Re: sshd responsiveness

[Justin Sherrill <justin@xxxxxxxxxxxxxxxxxx>]

--047d7b414e8cc2b24304d5634d7e
Content-Type: text/plain; charset=ISO-8859-1

Well, it's two different file systems, two different operating systems, and
(I assume) two different machines with different disks and different
network links.  Maybe even two different encryption setups?  I don't know
if the FreeBSD machine is using ZFS or UFS - if it's UFS, it's not
recording history so there's less overall work.  That can make a difference
too. Unfortunately this all adds up to a shoulder shrug, in terms of an
answer.

There's been a lot of discussion about the disk scheduler and how to
position reading vs. writing for Hammer over the past year or two.  If you
really want to dive into it, there's the 'dsched' man page, and the other
pages it links to from there.  Also, there's scheduler-related links on the
Digest:

http://www.shiningsilence.com/dbsdlog/index.php?s=sched



On Sun, Feb 10, 2013 at 5:17 AM, Carsten Mattner
<carstenmattner@gmail.com>wrote:

> When I compared FreeBSD 9.1 geli based full disk encrypted system
> with a dfly (HAMMER) 3.2.2 full disk encrypted system I noticed that
> using an ssh session is considerably less responsive on dfly if I
> rsync data from another machine to the dfly machine in parallel.
> On FreeBSD the ssh session doesn't get unresponsive. This is
> subjective and not scientifically measured and I'm just curious
> to hear your thoughts. Is it just probably caused by default settings?
>

--047d7b414e8cc2b24304d5634d7e
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Well, it&#39;s two different file systems, two different o=
perating systems, and (I assume) two different machines with different disk=
s and different network links. =A0Maybe even two different encryption setup=
s? =A0I don&#39;t know if the FreeBSD machine is using ZFS or UFS - if it&#=
39;s UFS, it&#39;s not recording history so there&#39;s less overall work. =
=A0That can make a difference too. Unfortunately this all adds up to a shou=
lder shrug, in terms of an answer.<div style>
<br></div><div style>There&#39;s been a lot of discussion about the disk sc=
heduler and how to position reading vs. writing for Hammer over the past ye=
ar or two. =A0If you really want to dive into it, there&#39;s the &#39;dsch=
ed&#39; man page, and the other pages it links to from there. =A0Also, ther=
e&#39;s scheduler-related links on the Digest:=A0</div>
<div style><br></div><div style><a href=3D"http://www.shiningsilence.com/db=
sdlog/index.php?s=3Dsched">http://www.shiningsilence.com/dbsdlog/index.php?=
s=3Dsched</a><br></div><div style><br></div></div><div class=3D"gmail_extra=
"><br>
<br><div class=3D"gmail_quote">On Sun, Feb 10, 2013 at 5:17 AM, Carsten Mat=
tner <span dir=3D"ltr">&lt;<a href=3D"mailto:carstenmattner@gmail.com"; targ=
et=3D"_blank">carstenmattner@gmail.com</a>&gt;</span> wrote:<br><blockquote=
 class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1px #ccc soli=
d;padding-left:1ex">
When I compared FreeBSD 9.1 geli based full disk encrypted system<br>
with a dfly (HAMMER) 3.2.2 full disk encrypted system I noticed that<br>
using an ssh session is considerably less responsive on dfly if I<br>
rsync data from another machine to the dfly machine in parallel.<br>
On FreeBSD the ssh session doesn&#39;t get unresponsive. This is<br>
subjective and not scientifically measured and I&#39;m just curious<br>
to hear your thoughts. Is it just probably caused by default settings?<br>
</blockquote></div><br></div>

--047d7b414e8cc2b24304d5634d7e--