Q1: the Journal is conceptually similar to tape. It's really good (performance and cost) for sequential reads and writes, but poor for random access. This is why we use it to record transactions. Indexed storage is designed for random access, but this is simply more expensive to implement. We bill for them separately to keep the cost of QLDB as low as we can make it.
Q2: the Journal size will not (ever) increase unless you are committing transactions. You may be noticing this in CloudWatch Metrics, which may be delayed by a few minutes (or longer, depending on how you are aggregating the metrics).
Q3: that's it, just your data. QLDB also offers some virtual tables (such as information_schema.user_tables
), but these are not "in indexed storage" - i.e. you're not paying for them.