I need to create storage server (file server) to store backups and user files from Windows machines. I'm thinking to take CentOS 6.3 as a new server OS and setup samba sharing. All of the sudden I came across GlusterFS, which seems to be more advanced and scalable...
So the question is should I consider using GlusterFS or should I go the easier way and just use samba?
The correct answer depends on how much you plan to scale.
GlusterFs is good if you plan to have multiple data nodes and requirement such as HA, no single point of failure and almost linear scalability.
An option in case you don't plan to have multiple nodes could be ZFS. it basically replaces xfs+lvm, and gives you out of the box volume management and aggregation, self healing capabilities, data deduplication, data compression, snapshotting and also smb, iscsi and nfs sharing.
If you desire you can also combine them together having gluster bricks over zfs disks.
Windows does not (yet) have a native Gluster client. So you either have to connect it via samba or NFS to windows. In other words, another server in between gluster and windows which mounts the gluster volume and re-exports it using samba or nfs.
You may want to also look into using Ceph. But again, no native windows client and you'll have to re-export as well. From memory Gluster has inbuilt NFS server (although it doesn't provide the clustering/failover features of gluster itself).
In your particular situation, I wouldn't be looking at Gluster or Ceph right now. These filesystems are of more use in a large organisation or data centre where you expect to grow your storage needs very rapidly. They are designed to run on multiple servers as well. If you use it to serve to windows you'd end up having to run a samba server anyway and connect your windows clients to that.
Simply get a server with a suitable number of storage bays and RAID the disks. In our situation we're actually using a standard PC but with a case that can take multiple drive bays (not hotswap). We're using linux software raid in raid 5 mode giving us around 8TB of storage. It's working really well as a simple NAS for backups and basic filesharing needs. It's not a critical server so the lack of hot swappable hardware isn't an issue at all.
Gluster allows clients to connect using Samba. In a single server environment you probably won't notice too much of a difference between regular Samba and Gluster. If you decide you'd like the share to scale out and be distributed/ high-availability, then it may be a good idea.
Using Gluster will probably be more complicated to begin with, but may be worth the investment if your backup server needs to scale in size.
Have a read of the few pages located here. It should help you decide whether it's right for you
I don't have much experience with glusterfs in production but in such scenarios I deployed a few times community edition of Openfiler NAS appliance which is based on rPath Linux.
But this it not so important. More interesting is the set of features it has. It supports file-based networking protocols NFS, SMB/CIFS, HTTP/WebDAV or FTP, block-based storage area network protocol iSCSI (initiator and target), filesystem quotas, integration with Active Directory and more. It leverages Linux LVM for volume management and all of these features are easily configurable from unified web-based management so it' not so necessary to really understand how it works under the hood.
