Search Typeahead Q: Alright, how do we shard the data then?
Sorry, not getting how will you split the trie across multiple servers. Will someone be able to elaborate more ? Also, this is discussed here also: http://stackoverflow.com/questions/30281037/how-to-scale-a-trie-across-multiple-servers
Streaming algorithms like sticky sampling and lossy counting should be perhaps used rather than maintaining tries. They will not cause vertical scaling issues like tries and also be much faster. Streaming algos are actually suited for this type of requirements.