This is the basic core of a sort-merge. It depends if the sorted lists are on disk (usually are with each about as big as memory), and how many you merge at onc

This is the basic core of a sort-merge. It depends if the sorted lists are on disk (usually are with each about as big as memory), and how many you merge at once (which is about dividing all available memory into disk buffers, or the disk bandwidth, whichever is the bottleneck), and merging the memory resident lists using a treelike structure to be compute efficient. Also depending on the OS it can take parameter twiddling to achieve full sequential bandwidth. If the lists are on other nodes, like in a petabyte sort, that’s where it gets complicated because you need to recover from node failures during the sort.