chenboat commented on PR #16364: URL: https://github.com/apache/pinot/pull/16364#issuecomment-3111519330
In the test @deemoliu mentioned above, we try to do regex (mostly wildcard) over a corpus of metric tags for time series data. Each time-series point usually has about around 7 tag/value pairs. We model them as multiple value columns. We find ngram length of 3 is really good tradeoff between space and search speed-up. As mentioned above, most of the user queries are relatively short wildcard search like (*dca*, *dca11*). Ngram index with our inverted index implementation did not result in storage space explosion (because the number of unique 3-grams are actually fairly limited) and search performance is 10x better than Lucene based FST term text index. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
