The idea is simple: most applications actively work with only a small part of their data. That is the hot set, which needs to be in fast DRAM.
gigaRAM sits a layer below the operating system and keeps hot pages in DRAM while cold pages live on inexpensive NVMe flash.
A predictive engine moves data between tiers ahead of time, so latency stays close to native memory while capacity grows several-fold.
To the operating system this looks like ordinary, but much larger, RAM. No changes are needed in your database, virtualization or analytics.