Difference between Bucket level lock and segment level lock in ConcurrentHashMap?

Question

I gone through below link and some online video tutorials but am unable to find exact differnece between Bucket level lock and segment level lock in ConcurrentHashMap.

ConcurrentHashMap

Answer 1

In Java 8 ConcurrentHashMap changed its logic to use bucket level locks instead of segment level locks.

What is the difference?

In Java 7 the map was subdivided among Segments, each of which itself is a concurrently readable hash table. Each segment contains some number of buckets inside.

The bucket itself is a list/array of key-value pairs.

The number of threads working with map in parallel was limited to the number of segments.

In Java 8 lock level was moved to a bucket level, while segments were removed (it is still in the code, but without use).

So, basically, in Java 7 inserting was locking a few buckets, in Java 8 - only one.

Another reason why it was done is that the number of threads, which are working in parallel with the map, can increase during map size increasing (more buckets - more threads). In Java 7 segments size was fixed, and it couldn't be changed without recreating a map.