Date: Thu, 6 Apr 2006 12:21:45 +0200 From: "Eric" <ericd@free.fr> To: "Mark Kirkwood" <markir@paradise.net.nz> Cc: Lucas Holt <luke@foolishgames.com>, freebsd-smp@freebsd.org Subject: RESOLVED : Low perf with smp on DELL server Message-ID: <002201c65963$e86e3990$65fd24c0@Eric> References: <009301c65418$cc029f30$65fd24c0@Eric><64659004-E643-4044-AAED-076E0CC91977@foolishgames.com><01a901c65493$77ad2740$65fd24c0@Eric><442DC9B6.2090200@paradise.net.nz><442EAA0A.8030704@he.iki.fi> <442F0E1B.9090505@paradise.net.nz><027f01c657f4$0d0a66a0$65fd24c0@Eric> <4432858F.90701@he.iki.fi><008001c65825$f3c27700$65fd24c0@Eric><44331024.7090807@paradise.net.nz><000a01c6588e$c3201da0$65fd24c0@Eric> <44339F80.3040704@paradise.net.nz>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi Mark,
I do a lot new test, re-install all system and arrive now to this :
Test without SMP : 29 sec
Test with SMP : 37 sec
Wich is very far from the 62 sec we have before with SMP and seems to be
"normal" for most people saying SMP and 4.11 are not so friend they seems to
be. Ok for you?
It seems that problem before, came from perc4 controller configuration. It
seems that if you change configuration on fly you don't have the same result
;=( as if you re-create the logical drive with good param and put same
param onfly .
Good configuration is for me disable any cache :
On creation and on configuration : Write Policy --> Writethru
Read Policy -->
non-adaptive
Don't know why difference was more important with SMP but it seems to be
now ok.
As there is still few difference with SMP, I now do same test but launching
1, 2, or 3 test in same time to see if it's usefulle for me to have SMP
kernel. Result is really near but I doubt that it's not hardrive that slow
down all system.
Do you know a way to know the hard drive utilisation as with top? (iostat
always return 0)
Do you know a way to do high ramdrive? (I try with mount_mfs but it always
limit me to 400 Mo)
Many thanks,
Eric.
----- Original Message -----
From: "Mark Kirkwood" <markir@paradise.net.nz>
To: "Eric" <ericd@free.fr>
Cc: "Lucas Holt" <luke@foolishgames.com>; <freebsd-smp@freebsd.org>
Sent: Wednesday, April 05, 2006 12:44 PM
Subject: Re: Low perf with smp
> Eric wrote:
>> Main problem is that query is done in a database of more than 2Go and it
>> will be difficult to mail it. ;=)
>> I try with several versions of mysql and php from mysql 4.1.18 to last
>> 5.x and php 4.4.0 to 5.x and don't find significant change in
>> performance.
>> If it can help here is the loop.
>>
>> for($debut=$deb;$debut < $fin;$debut += $range){
>> $end = $debut + $range;
>> $query = "select avg(total) from `histo_perfs_x` where id1 =
>> '".$id1."' and id2 = '".addslashes($id2)."' and date > '".$debut."' and
>> date < '".$end."'";
>> $result = mysql_query($query) or die("La requete :$query a echouee.");
>> if(!$result)$databar_total[$j][$i]= 0;
>> else {
>> $ligne = mysql_fetch_array($result);
>> if($ligne[0] != "")$databar_total[$j][$i]=$ligne[0]; else
>> $databar_total[$j][$i]= "-";
>> }
>> }
>>
>
> Excellent! Well yes, 2G of data is too big to mail! However, generating 2G
> of synthetic test data is no problem at all.
>
> However to do this, a little more information is needed:
>
> 1/ A create statement for the table histo_perfs_x, along with its indexes,
> explicitly specifying if using an engine other than myisam.
>
> 2/ An estimate of proportion of the histo_perfs_x table a query like:
>
> select avg(total) from histo_perfs_x
> where id1 = ? and id2 = ? and date > ? and date < ?;
>
> is likely to scan (for typical values of the 4 bind variables).
>
> 3/ (If possible) the output of explain of the above query so we can
> duplicate the access plan.
>
> 4/ The number of iterations you were using in the php loop for your test.
>
> Cheers
>
> Mark
>
>
> _______________________________________________
> freebsd-smp@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-smp
> To unsubscribe, send any mail to "freebsd-smp-unsubscribe@freebsd.org"
>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?002201c65963$e86e3990$65fd24c0>