...
[2] "Configuring and tuning HPE ProLiant Servers for low-latency applications": https://www.hpe.com/h20195/v2/GetPDF.aspx/c05281307.pdf
[3] https://downloads.linux.hpe.com/SDR/project/stk/
...
I haven't done any rigorous benchmarking but this ska::flat_hash_map (https://probablydance.com/2017/02/26/i-wrote-the-fastest-hashtable/) is clearly much faster than std::map, std::unordered_map, sparsepp (https://github.com/greg7mdp/sparsepp) when the map is relatively small (e.g. ~20 entries). I am now using this map for incomingRpcs, outgoingRpcs and MessageAccumulator::fragments in HomaTransport.