I thought we were all pretty much on the same page that the Backblaze numbers were pretty much useless.
The only people who should use Backblaze as a source of HD reliability information are folks that re going to intentionally install consumer hard drives in a an inadequately designed server environment in direct confict with manufacturers' written instructions and ignoring the fact that they are not designed for such usage.
You might as well complain about wearing a watch that was not designed to be waterproof when going diving to 200 feet of depth nd then complaining it failed.
When Backblaze published their data, some dude (Christopher Nelson) sued Seagate cause he had a drive fail and the warranty alsoi failed. He sues on the bais of Backplaze data alleging that the data proved the drive was faulty. It didn't turn pout so well:
1. Backblaze installed consumer drives in a high-volume enterprise-class environment that far exceeded the warranty conditions of the HDDs.
2. Backblaze used its own internally developed chassis, using rubber bands to hold the vertically mounted HDDs in place.
3. It is not as if testing consumer drives in a server envirnment is like a tougher test and is therefore in any way representative of perforemnce in a consumer environment. It does the opposite. Putting a server drive in a consumer environment makes it fail faster; putting consumer drive in a server environment makes them fail faster.
4. Server drives are designed based upon the assumption that they will be installed in data centers with thick concrete floors and racks securely bolded to said floors providing a vibration free environment. Server drives therefore do not include a head parking feature which parks the arm when I/O slacks to prevent the arm crashing into the disk when vibration occurs. BB's rubberband equipped cases left sitting on folding table didn't exactly qualify for this designation.
5. Most consumer drives are equipped with features suited to the consumer environment. One of these is the head parking feature which is intended to protect the drive from the copy paper delivery guy's handruck at the office or your dog jumping up when doorbell rings while he's sleeping under your desk. So here, the drive is parked when I/O activity lapses.
6. Consumer drives don't see the I/O activity that server drives do so they spend most of their time in the parked position. These drives are generally rated for 250-500k parking cycles, far more than they are likely do see in a small / medium office or home usage.
7. Server drives on the other hand, can see 30-90k cycles in a month. So the very feature that protects consumer drives in a consumer environment, causes pre-mature failure in a server environment.
8. BB used more drives per chassis than allowed for under manufacturer's design specifications leading not only far greater loadings but also environment temperatures.
9. BB took this route as low budget consumer drives made financial sense. Not only were they much cheaper so that it was cheaper to replace a drive 3-4 times than purchase one designed for that environment, but server Hds were in short supply at the time because the flooding in Thailand has wiped out many plants.
10. So what it comes down to is a) of what possible relevance is backblaze data when they are purposely installing drives in direct conflict with manufacturermwritten instructions, inmtentionnaly ignoring the fact that they are not designed for this service and b) why in the world would anyone look at this data when relevant data is readily available. At the time that the BB data was used as the basis of this lawsuit, other data ..., actually applicable data ... was readily available and published by behardware. The following data presents actual consumer drives bought by consumers thru retail channel(s) and RMA'd by brand and model number. The 1st number is the latest 6 month reporting date and the 2nd the 6 month reportimg period preceding:
- Seagate 0,72% (0,69%)
- Toshiba 0,80% (1,15%)
- Western 1,04% (1,03%)
- HGST 1,13% 0,60%)
To my eyes, based upon a historical perspective, anything < 22% is great. back in 2013 it was
Toshiba 1,15%
Seagate 1,44% (1,65%)
Western 1,55% (1,44%)
Samsung 2,24% (1,30%)
Hitachi 2,40% (3,45%)
