What would be the best Amazon instance size for a Solr collection with about 14M small documents, long tail queries, 2 shards, one on each server?
Index size is currently 360 GB.
Not sure what’s more important: lots of RAM and CPU, or a fast disk (EBS IOPS etc)
Typically the best performance in Solr is achieved when the entire index can be loaded into RAM on the server. So, once you determine the physical size of your index, if you can put that amount of RAM, plus whatever the operating system will need, should give you the best index performance. If you cannot fit it all on one server, then you might consider splitting the index across multiple shards. The fast disk will help with performance when updating the index.