Basically, this should be the same in-memory function of redis but instead stored on disk and syscalls are disk calls.
Well, technically, the on-disk redis is not for performance purposes, but rather, just for storage purposes where redis can act like NOSQL database.
In this manner, users can design web applications with in-memory and on-disk capability in one unified redis interface.
Possible approach: - Since redis is single threaded, disk calls should not block mem-calls, so when on-disk redis is enabled, another independent thread that handles disk calls should be launched @ configure start. Dev can opt to permit internal communication between on-disk and mem-redis, but I believe this is too complicated. My take is on-disk and mem-redis should be atomic, exchanging data should be done within the web app instead, as what is typically done now. - If the independent thread for disk and mem-redis is not favorable, on-disk redis should be lunched as an independent server. - There are also other non-blocking approach but I am not familiar of those.
Comment From: itamarhaber
Hello @jobs-git
It is not clear what feature you're asking for, or the type of issue you're experiencing. Please provide more information to make this into something that's actionable.