Thursday, April 7, 2011

RAID Fun and Games - Part One

Recently I decided that I wanted to put togther a RAID array at home so I could do some database benchmarking. I wll not bore you with the details. I like to get my hands dirty on things like this from time to time as I feel it enhances my role as a solutions architect.


I did not want to spend a fortune on doing this. I wanted a RAID enclousre that supported at least four drives. Also, my plans were to use an older laptop to interface to the enclosure so I needed the enclosure to have an onboard RAID controller. After looking at reviews on the newegg website I decided that the "Mediasonic HFR2-S3B PRORAID Box 4 Bay Raid Enclosure" was going to be my choice. It has an onboard RAID controller and it supports Firewire, USB 2.0, and eSata. USB will work well with the connecting laptop.


I was a bit puzzled by the reviews for this and other products. They were all over the map, and prompted me to do further research before I do the deal. Turns out there are two classifications of eSata drives:
  1. Desktop Drives
  2. Enterprise Drives
Enterprise drives have TLER, CCTL, ERC (depending on the hard drive manufacterer. Three acronyms for the same thing.). These are, according to Wikipedia, "firmware bugfixes to reduce the amount of time a drive will attempt to recover from an error." Most RAID arrays assume that after 7 seconds that the drive is no longer available to the RAID array and RAID will drop it from the array. The goal of these three techniques is to prevent drives from prematurely falling out of the RAID array.

This is good as rebuilding a disk in a RAID array can take a great deal of time (sometimes measured in days). Given that the general goal of a RAID configuration is performance, availability, and protection of your data, having a good drive fall out of an array would seem to defeat these three goals.


It used to be that one could enable TLER on desktop drives. In fact, for some time Western Digital offered a utility to do this. It no longer does. Additionally, recovery times of todays desktop hard drives can be up to two minutes. This and the fact that you can no longer enable TLER on them make them unsuitable for use in a RAID array.


RAID stands for Redundant Array of Inexpensive Disks. The key word for many of us is inexpensive. Let me put this in perspective for you. A 1TB Western Digital Blue Caviar hard drive is $65.00. This is a desktop disk. A 500mb Western Digital WD5002ABYS costs about $82. This is an enterprise drive. I am using Amazon for my source of pricing - researched on 4/6/2011. So, 1/2 the disk space costs about 25% more.


Apparently enterprise drives are hardier than desktop drives. I am not a hardware guy so I cannot speak to this.


In my research I was able to determine that we used to be able to enable TLER on desktop drives, but no longer can. I was not able to find a reasonable explanation as to why we can no longer enable TLER (or the other listed standards). In other words, those of us that wish to use desktop drives in a RAID array no longer can (at least not safely).


This is not my biggest issue on this matter. In researching this subject the lack of cohesive information as to what drives can be used in a RAID array is hard to come by. Ridiculoulsly and painfully hard to come by.


Also, all of the "all over the map" reviews and vendors forum discussions now make sense to me. Many claim great success using desktop drives in RAID arrays, while others write about their nightmares in using desktop drives in RAID arrays. The people claiming grand success have not yet encountered an error that takes their drive more than 7 seconds to recover from. Those experiencing problems have.


Vendors should be responsible and ...
  1. Provide a list of drives suitable for their RAID enclosures and/or RAID controllers.
  2. Acknowledge and provide better information on this subject
I have spent close to 8 hours researching this subject. I almost dropped $250 on desktop drives that given my luck would have worked like a champ in my RAID array until a day after they went off warranty.