[Gluster-users] Replicated vs distributed
countz at gmail.com
Sat Apr 17 20:14:16 UTC 2010
Another thing to note is that when you distribute, say to 3 servers, every server will hold a third of the files.
This means that when you want to read the files, the 1st server will have to grab some of the files from the 3rd server, over the network.
If you have an application which requires very high performance, you will have to use Replicate.
With Replicate, all files are on all servers, and you can instruct the client to always prefer the local server when reading the files.
This will give you excellent performance, with the benefits of the clustered (replicated) filesystem.
More information about the Gluster-users