Its great blog. Small doubt, In disk how spark stores the data is it in serialized form (byte format) or de-serialized form (object format). I felt this one is not clearly mentioned in the blog ? Because data spilling involves both serialization and de-serialization