CODEWITHSUNDEEP
optimizationsolrsolaris
GabrielCol
GabrielCol

Reputation: 181

SOLR One collection (core) VS. many

I have multiple entities from a MySQL database that will be indexed in SOLR.

What is the best method in order to have the best performance results (query time)?

Thanks

Upvotes: 4

Views: 1500

Answers (1)

Srikanth Venugopalan
Srikanth Venugopalan

Reputation: 9049

I would add a few more parameters for you to consider (mostly discouraging one core per entity approach, but not just for performance reasons that you are specifically asking for)

  1. More cores would mean more endpoints. Your application will need to be made aware of such. And you may find it difficult to run a query across cores. For ex, if you are searching by a common attribute, say name, you would have to run multiple queries to each core and aggregate the result. And this will miss the relevancy aspect that you get out of the box in querying a single core.

  2. Consider making minimal requests to your database. N+1 jdbc connections drastically slow down indexing. Instead, try to aggregate your results in a view and if you can fire a single query, your indexing will be much faster.

  3. Range queries on common attributes will not be possible across core. Ex - if you have price of Books and Music Cds stored in different cores, you can't get all products between X and Y price range.

  4. Faceting feature will also be compromised.

So, while you may perceive some index time performance gain by parallelizing in form of 1 core per entity, I feel this may reduce the features that you can benefit from.

Upvotes: 5

Related Questions