Evgeny Barskiy
Evgeny Barskiy
> It's great that this can be done by IO threads, but we should not forget single-threaded mode here. You get the most out of the CPU if scale horizontally...
> Instead of creating a new empty table, we maintain a a special type of iterator on the slot we are iterating. The iterator goes through each key in the...
I'd say some kind of centralized control flow among full_sync/move/defrag etc makes sense, so we can prioritize slots and don't overwhelm service by copying too many slots in parallel.
> What evidence is there to suggest that COMMAND > > > is expensive to process (too much string operations) > > How meaningful is the benefit of caching this...
Regardless the performance: Internally, with default valkey it takes 2ms (little to no variance) to regenerate a result. Add bunch of modules (with extra commands) and it shoots up to...
https://github.com/valkey-io/valkey/pull/2839