[patch 0/5] [PATCH CFT] more revlog experiments

[Chris Mason]

On Tuesday 21 March 2006 12:45, Zach Brown wrote:
> > Assuming the operating system reads and caches inodes and
> > directory data at least a block at a time, we shouldn't see a huge
> > seek win here.
>
> How about processing blktrace logs of the different loads to get some
> stats on their seekiness?

I'm really short on time this week, but might have some spare cycles over the 
weekend.  I did a comparison on the number of compressed bytes read from 
disk, and the packed stuff was not significantly less than an unpacked repo.  
This should mean most of the speed differences come from seeks.

But, blktrace or kprobe/systemtap tracing are the only way to know for sure.

-chris