cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Peter Schuller <peter.schul...@infidyne.com>
Subject Re: Multi-DC Deployment
Date Thu, 21 Apr 2011 11:48:40 GMT
> Again, for a lot of services, it is fully acceptable, and a lot better, to
> return an almost complete (or maybe even complete, but no verified by
> quorum) result than no result at all.

+1, except maybe "a lot" depending on how one chooses to define that.
There are definitely cases where sufficient information propagating
end-to-end between clients, services and storage systems can be very
conducive towards providing a better end-user experience; particularly
in terms of more graceful degradation when you do have failures that
exceed that which was planned for.

Introducing fallback-to-lower-consistency has significant issues; it's
difficult to get right in a way that is actually useful in degraded
conditions while also not at the same time contributing to your
service committing suicide by magnifying already existing problems by
adding to the load in an attempt to degrade gracefully.

So... just wanted to voice that I think you're raising valid use cases here.

-- 
/ Peter Schuller

Mime
View raw message