Hello John!
The versi=F3n of IDS is the same in my both servers (12.10 FC3). The diffe=
rences are in the hardware. One of the servers is a 7 years old phisical ma=
chine (HP Proliant 585 G1) and the other is a virtual server over VMWare on=
modern hardware (HP Proliant 385 G7).
I suppose that update statistics is a heavy read process and maybe I could=
be in a mistake. In any case, the update statistics process is significati=
vely slow in the virtual server compared to the "old" phisycal one, and thi=
s is unexpected to me, because all the phoronix tests (CPU, Memory and HDD)=
are better in the virtual server.
Thank you very much for your interest.
Best regards.
--
Javier Perez Arenal - jperez@uniovi.es
Jefe del Area T=E9cnica de Inform=E1tica y Comunicaciones=20
Vicerrectorado de Campus, Inform=E1tica e Infraestructuras
Universidad de Oviedo=20
Edificio Severo Ochoa=20
C/ Fernando Bongera s/n, Campus del Cristo
33006 - Oviedo, Asturias
-----Mensaje original-----
De: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] En nombre de John Mi=
ller iii
Enviado el: martes, 27 de mayo de 2014 19:31
Para: ids@iiug.org
Asunto: RE: Poor performance on VMWare [33108]
You example provided of update statistics is probably not the best exam=3D =
ple.=20
The assumption made is that update statistics of an older Informix serv=3D =
er is the same as a new Informix server and that a majority of the time it =
is doing I/=3D O.=20
Update statistics
has been changed significantly over time and it can end up doing work t=3D =
o cleanup older style indexes which are inefficient.=20
John F. Miller III
STSM, Lead Architect
miller3@us.ibm.com
503-747-1366
IBM Informix Dynamic Server (IDS)=20
ids-bounces@iiug.org wrote on 05/26/2014 07:26:09 AM:=20
> From: "Alexandre Marini"<alexandre@briug.org>
> To: ids@iiug.org,
> Date: 05/26/2014 07:27 AM
> Subject: RE: Poor performance on VMWare [33100] Sent by:=20
> ids-bounces@iiug.org
>=20
> Hello, Javier.=20
> My personal opinion is that you might have a better performance, but =3D
you=20
> should not use the same onconfig file.=20
>=20
> 1) Are you really using cooked files? If so, please forget them. Chan=3D
ge
your=20
> chunks to raw devices, and make sure you are not using a bad RAID lev=3D
el
in=20
> your storage.=20
>=20
> 2) You could start a performance tunning process after topic 1 is sol=3D
ved.=20
> Informix 12 have several features that might benefit your=20
> performance, even in=20
> VMs environments. Eg: are you using cpu vp memory caches? Are you usi=3D=
=20
ng=20
RTO=20
> policies? Are you using alice index cleaner mode (OLTP system)?=20
>=20
> 3) You might also check other topics, but as a start phase, I=3DB4d go =
=3D=20
for=20
all of=20
> mentioned ones first.=20
>=20
> Hope it helps.=20
> Regards.=20
>=20
> Alexandre Marini=20
> IBM Informix Certified Professional v10 / v11.50 / v11.70 / v12.10=20
>=20
> IBM Information Management Informix Technical Professional=20
>=20
> IBM Infosphere DataStage Technical Professional=20
> Informix Senior DBA - Orizon Brasil=20
> BRIUG website administrator=20
> Informix independent consultant=20
>=20
> > To: ids@iiug.org=20
> > From: jperez@uniovi.es=20
