Open Stack

> 
> Have you done anything on this since the initial draft you posted? 
No.
 
> My
> major design concerns with this are:
> 
> 1) How do you ensure that this new db service isn't a bottleneck?
> 
> 2) If the answer to #1 is to allow running more than a single instance
> of it, how do you ensure database updates are done in order?


It is certainly a best-practice to run more than a single instance of the daemon, I'd even go so far as to say it would be necessary (especially if an eventlet threadpool was not re-introduced).

Updates would be implemented through rpc 'call', which would provide synchronous writes from the perspective of the client making a series of related updates. Other cothreads and processes would already be lacking any guarantees of order amongst themselves, so I would see no difference there.

Regards, 
Eric Windisch