File System: RAID Arrays
From Debian Clusters
RAID
RAID stands for a Redundant Array of Inexpensive/Independent Disks. Basically, it allows multiple hard drives to act as one hard drive in order to have data redundancy and better performance. It comes in many different flavors, from RAID 0 and 1 up through RAID 5 and 6. RAID is by no means necessary for a cluster, but putting your users' data (probably the NFS mount) on a RAID "array" (a group of hard drives connected together through RAID) builds in a backup.
For more information, I recommend the RAID Tutorial from the University of Massachusetts.
RAID vs. NFS
Didn't we just cover file systems in the NFS tutorial? Well, yes and no. NFS is what allows a file system to be shared amongst multiple nodes. It "looks" to the system and to the users like copies of the same files are situated on each one of the worker nodes, but in reality, it only exists in one location: the hard drive of the head node. All of the other nodes "mount" this file system and access it over the network.
RAID is one step closer to the hardware than NFS. In NFS, the worker nodes don't really care what kind of file system they're getting from the head node. The head node is responsible for dealing with all of the particulars of it. RAID is on the hardware and software level of the file system of the head node.
Types of RAID
There are many different types of RAID, software RAID, and in between the two. Unfortunately there isn't time to cover all of them. My university happened to purchase HighPoint RocketRAID cards. I'm not endorsing the brand, but since that's the model I have, that is what I will be documenting here.