> > Subject: Poor performance on VMWare [33099]=20
> > Date: Mon, 26 May 2014 09:44:07 -0400=20
> >=20
> > Hello=20
> >=20
> > I want to know some expert opinion about one IDS 12.10.FC3WE test=20
> that I ha=3D3D=20
> > ve currently executing in two machines. This machines are:=20
> >=20
> > SRV1: (+7 years old) HP Proliant DL585 G1=20
> > - 4xAMD Opteron 820 2.3GHz (2 cores x processor for a total of 8 co=3D=
=20
res)=20
> > - 16 GB RAM=20
> > - Local HDD on 146GB Ultra320 SCSI=20
> > - Informix storage: cooked files=20
> > - Red Hat 5.9 x64=20
> >=20
> > SRV2: VMWare Virtual server=20
> > - 8xvCPU (reported as AMD Opteron 6176 SE 2,3GHz)=20
> > - 16 GB RAM=20
> > - SAN (HP EVA6500) HDD on 2TB 7.2K EVA SAS MDL disk pool (Mapped RA=3D=
=20
W=20
LUN)=20
> > - SAN conectivity: 1Gbps iSCSI (server side) 10Gbps (SAN side)=20
> > - Informix storage: cooked files=20
> > - Red Hat 6.5 x64=20
> >=20
> > With these technical data, I am expected that the SRV2 results=20
> were better =3D3D=20
> > than the SRV2 ones (and more knowing the SRV1 "age").=20
> >=20
> > When I ran some scripts, stored procedures (SPL) and update=20
> statistics, I w=3D3D=20
> > as surprised by the results. The massive write operations were=20
> slightly bet=3D3D=20
> > ter on SRV2 (import was 40% faster in SRV2), but the scripts=20
> envolving heav=3D3D=20
> > y reading operations were about 1,5 times slow in the best cases,=20
> reaching =3D3D=20
> > 2,5 times slow in the worse cases.=20
> >=20
> > In this unexpected scenario, I looked for an IDS independent=20
> benchmarking t=3D3D=20
> > ool and I found a suite called "phoronix test suite" (PTS). I was=20
> decided t=3D3D=20
> > o find the bottleneck in SRV2... but again the results of the PTS=20
> surprised=3D3D=20
> > me. I ran various test, with the following results:=20
> >=20
> > - pts/SmallPT (CPU test. Less is better) > SRV1: 192 > SRV2: 141=20
(better)=20
> > - pts/Apache (CPU test. More is better) > SRV1: 7017.65 > SRV2:=20
> 9547.57 (be=3D3D=20
> > tter)=20
> > - pts/aio-stress (I/O test. Random Write. More is better) > SRV1:=20
> 22.06 > S=3D3D=20
> > RV2: 881.51 (better)=20
> > - pts/iozone (I/O test. Read Performance: 1MB record / 2GB file.=20
> More is be=3D3D=20
> > tter) > SRV1: 728.09 > SRV2: 2008.14 (better)=20
> > - pts/iozone (I/O test. Read Performance: 1MB record / 4GB file.=20
> More is be=3D3D=20
> > tter) > SRV1: 824.98 > SRV2: 1985.23 (better)=20
> > - pts/iozone (I/O test. Read Performance: 1MB record / 8GB file.=20
> More is be=3D3D=20
> > tter) > SRV1: 768.30 > SRV2: 1906.90(better)=20
> > - pts/iozone (I/O test. Read Performance: 4KB record / 2GB file.=20
> More is be=3D3D=20
> > tter) > SRV1: 1016.75 > SRV2: 2113.92 (better)=20
> > - pts/iozone (I/O test. Read Performance: 4KB record / 4GB file.=20
> More is be=3D3D=20
> > tter) > SRV1: 1071.86 > SRV2: 1953.93 (better)=20
> > - pts/iozone (I/O test. Read Performance: 4KB record / 8GB file.=20
> More is be=3D3D=20
> > tter) > SRV1: 1086.07 > SRV2: 1794.42 (better)=20
> > - pts/iozone (I/O test. Write Performance: 1MB record / 2GB file.=20
> More is b=3D3D=20
> > etter) > SRV1: 25.38 > SRV2: 97.64 (better)=20
