A storage server should have a hard drive for the operating system and an array of drives for the shared storage. We feel that the most important feature for a storage hard drive is reliability. We went with IDE drives because of their superior price to performance ratio, as compared to SCSI. In our case, we don’t even need the bandwidth of the SCSI drives – quantity rather than blistering speed was important. With respect to SATA or parallel ATA, both are more then adequate for our needs.
With these needs in mind, we chose the Maxtor Maxline II Plus 250GB 7200rpm 8MB buffer hard drives. These drives are rated at 1,200,000 hours MTBF as compared to 600,000 hours for standard consumer drives. This does not mean that you can run your hard drive for 137 years, but does imply that it is more reliable than a standard desktop drive. Maxtor has advertised this drive as one designed for 24/7 applications, this is in stark contrast to the old line of IBM drives that did not recommend continuous usage. Currently 250GB is the maximum capacity for 7200rpm drives. The only other IDE/SATA drive with a similar MTBF rating is the Western Digital Raptor series, but the max capacity is still only 36GB, with a 74GB version coming soon.
We will use four of these drives for a nice and even one terabyte of storage, with a server design that will allow for an easy addition of another 4 drives for a peak of 2 terabytes. But, when we are ready for a storage upgrade there will likely be even higher capacity hard drives on the market, further extending our maximum storage capacity.
The Maxtor hard drives assembled
To RAID or not to RAID
The next decision we had was whether to RAID the drives or not. Since we were interested in reliability, RAID 1 or RAID mirror was considered. If you believe the numbers, running a drive in RAID mirror will double the effective MTBF, we have done that by choosing the Maxline series vs a standard consumer IDE hard drive. In addition, our budget constraints did limit our ability to implement a RAID 1 array.
Another possibility was RAID 5, which allows 5 drives to act as 4 drives. An additional parity track is written on each drive, so if one fails, then the other drives can recover the lost data. This is available through software or hardware. This is a great solution if you do not plan to upgrade your maximum server capacity. When the time comes to replace a drive with a higher capacity drive, you will be forced to replace the entire array. As this is a budget server, going to RAID 5 isn’t as important to us. One must realize that RAID designs really only protect against hard drive failures. With power supply or motherboard failures, more than one drive can be destroyed at a time. So it should be important to note that simply turning on the RAID option does not guarantee the safety of your data.
Our solution is to run the drives in a standard configuration. We plan to have 4 drives initially with room for another 4 drives. Since these drives are not in any disk arrays, we can remove drives as they fill up and place them into storage for backup. Better yet, our plan is to place full hard disks into external drive boxes. New technology in the form of higher density drives can easily be added to the server and the old drives can be removed and placed into USB 2.0 or Firewire enclosures, ensuring easy access to old data. This is also the reason Linux was not a good choice for our system -- it doesn't make sense to put XFS/ext3/ReiserFS drives into a USB2.0/Firewire external box. Since we anticipate going through 2 TB of data every year, this setup allows for that flexibility without a significant cost penalty. Remember, you do not need to populate a storage server to maximum configuration to start. In six months when we’ve filled up a terabyte, we’ll be able to buy another terabyte at a lower cost. With this setup, we are relying upon the reliability of the drives to last a year and will also attempt to maximize the reliability of the system as well. This was another reason why we chose standard ATA drives over SATA drives. DVD-Rs will serve as temporary backup along the way.
For the system drive, we went with a Seagate Barracuda 7200.7 SATA 120GB drive. We like the Seagate SATA series because they do not use an internal SATA to parallel ATA bridge. While this is more a theoretical rather than practical issue, with even the Raptor having a bridge, this is a nice design touch. In addition, the Barracuda is very aggressively priced. One consideration is that the Barracuda does not have a standard Molex power adapter and must be connected with a Serial ATA power adaptor. This makes it a less than ideal upgrade drive for older systems.
Seagate Barracuda SATA drive
Cost: 4 x $250 for the Maxline II Plus, $110 for the Barracuda