I have got 44 .ttl files with the total memory of 117 GB. I'm curious about the efficient method for executing SPARQL query for all .ttl files in real time. I tried to combine them, but loading the model takes a lot of time in Apache Jena as well as combinations. So, what should be the most efficient way?
Load them once into a triplestore such as Jena's TDB. You can do it into one big graph or into separate named graphs depending on needs.
This saves all the data parsing time when you want to execute a query.