[lustre-discuss] Terrible BW via lustre network

[Tim Schnedler]

hum so you're using one raid and one vdev for all 60 drives? so all that is
just on I am guessing one OST? I would honestly destroy it and start over.
Read up on ZFS and recommended amount of disks per vdev, Then do a bit of
math figure out how many vdevs you need for 60 disk's then make more
then one OST for each vdev. You will get much better performance just doing
that part.

On Wed, Jun 10, 2026 at 4:30 AM Peter Grandi via lustre-discuss <
lustre-discuss at lists.lustre.org> wrote:

> > [...] 1 OSS (60 HDDs configured as a single dRAID3 pool, one vdev)
>
> That is *amazing*. :-(
>
> > 1 Client (using the same Lustre server image)
>
> 1 OSS, 1 client: I guess Lustre is not a parallel filesystem
> anymore :-)
>
> > Problem: Raw bandwidth per drive: ~250 MB/s
>
> A good example of "up to" metrics :-)
>
> > ZFS pool raw throughput (local): 5-5.5 GB/s
>
> A very good example of "up to" metrics :-)
> https://github.com/axboe/fio/tree/master/examples/
>
> > Lustre client mount throughput: <300 MB/s
>
> What is the "throughput" if you mount the filesystem on the
> server itself via 'localhost'?
>
> Note: with questions like this I am always wondering whether to
> not comment at all (my usual choice), comment euphemistically as
> in this case, or comment clearly.
> _______________________________________________
> lustre-discuss mailing list
> lustre-discuss at lists.lustre.org
> http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss_lists.lustre.org/attachments/20260616/6463c591/attachment.html>