Consistent Hashing - Rebalancing Partitions | Systems Design Interview 0 to 1 with Ex-Google SWE

They say if you've been putting too much load on one damsel it's a good sign to shard that and get another to help balance the stress

[@MrDartfire]

Jordan has life of the great big sense. Thanks a lot for all of these videos!

[@aleksandr_t]

Great work, thank you for that video

[@user-rv5yb9kj6w]

How are we choosing the K locations that belong to Nodes?

[@jordanhasnolife5163]

Uniform random distribution sorry if that wasn't explicit

[@LeoLeo-nx5gi]

How can we avoid hot partition issues in consistent hashing if some nodes are removed ? Also this was quite cool video thanks!!

[@jordanhasnolife5163]

Well generally speaking hot partitions will come up when one key in particular is hot which means it's pretty darn hard to avoid - at that point you have to kinda be clever about sharding even deeper within that key! Great question

[@LeoLeo-nx5gi]

@@jordanhasnolife5163 makes sense!!

[@htm332]

🔥

[@aloha9938]

so from what i understand, consistent hashing is taking all the shards(partition) and further sharding each of them into k shards per shard?

[@aloha9938]

nope, the shards are not literally sharded again, they are hashed into k keys and placed on the ring in uniform distribution, you can think of them as virtual nodes, i think

[@aloha9938]

thanks

[@jordanhasnolife5163]

Nice