macOS' APFS File-System Performing Much Better Than The Dated HFS+

[aht0]

You can make ext4 slow-down to crawl, try mounting it with sync option. You'd pretty much emulate how HFS+ behaves. Because with fsynch() in linux, there is no guarantee that commit in a file system is a really commit in normal conditions. And since you brought up databases, it's a spot where that particular way of things could really bite your ass in Linux.
You do have that guarantee in OSX. Reason it is so sluggish. Trade-off between protecting data and performance.

[aht0]

Yep and most of them don't have any clue that the hardware specs place them at about $200-$300 dollar range if they were PCs. Most mac users overpayed by $800 dollars or more. That's why Apples 3% marketshare gets them huge profits. IMO its a damn shame.

You are ignoring the money Apple has to dish out per manufactured item to various patent holders. For example: Samsung gets 100+ USD for each iPhone manufactured and sold. Though, Apple still makes hefty profit.

[name99]

I distinctly remember reading that the file system introduced with the Apple III was called "HFS" and introduced to the Macintosh range in 1985 with the first Macintosh-compatible HDD from Apple. Even if the Mac HFS and Apple HFS, both of which originally came out in the first half of the 1980s, are separate their naming is close enough to make your insults unfounded. Other than that I also never claimed that there hadn't been any change and specifically talked about the evolution of the HFS file system.



If you've looked at any of the storage benchmarks on here then you'll know that HFS+ had a tendency to show a consistently sub-par performance year in and year out. If failing to perform year in and year out isn't a failure then you seem to have a pretty high bar for what a failure is for a file system.

I'm sorry! You are clearly ignorant about every dimension of this issue, but somehow I'm the one whose at fault?
There are THREE file systems of relevance here. The first, from 1980, was designed for SOS. The second, from 1985, was designed for the Mac.
It "borrows" from the 1980 file system in the same way ext4 "borrows" from UFS, that is it uses some of the same ideas, and that's pretty much it.
But even that is somewhat irrelevant because the 1985 HFS was replaced by HFS+, which is the name used by anyone who knows what they are talking about, and HFS was a dramatic rewrite of HFS. Again using some of the same ideas, if you want to insist on that (by which I mean concepts like hierarchical name space, use of B-Trees to organize the catalog and extents, or provision for multiple forks) but insisting that these define a file system is like claiming that pretty much every UNIX file system that ever existed is identical because they all support a hierarchical name space, hard links, and inodes.

And even the HFS+ of 1998 is not the end of the story because in 2002 Apple added journaling and this was considered a significant enough change that the file system is referred to technically (ie on-disk, when you create partitions, etc) by a different name, now JHFS+. And, as I said, along with this last naming change, there's been a constant run of improvements of all sorts.

To continue to assert that the JHFS+ of today is basically the same thing as the HFS of 1980 is as absurd as claiming that ext4 is basically the same thing as V7's FS.

As for "you'll know that HFS+ had a tendency to show a consistently sub-par performance year in and year out", well, let's see.
Look at https://www.phoronix.com/scan.php?pa...a-ubuntu&num=2
and consider the disk intensive benchmarks. What I see is pretty much a wash.
Mac wins SQLite,
Linux wins CompileBench,
Mac wins some LLVM compiles, Linux wins some other LLVM compiles.
Mac wins Postgres
with a similar sort of spread for both SSD and HD results.
I'd call that pretty much parity, not "consistently sub-par performance".

[L_A_G]

I'm sorry! You are clearly ignorant about every dimension of this issue, but somehow I'm the one whose at fault?

I admit to being mistaken and point towards confusing naming so I'm somehow blaming you for this? Do you identify yourself as "Apple Computers Inc" or something? Because that's the only way that sentence makes any kind of sense.

...

Again with that straw man? I know when you think you've got a really good thing going it can be hard to admit you were wrong and move on, but I never said that it's the same. Instead I talking about current day HFS a further development of something that came out in the 1980s and I was not wrong about it!

As for "you'll know that HFS+ had a tendency to show a consistently sub-par performance year in and year out", well, let's see.

I don't know about you, but I see OSX winning only by a slight margin where it does win, but losing very badly when it loses. Hell, you could even say that I was being nice to it with those downright abysmal results.

Seriously, I don't know what the hell your problem is, but you keep acting like I'm insulting your mother or something.

[aht0]

You are splitting hairs, guys. Unless you are using said file system (JHFS+) and dissatisfied with it, why even bother? To do so only because of being Linux/Apple apologist or fanatic /take your pick/, is down-right stupid. Raw performance is just one variable file system could have and not even the most important for many. I'd have liked to see for example, what the fuck would have happened to either of the file systems if Michael had included tests of "pull power from the box 10 times after 10min uptime each".

I am quite happy to use ext4 with sync option myself and literally don't care about it's impact to the performance - especially on SSD's, as long as files remain intact on hard drive itself. Not that I even have much reason to be afraid of power loss. Just a matter of principle. What's the fucking importance of 0.1sec time difference between some file operations.

[nils_]

As usual this was never the case, the actual situation was VERY different. But you don't want to hear about that, do you?

For those who ARE interested in reading about the issue:

https://postgrespro.com/list/thread-id/1658794

which explains that the REAL issue is ATA drives that don't follow the spec, and how Apple tried to deal with it

and

https://www.postgresql.org/message-id/20050222053741.GL86914%40decibel.org

who explains how ALL OSs have to suffer from this (because the damn bugs are in the drives themselves!)

That information is severely outdated, and since Apple is usually in complete control of their hardware also unlikely. There used to be a lot of hard drives that were broken (acknowledging writes that went to cache). Of course these days most consumer SSD suffer from the same problem.