Why concurrent Hashmap?? | Concurrent Hashmap Internals #shorts #concurrenthashmap #hashmap #java

Añadido: 2026-05-17

[00:00:00]Internal working of concurrent hashmap.

[00:00:02]How is it different from hashmap and what is the concurrency model that it uses under the under the hood? So, hashmap works great in a single-threaded scenarios and but under concurrency it can lead to race conditions, inconsistent state and performance degradation. Con- Concurrent hashmap is designed to solve these problems not by simply adding locks, but by rethinking the internal architecture for the scalability in here.

[00:00:30]Why not we synchronize the hashmap directly like the way that we have map.collections.synchronizedMap and we can pass the hashmap and that's how it will be uh thread safe, right? But there is a problem. Entire map is locked for every operation, no parallelism, high contention under load. So, how concurrent hashmap solves that one pre-Java 8. Map is divided into segments and each segment had its own lock. You see, the concurrent map has different segments there and each of the segment has its own lock. So, if right is happening, so if that is happening on that segment only and that will be locked, not other segments. They will be free to read.

[00:01:09]But there is a limitation. Concurrency is limited by the number of segments, right? But what is the improvement that has happened around the Java 8 plus?

[00:01:16]Segments removed and they replaced they got replaced with the compare and swap.

[00:01:21]That is the mechanism CAS. Bucket level synchronization is there. Lock free uh lock free reads are there now.

[00:01:28]And this is how the internal data structure looks like. Transient, volatile, then we have node of the key and value. That's how the table is there.

[00:01:36]Each bucket can contain node linked list or tree bin red black tree forwarding node during resizing. High level architecture looks like this one. So, concurrent hashmap table is there. The representation that we have seen this table in here, right? So, this will go like this one. Uh bucket 0 1 2. Each bucket will be locked at a time.

[00:01:56]Put operation internals. when you do the map.put in here, hash calculation happens. That is the first thing in here. Bucket indexing is going to happen. Insert logic is going to happen.

[00:02:06]That is where the CAS, compare and swap, no locking will happen in here. So, non-empty bucket in here, this is how it is going to do that one. This is how it is going to handle the collision in here. Bucket node one, node two in here.

[00:02:18]After threshold, the tree bin will be converted like the instead of linked list, that will be converted into the a tree bin in here. So, treefication happens in here. Resizing measure in here. So, this is how you can see the difference between the hash map and concurrent hash map. And huge difference and very important question for the interviews.