zfs very poor performance compared to ufs due to lack of cache?

[Ivan Voras]

On 09/15/10 16:09, Andriy Gapon wrote:
> on 15/09/2010 16:42 Steven Hartland said the following:

>>> General problem of double-caching with ZFS still remains and will remain and
>>> nobody promised to fix that.
>>> I.e. with sendfile (or mmap) you will end up with two copies of data, one in
>>> page cache and the other in ARC.  That happens on Solaris too, no magic.

>> Obviously this is quite an issue as a 1GB source file will require 2GB of memory
>> to stream hence totally outweighing any benefit of the zero copy sendfile offers?

>> Indeed. Where would this need to be addressed as ufs doesn't suffer from this?
>
> In ZFS.  But I don't think that this is going to happen any time soon if at all.
> Authors of ZFS specifically chose to use a dedicated cache, which is ARC.
> Talk to them, or don't use ZFS, or get used to it.
> ARC has a price, but it supposedly has benefits too.

(replying for the OPs benefit)

This has been a question since the beginnings of ZFS on Solaris - the 
authors wanted their own control over the cache and hence the ARC was 
implemented (modified from the IBM's original). This decision has been 
contested as possibly ineffective but at the end it stayed.

Here are some Googled references:

http://www.mail-archive.com/zfs-discuss@opensolaris.org/msg08692.html
http://mail.opensolaris.org/pipermail/zfs-discuss/2009-July/029370.html

There are also some problems which are curiously similar to ones people 
complain about in FreeBSD+ZFS:

http://mail.opensolaris.org/pipermail/zfs-discuss/2009-July/029654.html
http://mail.opensolaris.org/pipermail/zfs-discuss/2009-July/029362.html

Random other references:

http://www.almaden.ibm.com/cs/people/dmodha/arcfast.pdf
http://nilesh-joshi.blogspot.com/2010/07/zfs-revisited.html
http://www.mail-archive.com/zfs-discuss@opensolaris.org/msg38362.html
http://www.thezonemanager.com/2009/03/filesystem-cache-optimization.html