You need multi-core because you cannot make stops and stop words in a multilingual database.
Common stop words in English are "by" and "is", but these words mean "city" and "ice" in many Scandinavian languages.
If you use multicore, each language can be on its own kernel with a custom schema.xml that selects the correct stemmer, stop words and protected words. But the same JVM runs all this on one server, so you do not spend extra money on servers for one specific language. Then, if the load is too large for a single server, you replicate your multi-core configuration and all indexes benefit from replicas.
source share