How cluster data works
Hello. I just started using Couchbase and I have questions about its clustering technology, and how it differs from replication.
Is it okay if all members of the cluster are geographically separated, for the purpose of bringing the data closer to the end-user? I want a database to live entirely in several cities (so all nodes individually have at last 99% of the database), where little data is lost if a city goes dark, and it's easy to add/remove cities when necessary. I think Couchbase clustering seems to do the job, but is it okay if all members of the cluster are 50 to 200ms apart and could drop off once in a while when network conditions degrade? I know replication is coming out in 2.0, but from what I know of other databases' replication techniques, I fear it may be difficult to rebuild/re-add a server when one goes down versus clustering which appears to be simple.
>> Is it okay if all members of the cluster are geographically separated, for the purpose of bringing the data closer to the end-user?
If you do this you will want to use our cross data center replication feature that will be a part of 2.0 developer preview 5. Normal replication would be a bad idea.
>> I know replication is coming out in 2.0, but from what I know of other databases' replication techniques, I fear it may be difficult to rebuild/re-add a server when one goes down versus clustering which appears to be simple.
The cross data center replication feature should be very easy to use and I wouldn't recommend clustering a server with nodes that are far away from each other. We didn't design our normal replication for this use case and don't do any testing to help us understand the potential issues.