Supercharging the Elasticsearch Percolator

#1

Post by traxmaxx » Thu, Oct 15, 2015, 11:54 AM UTC

Supercharging the Elasticsearch Percolator
underthehood.meltwater.com

Re: Supercharging the Elasticsearch Percolator

#2

Post by petard » Thu, Oct 15, 2015, 3:36 PM UTC

What are your memory settings for utilizing the 360GB RAM with ElasticSearch?

Re: Supercharging the Elasticsearch Percolator

#3

Post by rchakra3 » Thu, Oct 15, 2015, 5:15 PM UTC

Can you explain how a positional query is a subset of boolean queries?

Intuitively it feels like the reverse would be true.

Re: Supercharging the Elasticsearch Percolator

#4

Post by alxnlssn » Thu, Oct 15, 2015, 5:16 PM UTC

petard wrote: post #2

What are your memory settings for utilizing the 360GB RAM with ElasticSearch?

Currently running with three ES data nodes per physical machine, with 30GB -Xmx allocated to each instance.

Re: Supercharging the Elasticsearch Percolator

#5

Post by bergetp » Thu, Oct 15, 2015, 6:50 PM UTC

rchakra3 wrote: post #3

Can you explain how a positional query is a subset of boolean queries? Intuitively it feels like the reverse would be true.

You're right, but the point here is that positional queries are boolean queries plus positions. For example, document that match 'a NEAR b' are a subset of documents matching 'a AND b'. So if 'a AND b' has no matches, there is no need to execute the more expensive 'a NEAR b'.