[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: partman-zfs warning template wording

On 2011-02-08, Robert Millan <rmh@debian.org> wrote:
> This is a common missconception.  ZFS has stability and performance
> problems on Intel 386 due to limitations of this architecture, it's not
> because of word length. I386 carries on with 30 years of ill design
> decisions and legacy baggage. Other 32-bit architectures such as
> ARM or MIPS don't suffer from this kind of issues.
> Specifically, the problems have to do with small number of
> general-purpose registers and with inability to implement
> privilege separation at the MMU level.

General-purpose registers, yeah, sure.  This might create performance
problems, but i386 is as capable to get the processing done as amd64.

Could you elaborate on the MMU bit?  I'm not aware of huge changes that
cannot be emulated by PAE, too.

I don't defend i386, but your overly brief formulation of issues look
like FUD to me.  Your references (in the blog post) also didn't back
that up.  You might run into addressing trouble with virtual memory
management, due to the size of the caches ZFS wants.  But then state
this instead of refering to unrelated issues at hand.

Philipp Kern

Reply to: