vendredi 23 janvier 2015

Replication Strategy for Multi-Office Database


I'm considering options for setting up a repository of large spatial datasets that can be used by employees across our company. We have offices scattered around the US. Our employees utilize desktop-based GIS software to do a variety of things so a web-based solution isn't feasible at this time.


If all data is stored in one physical location on the network, the data is large enough to cause significant lag, which is not acceptable for employees who need the data for conducting analysis.


If data is stored in each office location with no replication across servers, we will have many duplicate datasets with potentially many versions of data, which isn't really an improvement over current operations.


What I'm hoping to set up is a network of Master-Master databases where each office has their own server (solves the lag problem) but all servers are synced.


I expect additions/edits to the data to be relatively infrequent (perhaps an average of less than 1,000 row changes per day, with spurts of activity when a new dataset is added or a large analysis is performed). Asynchronous replication would be OK. The majority of usage would be read-only.


We use PostgreSQL. I have looked into Bucardo but I'm not sure it's the right solution for us and the documentation left a good deal to be desired.


My questions are:


1) Is there some standard strategy for dealing with this problem that I am missing? Most of the information I've found on replication deals with massive data centers and failover solutions, which isn't our use case.


2) What currently-maintained tools should I investigate to achieve what we need? Ideally, it would be something baked into PostgreSQL itself, but I get the sense that functionality is currently lacking in the core.





Aucun commentaire:

Enregistrer un commentaire