When and How to Upgrade Your Server
Learn the signs your server needs more power and upgrade without downtime.
By Lodgichost Team · Last updated: August 19, 2026
Every Minecraft server eventually hits a resource wall. The signs are consistent: TPS dips below 20, MSPT creeps past the 45-millisecond-per-tick target, memory pins at the plan ceiling, CPU pegs during peak hours, and join spikes are greeted with lag. Upgrading fixes these — but only if you buy the right resource. RAM helps more players and heavier modpacks, CPU cores help parallel workloads like async chunk generation, and storage speed helps pre-generation and schematics. This guide explains how to read those signals, what each resource scales, and how to upgrade on Lodgichost without players noticing.
Signs It Is Time to Upgrade
Measure before you spend. On Paper, /tps shows ticks per second — a healthy server holds 20.0, and any sustained drop means the main thread is overloaded. MSPT, milliseconds per tick, is more precise: one tick is a 50 ms budget, and staying below 45 ms is the target for headroom. When the memory graph is pinned and GC is thrashing, RAM is the constraint. When CPU is pegged but TPS is fine, the bottleneck is usually view distance or redstone, not raw power. Collect readings over a few days at peak time before deciding.
[ ] TPS dips below 20 during peak hours for 3+ days [ ] MSPT averages above 45 ms per tick at peak [ ] Memory pinned at the plan limit; GC pauses over 100 ms [ ] CPU pegged while TPS stays low (check view distance first) [ ] Join spikes at known times are followed by lag
What Each Resource Actually Scales
Buying the wrong resource wastes money and does not fix the lag. RAM is the first upgrade for most servers: it raises the ceiling on player count, view distance headroom and modpack memory, but does not directly speed up ticks. CPU cores help what parallelises — Paper's async chunk generation, plugin thread pools and pre-generation — though the vanilla tick loop is mostly single-threaded, so one fast core matters more than several slow ones. Storage speed shows up when you pre-generate chunks, paste schematics or restore backups, since those are disk-bound. Add RAM when join spikes hit the memory ceiling, upgrade CPU when chunk generation is the limit, improve storage when disk operations stall.
- RAM — player count, view distance headroom, modpacks.
- CPU cores — async chunk generation, plugin parallelism and pre-generation.
- Storage speed — pre-generation, schematic pastes and backup restores.
- None fix a badly configured server — check view distance and plugins first.
Why RAM Helps Players More Than Tick Speed
Extra RAM stops helping tick performance beyond a threshold. A vanilla server runs comfortably in 4 GB and most townships in 8 GB; 16 GB over 8 GB means more players, chunks and modpacks at once, not faster ticks. Tick speed depends on the CPU's single-thread performance and the work the main thread does per tick. So low TPS with memory to spare is a view distance, redstone or plugin problem, not a RAM problem. Hitting the ceiling while TPS stays healthy is the pure RAM upgrade case.
Upgrading Without Downtime
On Lodgichost, upgrades happen through the panel with your world, plugins and IP untouched. Choose the new plan and confirm — the panel restarts the server on the new resources. Your IP stays the same, so players reconnect as before. Time the upgrade for off-peak hours and warn your community first. Before upgrading, check your startup flags stay valid — a plan change can reset the memory allocation, and you want -Xmx matching the new total.
- Choose the new plan in the panel and confirm.
- The server restarts on new resources; IP, world and plugins are unchanged.
- Verify the startup RAM flag matches the new plan's total.
- Warn players and run the upgrade at off-peak time.
Migrating to a Different Host
If you are moving to a bigger server on another host instead of resizing, the safe path is backup, transfer, verify. Create a full backup through the panel — or stop the server and copy the folder via SFTP — transfer it, point server.properties at the world and start. A RAM-only change does not touch mods or plugins, so you normally do not need to reinstall anything; just confirm the startup flags and that the world folder name matches level-name. Keep the old server running until the new one is verified.
Troubleshooting
- Upgraded RAM but TPS is still low: RAM was not the bottleneck. Low TPS with free memory points at view distance, redstone or a plugin, not memory.
- Server restarted but still shows the old memory: a hard-coded flag like -Xmx4G overrides the new plan. Update the JVM arguments in the panel and restart.
- Players report disconnects after the upgrade: it applied during peak hours. Schedule upgrades off-peak and broadcast a warning before the switch.
- Migration lost my world: the world folder name does not match level-name, or the transfer path was wrong. Restore the original backup and verify before starting.
FAQ
How do I know if I need more RAM or CPU?
Watch panel graphs at peak time. If memory is pinned and GC pauses grow, add RAM. If TPS is low while memory has headroom, the main thread is overloaded — that is CPU work, not RAM.
Will upgrading reset my world or plugins?
No. A plan upgrade through the panel only changes allocated resources; your world, plugins, configs and IP stay as they are. The server simply restarts on the new plan.
What is a good MSPT number?
One tick is 50 ms, so staying below 45 ms per tick leaves headroom at 20 TPS. Sustained averages above that mean the main thread is near its limit.
Can I downgrade later if I don't need the resources?
Yes, move between plans in the panel at any time. Just confirm your world fits within the new plan's memory before you downgrade.
Does more RAM mean better TPS?
Only up to a point. Once the server has enough memory to run its load comfortably, tick speed is determined by CPU and how much work the main thread does per tick. RAM helps player count and modpacks more than raw tick rate.
Related Guides
- Optimising Server Performance — fix lag before you spend money
- Scheduled Server Restarts — flush memory leaks cheaply first
- Reading Server Logs — read the performance data yourself
