directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alex Karasulu <>
Subject Re: [replication] Master slave replication will not suffice
Date Thu, 24 Feb 2005 18:14:08 GMT
Robb Penoyer wrote:

>Adding some food for thoughts:
>- take care to consider uniqueness conflicts. Server A 
>receives an add operation, and server B also receives an add 
>operation at around the same time. When replication occurs, 
>would not those entries have the same PK in the backend?
>- you might want to allow conflict resolution to be costly 
>in order to increase performance of conflict detection. 
>Perhaps consideration of a "tolerant" replication, meaning, 
>tolerate and capture ALL changes in an uncommitted region, 
>and allow detection to occur in thread, and resolution to 
>occur "outside".
>- some protocols may not be suited for MMR, some may be 
>tolerant. Is there any reason why the base server couldn't 
>support MSR and MMR on a per service configuration?

Good points Rob. Actually MSR is ok if the service sitting on top of the 
directory backing store is Read only. Then it does not ever need to go 
outside of the JVM to update data in MSR.


View raw message