Assuming a file sitting on an NVMe drive is structured in a very standard way, e.g. a binary file filled with 8-byte integers: is it faster these days to read those integers into an array in parallel using threads?
I read that NVMe drives are very effective at concurrent, random access reading, but is it to such an extent that reading a file via 8 or so threads would be quicker than reading the file sequentially?
On a regular HDD or older SSD drives, reading non-sequentially would obviously be slower, given the bouncing reader head or whatever causes the bottleneck in the case of older SSD drives. However, is this still the case for NVMe drives, or can we start reading files from NVMe disks in parallel and expect faster results?