Improving Response Latency in Elasticsearch with Adaptive Replica Selection | Elastic Blog