Saturday, April 14, 2007

Finally affordable RAID based NAS

In this rambling I will review the Buffalo Terastation Pro II Nas 1TB (4x250GB-SATA)

I have been looking for a product like for this for nearly an year. The concept of a low-end RAID based Network Attached Storage (NAS) is into its second generation and quite matured now, and it became quite affordable as well.

This product was very simple to install. It took me less than 10 minutes to get it out of the package, install the software, change the RAID setting from 5 to 10, and to start copying the files. It came pre-configured to use the subnet address 192.168.1.150, but since I didn't use that address, I was fine. Performance looked ok on my standard 10/100 Ethernet connection - I have yet to install Gigabit routers in my network.

Anyhow, I did quite a bit of reading on the various levels of RAID while researching the product. Below is a quick summary of what I'd learned. This summary assumes that all the disks in the array are of the same size.

RAID 0 - Only striping, no mirroring
READ PERFORMANCE Very good as reads are distributed across multiple disks
WRITE PERFORMANCE Very good as writes are distributed across multiple disks.
RECOVERY None
CAPACITY 100% of the total space is available.
SUMMARY Good for high-performance systems with no need for recovery.

RAID 1 - Only mirroring, no striping
READ PERFORMANCE Standard - same as a single non-RAID disk. Could be better some times if special hardware based controllers take advantage of multiple disks in the mirror (more than two).
WRITE PERFORMANCE - Some performance hit is there as everything you write has to be written to two or more disks. So, Hardware level controller is a must.
RECOVERY If one of the two disks fails, then the new disks gets built from its mirrored counterpart.
CAPACITY 50% of the total space, as the other 50% is used for mirroring.
SUMMARY Good for non-performance-critical systems with strong need for recovery.

RAID 5 - Distributed parity. Typical configurations involve 4 or more disks.
READ PERFORMANCE Pretty good, almost as good as RAID 0.
WRITE PERFORMANCE - OK to poor as parity information needs to be calculated and updated on each write.
RECOVERY Allows at most one disk failure at a time. Recovery could be painfully slow, as when one of the disks fails, the new disk's data is calculated using the parity info stored across all the other disks.
CAPACITY You lose 1 disk's worth of storage to store the parity info. In other words, a 4x250GB configuration will get you 750GB worth of storage space.
SUMMARY Not really good for anything - unless you are the NAS vendor trying to sell to customers who don't know any better - they can be easily misled by the higher RAID number, and the higher usable capacity (n-1 * disk capacity).

RAID 10 - Striping over mirrored disks. A 4 disk configuration involves two striped sets with each set containining two mirrored disks.
READ PERFORMANCE Very good. As good as RAID 0.
WRITE PERFORMANCE - Very good. Almost as good as RAID 0.
RECOVERY All but one disk can fail in each RAID 1 (mirrored) set.
CAPACITY 50% of the total storage space - the other 50% is lost for mirroring.
SUMMARY Good for high-performance systems with solid need for recovery. Unless you are have super-mission-critical system, in which case you should be looking for more advanced RAID configurations, this solution offers the most affordable yet high-quality solution for small businesses/home offices, IMO.

My ideal setup? I would use a two fast hard disks in RAID 0 configuration on my workstation for performance reasons, and a RAID 10 based NAS storage for backup/network storage purposes. If I want more peace of mind, I can do daily/weekly back-ups from the NAS storage to a USB based external hard disk.

You can read more about RAID levels in detail at Wikipedia:
http://en.wikipedia.org/wiki/Standard_RAID_levels
http://en.wikipedia.org/wiki/Nested_RAID_levels

Here's an excellent article that settled the RAID 5 vs RAID 10 debate for me:
http://www.miracleas.com/BAARF/RAID5_versus_RAID10.txt

0 Comments:

Post a Comment