Cached file read performance with 6.2-PRERELEASE

[Bill Moran]

In response to Mark Kirkwood <markir at paradise.net.nz>:
> JoaoBR wrote:
> 
> > I am not convinced that this kind of test is of any value for comparing 
> > systems at all because there are too much factors involved - unless the 
> > competitors are installed on identical hardware. On the other side I think it 
> > is usefull to compare tweaked settings on a particular machine. For example 
> > you may change fsize/bsize of the filesystem or any other and can compare 
> > this results then.
> > 
> 
> Exactly, that's why I did the comparison - I think you missed the part 
> where I mentioned the 2 systems were *identical* with respect to cpus, 
> memory, mobo - in fact even the power supplies are identical too! the 
> only differences are that the Gentoo system has a different RAID 
> controller from the FreeBSD one (a cheaper one in fact), and the FreeBSD 
> system has larger capacity disks (slightly newer variants, same brand!) 
> given this comparison is about *cached* read rates, the RAID controllers 
> and disks are not significant I think.
> 
> Specifically:
> 
> Gentoo :
> - Supermicro P3TDER
> - 2xSL5QL 1.26 GHz PIII
> - 2xKingston PC133 RCC Registered 1GB DIMMS
> - Promise TX4000 4x Maxtor plus 8 ATA-133 7200 40G
> 
> FreeBSD
> - Supermicro P3TDER
> - 2xSL5QL 1.26 GHz PIII
> - 2xKingston PC133 RCC Registered 1GB DIMM
> - 3Ware 7506 4x Maxtor Plus 9 ATA-133 7200 80G
> 
> In fact, to indulge your skepticism ('cause I think this is a real issue 
>   worth sorting out), I booted the FreeBSD system with a Gentoo livecd 
> and  ran the same tests there... and guess what - identical results to 
> the installed Gentoo system...so... errm - *my* experimental method is 
> sound...so how about we just get together and see how to make FreeBSD 
> kick Gentoo eh?

I looks like your testing methodology is sound, and that you've
uncovered an issue worth pursuing.

I recommend starting this thread up on freebsd-performance at .  The folks
on that list are more likely to jump all over this kind of thing.

You might also find helpful people on the current@ and hackers@ lists.
My gut tells me that any changes that can improve this will be large
enough that they'll have to go through CURRENT first, then get MFCed
back in to 6.

Keep in mind also that the holidays tend to slow things down, it might
be early January before you get a lot of people looking at this issue
seriously.

-- 
Bill Moran
Collaborative Fusion Inc.