ZFSisms that are not true, or no longer true

In a tool like this which is designed to compare RAID levels, the value of p isn’t very important. The relative differance between the RAID levels is.

Assuming the bathrub curve here can be done by the user and results adjusted if you are curious, but this would be an inaccurate representation. Not all drives fail exactly the same. Take a look at the data from Backblaze.

I wouldn’t really take the data from Backblaze as the gospel. They use mostly consumer drives and their data is relatively young (at least from the last data I saw published).

I’m not sure why that negates my point. Some hard drives have differant failure mode patterns than others. Perhaps I should have said some hard drive skus have differant failure mode patterns than others

I kinda agree. But still, the use case for this tool is to compare the reliability aka “how big are the chances of me loosing the entire pool”.

And for that, the p value and even how you split vendors in vdevs is important.

Of course not, but similar. Take a look at the google study. It said that temps are surprisingly irrelevant but age is. And after some age, they die “pretty” fast.
It is not like one (from the same batch) dies after 6y and the other after 8y. They mostly fail in year 6y (just a random example).

This is exactly the data where my “bad batch problem” concern stems from :slight_smile:
and that is excluding all the SSDs with broken firmware and sync write bugs.

Can you take the drive failure discussion elsewhere?

Unless you have both items to add for the original post, AND real reasonable consensus, this discussion distract from the original purpose of this Resource. (Aka Resource threads are not general purpose discussion threads.)

2 Likes

@Arwen This thread should have been placed elsewhere.

Resources is more of a location for ‘finished books’. The opening discussions and drafts should have taken place in a different category and then you would publish the book and minor edits in Resources.

2 Likes

@SmallBarky - Perhaps. The old forum had on-topic discussions following the first post.

Part of the problem is that people went off and had basically off-topic discussions.