Solr scaling multiple instance and shard or multiple solr for different collection

Question

I'm new with solr, and have some case with scaling. I have plan with 2 collection in solr and planning to scale it

• Song Collection (primary for searching). currently ~20m data, indexing 100k data/daily
• Traffic Collection (primary for faceting later), currently ~100m data, indexing 1m data/daily

currently i have 3 solrcloud connect with zk, and have rare case solr down because too many indexing data to traffic collection.

which one is good for performance all, moreover the server only have 3 Server each is 32x3GB RAM and 16 core (shared)

• deploy solrcloud with multiple instance, i.e i have 6 solr connect with zk
• deploy multiple solrcloud and separated collection,
  • 3 solr with zk for song collection only
  • another 3 solr with zk for traffic collection only (this separated solr with previous one)

Or maybe any other solution.

Answer 1

You should look at keeping a single cluster as long as is manageable. But you can improve how you deal with the traffic collection (songs collection sounds ok):

• at this rate it will be too large for a single collection eventually
• show why not index traffic in daily/weekly/or yearlly collections?
• your collections will remain manageable size, and when querying you will be more efficient, as many times you don't need to query the whole history, just last day, last month etc

This is a very recurrent set up.