Myth
Myth
Hi @xiaobiaozhao, I have two questions: 1. Why can multiple disks improve performance? Multiple paths do not seem to work in parallel. 3. How do we judge hot and cold...
> In fact, the level at which SST is located represents the hot and coldness of the data. Because rocksdb uses the LSM tree, it will naturally merge cold data...
> I am interested in this problem. Could you assign the issue to me? @AntiTopQuark Sure, go for it!
I prefer a simpler approach, count the size of the output buffer each time you put something into it. Of course, it's better to use a memory allocator.
@PokIsemaine I suggest that before you start writing code, outline your solution and let's discuss it together.
I agree with this way.
I prefer 1-SlotMigrationJob,1-Slot Range. We can seek only once for a range (the same namespace), making the changes relatively small and making it easier to determine whether a slot belongs...
I think ok, just pass a valid slot range. We can extend it in the future to make it more diverse.
> how do we handle failure situations? The migration slot range is migrated from front to back, if a slot fails, `[start_slot, failed_slot)` should be successful, `[failed_slot, end_slot)` should fail....
I want to discuss how we can implement this feature. My idea is to count client memory within each worker (using `libevent`'s callback to count `evbuffer`) to avoid locking across...