proxy type used by the app. If both the system proxy and virtual network adapter are enabled, first find out which route the test app uses, then test each mode separately. - Temporarily disable extra routing rules. Back up or copy the current profile, then set an explicit outbound route for the test destination. Start by disabling recently added broad domain or IP rules while keeping necessary exceptions, such as local network ranges. Check the match result or outbound logs for the same URL, if your version provides them. If disabling a rule changes the route from direct to proxied, the speed difference reflects a change in route—not the amount of overhead removed by deleting the rule.
- Adjust DNS based on where the failure occurs. Distinguish the node’s hostname from the destination website’s domain. If the node hostname fails to resolve, the client may not have connected to the node yet. The destination domain may be resolved by the app, the local core, or a remote resolver. Change the resolver responsible for that stage only when logs show a lookup timeout, resolution error, or unexpected resolution path. After changing it, create a new connection and retest the same domain. If it resolves to a different address, note that change instead of attributing the entire improvement to lookup speed.
Check routing and DNS against the routing rules and DNS settings guides. Changing the resolver usually won’t improve the speed of a download that’s already connected and transferring slowly. If a new request is faster, it may have connected to a different server.
Log errors: identify where the failure occurs
Use the timestamp of a retry to find the underlying error for that same request. The top-level message often only says that an attempt failed; details such as a hostname, port, or operation name can help distinguish listener, DNS, and connection problems. Before sharing logs, redact subscription URLs, credentials, and sensitive request parameters.
Common errors and what to check
Error: failed to find an available destination
What it means: This top-level message indicates that an outbound attempt couldn’t find an available destination; it doesn’t necessarily mean DNS failed. What to do: Check the underlying error in the same log entry. If it reports that the node hostname couldn’t be resolved, check the node address and DNS resolution path.
Error: connection refused
What it means: Nothing is listening on the destination port, or the connection was actively rejected. What to do: Use the address in the error to determine whether the problem is local or remote. For a local address, check that the core is running and listening on the expected port. For a remote address, verify the port provided by the server.
Error: i/o timeout
What it means: A network operation exceeded its time limit. This can happen during DNS lookup, connection setup, or data transfer. What to do: Check whether the log identifies a lookup, dial, read, or write operation, then retest the same node on another network. Don’t simply increase the timeout.
Common questions about speed tests and troubleshooting
Latency, proxy coverage, and retesting
Why is my download still under 2 MB/s when the client reports 80 ms?
First check whether the test measures connection latency or response time for a probe URL; neither tells you the download speed. Use the same file for three alternating download rounds and compare stable throughput. If a low-latency node consistently delivers only about 16 Mbps, choose a node based on throughput instead.
Why is my download tool slow if websites load in my browser?
Check whether the download tool uses the system proxy, has its own proxy configured, or connects directly. Make one short request from each setup, then check the client’s connection log for the expected destination and outbound route. Retest against the same destination.
Can I leave custom routing disabled if speeds improve?
Restore the original profile, then re-enable recently changed rules one at a time until the same destination slows down again. Fix that rule’s match scope or order, then retest both the test destination and one that should remain direct. Make sure both use the expected routes.
Should I switch nodes if a page is slow only on its first load?
First distinguish connection reuse from DNS caching. Measure the initial response on a new connection and compare it with a reused connection. If the logs show an unusually slow DNS lookup, investigate resolution. If DNS is fast but the connection or handshake times out, compare nodes and network routes.