Keep the world correct.
Trace differences to the stage where they first appear. A limited matching test is useful evidence, but it does not prove universal equivalence.
Independent development · Minecraft & compute
Can another PC
help a Minecraft
server?
We’re developing RemoteCompute to find out which server calculations are worth moving to another computer. The local Java–Rust integration works; the performance benefit still needs to be demonstrated.
Meet RemoteComputeCurrent work: profiling & world-difference diagnosis
01Current project
Research & developmentRemoteCompute
Separate compute for Minecraft servers.
RemoteCompute investigates whether selected Minecraft server calculations can be handled by separate workers, while the server retains control of the world.
The goal is practical: understand which workloads are worth moving, what the transfer costs, and whether players ultimately benefit.
See where the research standsRESEARCH CONCEPTRemote execution is a development goal.
Minecraft server
World state & final decisions
Compute worker
Selected calculations
Local Java–Rust integration testedRemote path planned
Moving a calculation is only useful when computing, transferring, checking, and applying the result beats the baseline.
02How we work
Three checks before
moving more compute.
We compare against Java, check whether the same workload finishes, and trace changes in the world. These checks decide what we build next.
Count the entire cost.
Measure preparation, copying, computation, validation, and application. A fast calculation does not automatically mean a faster server.
Separate the questions.
Test complete-work throughput separately from overload behavior. Later, verify remote connectivity and performance gains independently.
03Research notes
EXPERIMENTAL / CURRENT FINDINGSWhat works.
What still needs work.
The initial experiments established a real integration path. They have not yet established a speedup or production readiness.
A real server integration
A Paper server was confirmed to consume Rust computation results. This establishes that the integration path can work.
Limited output agreement
Outputs matched in 12 limited chunk-generation tests. Differences in normal world simulation remain under investigation.
Performance is still an open question
The measured Rust execution path took about 57% longer than the Java baseline at the median. Profiling comes before further optimization.
Read the experiment context
These findings describe an experimental Linux setup reported for Paper 26.2 build 132 and Java 25. They are not a general compatibility or performance guarantee.
The offered-load test completed 53–59 of 64 requests under the existing admission limits. The Java baseline also rejected requests under those limits. Request rejection must be distinguished from calculation errors, and a separate test must measure how quickly the same complete workload finishes.
The initial block differences and later water or leaf differences seen with Rust disabled have not been shown to share a cause. The next investigation tracks when each difference first appears.
About MineMoru
MineMoru · 마인모루
MineMoru is an independent software development team working on Minecraft server technology. RemoteCompute is a project within the team. This page shares its development goals and the current experimental findings.