Marc Herbert wrote:
Whatever the WaitForCompletion policy is, if you have two connections
writing the same table, then on each backend the second write will
have to wait for the first write to complete. Is that right? (or: is
that write?)
Yes, this is serializability and Sequoia provides
1-copy-serializability. I don't think that synchronous/asynchronous is a
term that apply to that situation.
So this seems faster than waiting for remote backends to answer, but
you still have pay the cost of total order broadcast across
controllers compared to "really" asynchronous replications like slony.
Yes, you are absolutely right, at the conceptual level, Slony uses
master-slave (or asynchronous) replication whereas Sequoia uses
multi-master (or active/active or synchronous) replication. I probably
misunderstood your previous email and I thought that you were addressing
[a]synchrony inside Sequoia.
In fact Slony does not care about the slave state and you are likely to
lose data on master failure which is one of the major drawback of
master/slave replication.
Thanks for the clarification,
Emmanuel
--
Emmanuel Cecchet
Chief Architect, Continuent
Blog: http://emanux.blogspot.com/
Open source: http://www.continuent.org
Corporate: http://www.continuent.com
Skype: emmanuel_cecchet
Cell: +33 687 342 685
_______________________________________________
Sequoia mailing list
Sequoia@lists.forge.continuent.org
https://forge.continuent.org/mailman/listinfo/sequoia
