Ask a Pack Rat:
Reader Roger Pelizzari asks,
Enjoyed your detailed review of the Mac Mini. I got the late 2009 version.
What concerns me about the 320gb hard drive isn’t so much it’s slow speed as it’s reliability. I just read an article about 2.5 hard drives and the following paragraph gave me pause.
Any cheap 2.5” harddisk of 320GB or above is unreliable by definition. That’s just physics.
Data density for this small platter is so high (when over 125GB per platter), it’s reached the point after which the drive’s firmware relies heavily on ECC or other error correction methods. Your data is on a brink of becoming unreadable. Always use a business class drive if you need 320GB or higher in a laptop or 500GB in a desktop. If the hard disk is below 320GB, a mainstream drive is OK.
So what do you think. Is this 320GB drive in the Mini doomed?
I don’t think we can draw a line at 320 GB, but it’s certainly true that higher disk density increases the reliance on error correction and thus the risk of data loss. My friend Robin Harris talked about this just last month on ZDNet. Industry numbers suggest that one would have to fill a 320 GB SATA drive 37 times before encountering an unrecoverable read error, potentially losing data. This Microsoft test suggests that drives might be 10x more reliable than this, but that controllers and software above the drive are equally critical.
The Bit Error Rate (BER) for consumer SATA drives averages 1 in 1014. This means that for every 12.5 TB written, one unrecoverable read error (URE) will be encountered. Since it takes about an hour to write 1 GB to a consumer-grade drive, it would take on average 12,500 hours (520 days) to write enough data to encounter a URE.
Real-world applications don’t write anywhere near this much data, so the chance of a URE is much lower. As a rule of thumb, an average hard disk drive will be filled only once in its lifetime, leading to a shorthand calculation where 12.5 TB / (drive capacity) = chance of URE.
Therefore, I’d say that unrecoverable read errors aren’t yet at a crisis point for 320 GB drives. Whole-drive failure is much, much more likely. But drives are getting bigger all the time, and will soon reach the point where URE is inevitable. At current specs, filling a 1 TB drive has a 1 in 12 chance of URE, a 2 TB drive 1 in 6, and so on. RAID and other data integrity solutions will become an absolute necessity very soon, and even single-drive redundancy becomes problematic as drives near the 12 TB mark.
Parity RAID protects against URE by calculating a checksum for all data and recalculating the data if an error is detected. As long as the parity drive is running, two drives would have to hit a URE on the same data at the same time, an extremely remote occurrence.
But the risk of a URE in a RAID-5 set when one drive has failed is equal to the sum of the URE risk for the remaining drives, just like a non-protected set of drives that is one drive smaller. So a RAID-5 set consisting of 4 1.5 TB drives, once one drive has failed, is equal to the URE risk of a hypothetical 4.5 TB drive (3 x 1.5 TB), or one in 2.7. This is why dual-parity RAID-6 and other erasure code schemes are so valuable – they all but eliminate the chance of a URE for reasonably-sized drive and sets.
That’s why I’m a strong advocate of backups (thank you, Time Machine!) and RAID (thanks, Drobo!) Even today, with URE a fairly small concern, drive failures and data loss do happen. And these become more likely with motion (portable drives?) heat (planes and cars?) vibration (ibid) and other environmental conditions. Don’t panic about using 320 GB drives, but do back up your data anyway!
4K sectors are touted as a way to improve the ECC gain. Do you think 4k sector drives will pull us back from the brink, or will manufacturers use it as a way to reduce the space consumed by ECC?