Super interesting. A major point to be aware of from the docs:
Note what this is not: a distributed SQL engine. [...]
SQLite on RADOS is meant to be accessed by a single SQLite client
database connection at a given time.
So it distributes the backend storage across multiple nodes for the increased reliability, but any individual SQLite database is only accessible from a single node at a time.
That still sounds like a good trade off in a lot of cases.
Performance wise, the docs mention this when SQLite is configured for WAL journaling:
You can expect around 150-250 transactions per second ...
That's easily good enough for a lot of things. Obviously that would vary too depending on how the Ceph cluster itself is set up.
Wonder if this could be extended somehow, maybe with range locking?, to enable simultaneous operations from multiple nodes?
Thereby to create a cheap mans Oracle RAC equivalent... :)
That still sounds like a good trade off in a lot of cases.
Performance wise, the docs mention this when SQLite is configured for WAL journaling:
That's easily good enough for a lot of things. Obviously that would vary too depending on how the Ceph cluster itself is set up.Wonder if this could be extended somehow, maybe with range locking?, to enable simultaneous operations from multiple nodes? Thereby to create a cheap mans Oracle RAC equivalent... :)