I notice we use
std::fs for file i/o, in the state db particularly (although it may be used elsewhere, I’m not sure as of right now). I know
tokio offers an
async file i/o api as a feature flag, so I’m wondering if we could speed vault startup by using the
async file i/o of
tokio and allowing a vault’s I/O scheduler to batch up the file reads/writes better and allowing execution to continue elsewhere.
Is there a reason we currently don’t use the async versions, or is it more that the async conversion isn’t quite complete yet (as part of the “minor issues” mentioned in the September 3 update)? Wouldn’t be a massive effort to convert afaik, but just curious as I stumbled across it earlier.