In the table that describes multi-dimensional scaling, it says that the hardware requirements for the Index service are: basic processor, less memory, SSD. Does this mean that indexes are only stored on disk? If they are not only stored on disk, how is memory / disk used for the indexes (e.g. is a part of the index stored in memory, or is it only stored in memory if the whole index can be stored in memory.
@Eugaia indexes are stored both on dis and in memory. They use the underlying ForestDB storage engine, which takes care of that. Every cluster has an index memory quota and it will try to use as much ram as you want to give it to store indexes in RAM so access is fast.
That’s what I assumed, but just wanted to make sure.
Assuming there is not enough RAM to store an entire index in memory, what is the decision process on what data is stored in memory / not? I’m assuming that it somehow involves the most frequently used data, but could you explain in what way?