Versions (relevant - OpenSearch/Dashboard/Server OS/Browser): OpenSearch 2.17
Describe the issue: I see that when using lucene engine with hnsw and using a disk based index (4x compression) the total size of the index is exactly same as the index which is not disk based. Is this possible, I am unable to understand why and how this might happen?
Related information - I have 768 dimension embeddings and around 90 million documents in the index. I have 60 shards having (27.5GB per shard)