IDS 200 Lecture Notes - Lecture 21: Spell Checker, Eventual Consistency, Google Search Appliance
Document Summary
When a search query is received (and the system has determined it will be handled locally): Google checks if the search results are already available from prior searches. Otherwise, the search is processed from scratch: Many pages may be associated with a particular indexed term. Spelling servers check for likely, alternate meaning. Index servers retrieve document (webpage) ids based on search terms. For english, several hundred thousand terms, each might be associated with many millions of pages. The index will be divided over many servers, each holding some subset of entries & page references. Queries not in the local language will, most likely, be forwarded to a different cluster. When a search query is received, the system: Performs semantic analysis to identify which index entries correspond to the search terms. Forwards the query to the appropriate servers. Although google doesn"t reveal precise system details, we can read its patent documents.