Closed parkan closed 7 years ago
SQLite should be sufficient for the immediate future. I don't think rocks.db is a good choice here (except perhaps for raw statement storage); we need an index and query capabilities.
In terms of scaling, SQLite can takes us far enough I think, and it should be easy to swap in a more heavyweight db solution if/when the need arises.
What protocol level metadata are we thinking about? Signer, timestamp, what else?
Down with SQLite, but should keep the scaling to billions of rows/batch processing/potentially columnar store case in the back of our minds.
As a baseline I think we need indexing by namespace, id, publisher, timestamp, tags, and WKI references.
Well, the plan is to have it simple to use for lightweight usage. For aggregators/big nodes we will probably want to use a database server or RDS.
Local store for statements and index
Probably SQLite but maybe something like rocksdb
Do we need to consider long-term/large dataset storage optimizations at all yet?