@techreport{oai:ipsj.ixsq.nii.ac.jp:02000720,
 author = {Hiroki,Hanada and Keisuke,Ishibashi and Hiroki Hanada and Keisuke Ishibashi},
 issue = {34},
 month = {Feb},
 note = {Timeout configuration is critical for ensuring the reliability of microservices communication. Under normal load, shorter timeouts reduce latency but increase failure rates, while longer timeouts reduce failure rates at the cost of higher latency. Under overload conditions, shorter timeouts can reduce failure rates by alleviating server resource contention. This study proposes an adaptive timeout mechanism that dynamically adjusts timeout values based on real-time metrics. The mechanism minimizes latency and ensures failure rate SLO compliance during normal load by modeling latency distribution using a method inspired by TCP retransmission timeout. During overload, it minimizes failure rates and maintains latency compliance by estimating server capacity and allocating part of the timeout for request delays. Evaluation using Online Boutique benchmark showed an 18% reduction in failure rate deviations and a 25% reduction in tail latency., Timeout configuration is critical for ensuring the reliability of microservices communication. Under normal load, shorter timeouts reduce latency but increase failure rates, while longer timeouts reduce failure rates at the cost of higher latency. Under overload conditions, shorter timeouts can reduce failure rates by alleviating server resource contention. This study proposes an adaptive timeout mechanism that dynamically adjusts timeout values based on real-time metrics. The mechanism minimizes latency and ensures failure rate SLO compliance during normal load by modeling latency distribution using a method inspired by TCP retransmission timeout. During overload, it minimizes failure rates and maintains latency compliance by estimating server capacity and allocating part of the timeout for request delays. Evaluation using Online Boutique benchmark showed an 18% reduction in failure rate deviations and a 25% reduction in tail latency.},
 title = {Adaptive Timeout for Microservices Communication Reliability},
 year = {2025}
}