ZFS Comes to OS X Courtesy of Apple's Former Chief ZFS Architect

[fmaxwell]

Yeah bro, I know how it works. you might wanna consider a custom OSX Installer disc if you're gonna rely heavily on ZFS, and won't have internet access.

Apparently you don't know how it works.

If you install a new OS X and your system won't boot, you don't have Internet access.

Many OS X upgrades are one-way: They modify files to make them compatible with new versions of the OS and the applications. For example, the email database was redone with Snow Leopard. Same thing could happen with iTunes, iPhoto, or any other app. So you can't just reinstall the old OS and keep working.

----------

This is why you resist the fanboy urge to upgrade right away and check into compatibility beforehand.

This is why you don't put your OS on an unsupported filesystem. What if it NEVER becomes compatible and you've got 4TB of data on it?

Why don't you control your ZFS fanboy urges and wait until Apple releases ZFS or something equivalent?

 

[marcusj0015]

Apparently you don't know how it works.

If you install a new OS X and your system won't boot, you don't have Internet access.

Many OS X upgrades are one-way: They modify files to make them compatible with new versions of the OS and the applications. For example, the email database was redone with Snow Leopard. Same thing could happen with iTunes, iPhoto, or any other app. So you can't just reinstall the old OS and keep working.

----------



This is why you don't put your OS on an unsupported filesystem. What if it NEVER becomes compatible and you've got 4TB of data on it?

Why don't you control your ZFS fanboy urges and wait until Apple releases ZFS or something equivalent?

... Did I not suggest you to add the ZFS Kext to the boot image over a month ago?

 

[odedia]

Interesting thing I read was actually quite the opposite. Can't seem to find it now, but from watching behaviors, seems quite likely.

Because ZFS does a heavy amount of caching in memory when writing, it writes in relatively big sequential swaths. For example, if you're copying a large number of small files to a ZFS volume, a bunch of these will go into RAM first and then get written in chunks around 512K at a time.

Why does this matter? Well, remember the old crappy JMicron 601/602 SSDs and the stutter issues? Writing in large sequential blocks (relative to the NAND block size) cuts down on the stutter dramatically and increases the performance.

I read this after getting a 16GB one for the boot drive... maybe should have moved it to being the ZIL. But oh well, too lazy.

SSD's are great for being ZFS second-level cache, but I was refering to actually building SSD RAID-Z systems. Oracle built something like this which was the fastest drive in the world or something... can't find the link right now.