Find what is actually lagging your server

Three different faults, one complaint#

• Slow ticks. The server's own loop is taking too long. Blocks break late, mobs stutter, everyone sees the same problem at the same moment.
• A bad route. The server is fine; the path between it and one player is not. That player rubber-bands while everyone else is smooth.
• A client problem. One person's own machine or connection. Nothing on the server will fix it.

The quickest test costs nothing: ask whether the stutter is happening for everyone at the same second. If it is, it is the server loop. If it is one or two people, it is route or client. People rarely volunteer this, so ask directly.

Read tick time before changing anything#

On a Minecraft server, the numbers are already there. A healthy server runs 20 ticks per second, and the useful measurement is how long each tick takes in milliseconds — MSPT. Twenty ticks per second means a 50 ms budget; a server averaging 45 ms is near the edge even though TPS still reads 20, and a server averaging 60 ms cannot keep up.

The spark profiler is the standard tool and it is free. Its documentation describes the subcommands plainly:

• /spark tps prints the server's ticks-per-second figures and how busy the machine is.
• /spark health — opens a live report that refreshes itself every 10 seconds; /spark health show prints one to the console instead.
• /spark profiler start then /spark profiler stop — "start the profiler in the default operation mode", then "stop the profiler and view the results".

Two flags are worth knowing. /spark profiler start --timeout <seconds> stops the profiler on its own after a set time, which is the honest way to profile a busy evening without forgetting it is running. /spark profiler start --only-ticks-over <milliseconds> records "only ... samples from ticks which take longer than the given duration" — the single most useful option when the server is fine for minutes and then hitches.

Profile while the problem is happening, not afterwards. A profile taken on an empty server tells you nothing about a full one.

The usual culprits, in the order they usually turn out to be true#

1. One add-on. Nine times out of ten, the profile names it. A single plugin or mod doing work on every tick — scanning entities, hitting a database, walking chunks — will flatten a server that is otherwise idle.
2. Entity accumulation. Dropped items, mob farms, tamed creatures, deployed structures. This is the one that creeps up over a season and explains why a server that was fine in week one is not in week six.
3. Chunk or world loading. Players exploring in different directions force new terrain generation. It shows up as hitching that follows people around the map rather than a steady slowdown.
4. Distance settings set too high. Minecraft's defaults are view-distance=10 and simulation-distance=10. Simulation distance is the expensive one — it decides how much world is actually being ticked. Raising both because a guide said to is a common self-inflicted wound.
5. World size. On Rust, Facepunch's server guide points out for +server.worldsize that a larger map needs more disk space and more of the machine, with a documented usable range of 1000 to 6000. A large map with a small crew is expensive and empty.

Change one variable at a time#

The reason lag investigations drag on for weeks is that people change five things at once, the problem improves, and nobody knows which change did it — so nothing can be undone safely later.

Do this instead. Take a baseline measurement with the server busy. Change exactly one thing. Wait for a comparable session — same time of day, similar player count. Measure again. Write both numbers down. If a change did not help, put it back before trying the next one.

Disabling add-ons is the fastest bisect: turn off half, measure, then half of whatever half was guilty. Four rounds finds one bad add-on out of sixteen.

When it is the route, not the server#

If the tick time is healthy and one player is still rubber-banding, you are looking at the network path. Have the affected player test at a different time of day and, if possible, on a different connection — a phone hotspot is a rough but decisive test. If the problem follows their connection rather than the server, no plan change will fix it.

Distance matters, but it is not the only thing; routing between two providers can be worse than a longer straight line. Treat it as something to measure with your own crew rather than something to promise in advance. Ours run in Beauharnois, Quebec, and the sensible move is to test your own route before committing.

When more resources genuinely is the answer#

Sometimes it is. The honest signs are all of these together:

• Tick time is consistently over budget with the crew online, and drops to normal when they leave.
• The profile does not blame one add-on — the cost is spread across normal world work.
• Entity counts are already under control and the distance settings are already conservative.
• The problem tracks player count, not player location or time of day.

If those four are true, the server is genuinely doing more work than its plan supports, and moving up a tier is the correct answer rather than a purchase you will regret. If they are not, a bigger plan buys you a slightly later version of the same evening. It is a strange thing for a host to write down, but the sale we do not want is the one that does not fix your problem.

Game-specific notes sit on the Minecraft, Rust and ARK availability pages.

Where RadarHost fits#

RadarHost rents game servers hosted in Beauharnois, Quebec, Canada, priced in Canadian dollars, with no setup fee and cancel any time. Sixteen games are listed in Standard, Premium and Diamond plans — Minecraft (Paper, Forge or Vanilla), Rust, ARK: Survival Evolved, Garry's Mod, Terraria, Valheim, Palworld, Enshrouded, Satisfactory, Factorio, Project Zomboid, 7 Days to Die, DayZ, Conan Exiles, Arma 3 and SCUM. Counter-Strike 2 and every other title are request-only and are never priced in advance. After payment, our team reviews the order and emails setup details or next steps within 24 hours. The three tiers and their prices are listed on the plans section, so you can compare before you move up.