On Sun, Sep 09, 2018 at 03:00:11PM +0200, Manuel Bouyer wrote:
> > I'm doing some benchmarks now (just for vnd, not xen yet). I can see
> > a penalty of about a factor of 2 for linear I/O, much less for random I/O.
>
> On my setup it's between 5 and 10. enough to make the disks 100% busy
> with less than 1MB/s of real I/O (the disk itself does about 3MB/s read
> and 5MB/s writes, but it seems to realy disklike doing a read followed
> by a write to the same sector(s)).
The handle_with_rdwr code does lots of weird things. It tries to
kill any kind of read-ahead or caching and flushes writes synchronously.
The result is that you read a file block multiple times. Each time
digesting a single 'sector' and throwing away the rest.
But there must be something else as the speed doesn't change when the
data is kept cached and no physical I/O is done.
--
Michael van Elst
Internet: [email protected]
"A potential Snark may lurk in every tree."
