I read the Tweaktown piece and while some of their points may be slightly correct, they haven't really reported anything that everyone doesn't already know because BackBlaze freely shares all of their info.
By Backblaze's own admission some of the sample pools are too small. They also have made it perfectly clear that different drives are in different pods. If Storage Pod v1 was so awful then why was it so awful to only Seagate drives. They have 1TB WD drives that have a higher average age (almost 6 years now) than the "bad" Seagates. The WD Greens have failure rates in the single digits while the Seagates are in double and even triple digits. Pretty sure they were all in v1 pods.
Backblaze is also pretty clear that they understand that these drives weren't meant for this kind of workload. I for one, though, love the idea of huge sample sizes in harsh environments to help reveal which ones are the strongest and the weakest. Backblaze undoubtedly reveals this. It's not like they pamper the HGST drives, fanning them with palm leaves and feeding them grapes while using the Seagates as door stops. No way would expect to see the same failure rate as they see.
If we look at just 3TB drives it's a lot like the old 1TB comparisons. The HGST drives are actually older than the notorious Seagate drives, increasing the likelihood that they were stuck in v1 or v2 pods. Yet, their failure rate is closer to 0% than the Seagates are to 25%.
Only a Seagate apologist could deny that the HGST drives are plainly more durable. Does the BackBlaze data mean that the Seagates will fail more often in the home use environment? No, and they never claimed they would. But if my data is important then it might be worth the extra $15, just in case. Personally, I have migrated to Toshiba drives since they are only about $5 more but seem more durable (even though the sample size is quite small).
I am not a brand loyalist. I started my my server four years ago with 3x2TB Seagates. I have since added a couple of WD Greens (1TB and 2TB) and 4 Toshibas (1x2TB and 2x3TB). None of them have failed, yet. Get another year under their belts and those 6TB Seagates are looking pretty good if they don't start dropping out.
I'm not advocating either way. The Backblaze data is enough to move me away from the Seagate drives, given that the alternatives aren't any more expensive. I actually just bought a couple 3TB Toshiba drives as well to use in my home server, given the lack of any real price difference.
I also like the idea of a huge sample size, but as you say the added failure in their specific use case does not necessarily correlate to failure numbers in consumer drives. A drive from manufacturer X could have a longer life in a client workload, better performance, etc than Brand Y, but be more susceptible to continuous high levels of vibration. If you use them in an extremely high vibration environment X would fail at a significantly higher rate relative to Y than you'd see in a consumer environment.
Similarly, neodymium magnets are graded by strength and (effectively) a temperature rating. Magnets with the highest strengths aren't available in the higher temperature ranges. Manufacturer Y might have went with a smaller, higher strength magnet with a lower temperature rating than manufacturer X in a client drive, with a standard use in mind for such a drive. If those drives are used in atypical applications that expose them to higher than designed for temperatures that could cause early failures on Y's drives, again without a corresponding link to the failure rate you'd see if they were used in consumer drives.
Not saying either of those happened or are even likely in this case, they're just off the cuff examples of trying to repurpose something for a different environment than it was designed dfor. The Tweaktown piece is useful simply because it provides a check for the people that would look the Backblaze data and say "Wow, those Seagates 3TB drives suck. They're 100x worse than Hitachi." Those drives likely still do have issues but unless a postmortem is done to see why they failed, it can't be concluded that that issue would be a concern tucked under someone's desk in a minitower PC. I'm guessing the majority just looked at the chart and moved on from there without delving too much into the background information.
It is funny to see the IBM drives do so well though, given that I still get a little squeamish and won't but a Deskstar even if it's the cheapest alternative. Old habits and all.
Facebook
Twitter