Not too long ago I wrote a small piece for a forum describing view on grid storage. I liked so now I am posting it here as well. Enjoy!
Capacity and performance
Grids can scale to incredibly large capacities as each storage node normally will add more resources to both capacity and computational power. Other framed storage systems relies on a couple of controllers to do all RAID and addressing computation for all disks in all arrays. This limits both capacity and performance.
Data migration
With very large amounts of data you don’t really want to migrate data to a new storage system every time your kit runs out of service and warranty. With the grids you leave the data and instead deploy new nodes or retire old nodes individually as time and wear requires. No more data migrations…
Backup
Vast amounts of data are hard to backup. Policies on data availability in a grid eliminate the need for back-up. For example with point-in-time data replication and a minimum of four copies of each piece of information in the grid you don’t really need to back up your data.
Distribution of information
Grids are by nature made to be scattered all over the place. Therefore global data availability can be achieved relatively cheap and easy and even over WAN lines by putting storage nodes/grid nodes where you need the data.
Discovery
Finding data in a grid is nothing new. Internet search engines are doing that all the time. Storage CAS/FCS will surely help finding your data in the grid (if you know what you are looking for).
For compatibility I know one grid (Caringo) that uses HTTP for communicating with the users. That to me seems to be a protocol that everybody speaks and will speak for some time in the future too.
I think storage grids are great in this age of clouds and globalization but for single applications in our data centres traditional F.C. SAN and even iSCSI deployments are still number one. But for large amounts of files and other fixed data types the grid is really the way to go.
