In e-commerce, we use a lot of search requests that could be forwarded by the coordinator node to all Elasticsearch tiers (hot, warm, cold, frozen). Thus, the frozen node will penalize the response time since we have to wait ,each time, for its response which will be the longest. my question, doens the previous architecutre is recommanded in the case of e-commerce and if not what is the recommanded architecture, in case we have a load that requires 10 data nodes.