Github user andrewor14 commented on the pull request: https://github.com/apache/spark/pull/1165#issuecomment-49685219 @mateiz Yes, currently we unroll it in deserialized form even if we only want to store it in serialized form. One issue with storing it directly in bytes is that `CacheManager` still needs to return the original iterator, meaning, after we serialize the values we need to deserialize it back. This may cause a performance regression for the `MEMORY_*_SER` storage levels because the deserialization process may be expensive. (This was attempted in #1083, but we later decided against it for this reason).
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---