Re: HyperThreading
--002354435cf01ffdb6049c9350ea
Content-Type: text/plain; charset=UTF-8
On Fri, Feb 18, 2011 at 1:00 PM, Saman Barghi <saman.b@gmail.com> wrote:
> It would be great if you had the SMP results as well. Anyway, I think you
> would love this:
> http://www.cs.uwaterloo.ca/~sbarghi/bench/transactions1.jpg
> (keep in mind that this is a quad core xenon with 2 GB of RAM, and FreeBSD
> is in SMP mode)
>
Could you try -master? There has been fairly substantial work since the 2.8
release; all of the tokens in 2.8 are still taking the MP lock as a
precaution. That is no longer the case in master. The process token has been
broken up fairly dramatically in master as well, so I expect to see
different results.
Can you give us a bit more information about the h/w and configuration? In
Kris Kennaway's tests, he was seeing FreeBSD 7 sustain 8 times the number of
transactions you're reporting for 8.1, and that was for hardware that is
three or four years old at this point. Also, last I tried sysbench/OLTP with
MySQL on dfly, I saw MySQL making many 256K allocations, which fall back to
mmap() in our system. mmap() acquires the VM token, so I expected to see a
much more dramatic falloff than you saw.
I guess I wasn't clear earlier, the results on my graph were SMP; the UP
lines were not present; the results were from very early in the 2.8 cycle,
but I don't have the exact rev handy.
> As I said before; I kept the environment the the same as what has been
> mentioned here: http://people.freebsd.org/~kris/scaling/dfly.html . I even
> used the same version of MySQL (5.0) although MySQL 5.5 shows better
> scaling behaviour.(
> http://mikaelronstrom.blogspot.com/2010/04/mysql-554-m3-scales-to-32-cores.html).
> IMHO if "The multiprocessor work that has been ongoing in DragonFly is
> really starting to bear fruit", the fruits are indeed juicy and sweet. It
> would be lovely to see the results on a machine with more cores though.
>
> Btw, is there any changelog that shows what has been changed about tokens
> !?
>
>
Yes --
http://gitweb.dragonflybsd.org/dragonfly.git/commit/2a4189307741dbcfbe11b31d6cc51a4fb39a8cde
was the commit that reworked the token contention path.This was not in 2.8,
but is in master. Matt has seen improvements from in on a 48-core system he
runs; I don't know that we've seen great things from it on smaller systems
http://gitweb.dragonflybsd.org/dragonfly.git/commit/3b998fa96afe52828957ea4f65d15320eb0fe240
was the other one in particular I was referencing; that one produced a
fairly noticeable bump in apache performance, among other things...
-- vs
--002354435cf01ffdb6049c9350ea
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
<div class=3D"gmail_quote">On Fri, Feb 18, 2011 at 1:00 PM, Saman Barghi <s=
pan dir=3D"ltr"><<a href=3D"mailto:saman.b@gmail.com">saman.b@gmail.com<=
/a>></span> wrote:<br><blockquote class=3D"gmail_quote" style=3D"margin:=
0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left=
: 1ex;">
<div dir=3D"ltr">It would be great if you had the SMP results as well. Anyw=
ay, I think you would love this:<br>=C2=A0<a href=3D"http://www.cs.uwaterlo=
o.ca/%7Esbarghi/bench/transactions1.jpg" target=3D"_blank">http://www.cs.uw=
aterloo.ca/~sbarghi/bench/transactions1.jpg</a> <br>
(keep in mind that this is a quad core xenon with 2 GB of RAM, and FreeBSD =
is in SMP mode)<br></div></blockquote><div><br>Could you try -master? There=
has been fairly substantial work since the 2.8 release; all of the tokens =
in 2.8 are still taking the MP lock as a precaution. That is no longer the =
case in master. The process token has been broken up fairly dramatically in=
master as well, so I expect to see different results.<br>
<br>Can you give us a bit more information about the h/w and configuration?=
In Kris Kennaway's tests, he was seeing FreeBSD 7 sustain 8 times the =
number of transactions you're reporting for 8.1, and that was for hardw=
are that is three or four years old at this point. Also, last I tried sysbe=
nch/OLTP with MySQL on dfly, I saw MySQL making many 256K allocations, whic=
h fall back to mmap() in our system. mmap() acquires the VM token, so I exp=
ected to see a much more dramatic falloff than you saw.<br>
<br>I guess I wasn't clear earlier, the results on my graph were SMP; t=
he UP lines were not present; the results were from very early in the 2.8 c=
ycle, but I don't have the exact rev handy.<br>=C2=A0</div><blockquote =
class=3D"gmail_quote" style=3D"margin: 0pt 0pt 0pt 0.8ex; border-left: 1px =
solid rgb(204, 204, 204); padding-left: 1ex;">
<div dir=3D"ltr">As I said before; I kept the environment the the same as w=
hat has been mentioned here: <a href=3D"http://people.freebsd.org/%7Ekris/s=
caling/dfly.html" target=3D"_blank">http://people.freebsd.org/~kris/scaling=
/dfly.html</a> . I even used the same version of MySQL (5.0) although MySQL=
=C2=A0 5.5 shows better scaling behaviour.(<a href=3D"http://mikaelronstrom=
..blogspot.com/2010/04/mysql-554-m3-scales-to-32-cores.html" target=3D"_blan=
k">http://mikaelronstrom.blogspot.com/2010/04/mysql-554-m3-scales-to-32-cor=
es.html</a>). IMHO if "The multiprocessor work that has been ongoing i=
n DragonFly is really starting to bear fruit", the fruits are indeed j=
uicy and sweet.=C2=A0 It would be lovely to see the results on a machine wi=
th more cores though. <br>
<br>Btw, is there any changelog that shows what has been changed about toke=
ns !? <br><br></div></blockquote><div>=C2=A0</div><div>Yes -- <br><br><a hr=
ef=3D"http://gitweb.dragonflybsd.org/dragonfly.git/commit/2a4189307741dbcfb=
e11b31d6cc51a4fb39a8cde">http://gitweb.dragonflybsd.org/dragonfly.git/commi=
t/2a4189307741dbcfbe11b31d6cc51a4fb39a8cde</a><br>
was the commit that reworked the token contention path.This was not in 2.8,=
but is in master. Matt has seen improvements from in on a 48-core system h=
e runs; I don't know that we've seen great things from it on smalle=
r systems <br>
<br><a href=3D"http://gitweb.dragonflybsd.org/dragonfly.git/commit/3b998fa9=
6afe52828957ea4f65d15320eb0fe240">http://gitweb.dragonflybsd.org/dragonfly.=
git/commit/3b998fa96afe52828957ea4f65d15320eb0fe240</a><br>was the other on=
e in particular I was referencing; that one produced a fairly noticeable bu=
mp in apache performance, among other things...<br>
<br>-- vs<br></div></div>
--002354435cf01ffdb6049c9350ea--
討論串 (同標題文章)
完整討論串 (本文為第 11 之 16 篇)：