A Fleet That Fixes Its Own Ships express
A fleet of many ships, like a galaxy of stars, carries out a common mission. Sometimes a ship explodes, like a supernova. If not replaced in time, the fleet weakens. But the trouble is we can't see which ship will fail next — failures are invisible, like dark matter. Their accumulation leads to disorder, entropy. It's crucial to get ahead of chaos.
The usual method — the standard model of scheduled repair — doesn't account for randomness. The new algorithm makes each ship anticipate its neighbors' moves and independently decide when to repair. The fleet gets by without a captain: the decision emerges from interaction. Tests showed: for a squadron of 70 ships, costs are a quarter lower than with old approaches.
Original abstract
as written by the authors arXiv:2607.11187
We study recovery control in replicated computing systems. Such systems consist of replicas that collectively provide a service to a client population. This redundancy enables the system to withstand failures provided that failed replicas are recovered faster than new failures occur. We show that the problem of deciding when to initiate recovery of selected replicas can be formulated as a partially observable Markov decision problem (POMDP) with a multiagent structure. We exploit this structure to apply a multiagent rollout method for approximating optimal control policies. Our method uses precomputed signaling information that reduces the need for replica coordination and facilitates parallel computations. Experiments show that our method scales to systems with up to 70 replicas and reduces costs compared to the recovery policies currently used in practice.
In large data centers, server failures are business as usual. Hundreds of hard drives fail every day, but thanks to redundancy, we don't notice the outages.
In the series 'Star Trek,' Federation ships could automatically restore their shields. This work brings regular servers closer to such self-repair.