> > - pts/iozone (I/O test. Write Performance: 1MB record / 4GB file.=20
> More is b=3D3D=20
> > etter) > SRV1: 26.28 > SRV2: 114.04 (better)=20
> > - pts/iozone (I/O test. Write Performance: 1MB record / 8GB file.=20
> More is b=3D3D=20
> > etter) > SRV1: 28.36 > SRV2: 144.49 (better)=20
> > - pts/iozone (I/O test. Write Performance: 4KB record / 2GB file.=20
> More is b=3D3D=20
> > etter) > SRV1: 26.17 > SRV2: 91.99 (better)=20
> > - pts/iozone (I/O test. Write Performance: 4KB record / 4GB file.=20
> More is b=3D3D=20
> > etter) > SRV1: 26.35 > SRV2: 108.19 (better)=20
> > - pts/iozone (I/O test. Write Performance: 4KB record / 8GB file.=20
> More is b=3D3D=20
> > etter) > SRV1: 28.62 > SRV2: 144.37 (better)=20
> > - pts/ramspeed (RAM test. Add integer. More is better) > SRV1:=20
> 3533.75 > SR=3D3D=20
> > V2: 6782.31 (better)=20
> > - pts/ramspeed (RAM test. Copy integer. More is better) > SRV1:=20
> 3418.95 > S=3D3D=20
> > RV2: 6782.31 (better)=20
> > - pts/ramspeed (RAM test. Scale integer. More is better) > SRV1:=20
> 3448.33 > =3D3D=20
> > SRV2: 6776.20 (better)=20
> > - pts/ramspeed (RAM test. Triad integer. More is better) > SRV1:=20
> 3499.36 > =3D3D=20
> > SRV2: 6536.23 (better)=20
> > - pts/ramspeed (RAM test. Average integer. More is better) >=20
SRV1:3453.82 =3D3D=20
> > > SRV2: 6851.13 (better)=20
> > - pts/ramspeed (RAM test. Add floating point. More is better) >=20
> SRV1: 3955.=3D3D=20
> > 16 > SRV2: 7241.65 (better)=20
> > - pts/ramspeed (RAM test. Copy floating point. More is better) >=20
> SRV1: 3599=3D3D=20
> > ..66 > SRV2: 7178.45 (better)=20
> > - pts/ramspeed (RAM test. Scale floating point. More is better) >=20
> SRV1: 339=3D3D=20
> > 9.08 > SRV2: 7286.11 (better)=20
> > - pts/ramspeed (RAM test. Triad floating point. More is better) >=20
> SRV1: 385=3D3D=20
> > 5.18 > SRV2: 7461.69 (better)=20
> > - pts/ramspeed (RAM test. Average floating point. More is better)=20
> > SRV1: 3=3D3D=20
> > 647.64 > SRV2: 7351.95 (better)=20
> >=20
> > In these test, SRV2 outperforms SRV1 in a clear and consistent wayi=3D=
=20
n=20
any o=3D3D=20
> > f the evaluated fields (Processor, I/O and Memory). But the IDS=20
> 12.10.FC3WE=3D3D=20
> > still runs slowly in SRV2 when I execute massive read operations (b=3D=
=20
y=20
examp=3D3D=20
> > le, update statistics over my test db is 136% more slow than the=20
> same opera=3D3D=20
> > tion on SRV1). The IDS config is exactly the same in the two=20
> servers (oncon=3D3D=20
> > fig copied from one to other).=20
> >=20
> > One thing I observed is that the SRV2's virtual processors shows=20
> more activ=3D3D=20
> > ity than the SRV1's ones when I execute any IDS operation. Also ist=3D=
=20
he=20
same=3D3D=20
> > thing with the "top" utility in the operating system command line.T=3D=
=20
his=20
is=3D3D=20
> > a "contradiction?" with the SRV2's better PTS CPU tests .=20
> >=20
> > I do not identify or suspect any possible reason who explain the=20
> poor SRV2 =3D3D=20
> > performance... any ideas?=20
> >=20
> > Thanks in advance and excuse me if the english is not quite good...=3D=
=20
:-(=20
> >=20
> > --=20
> > Javier Perez Arenal - jperez@uniovi.es<mailto:jperez@uniovi.es>=20
> > Jefe del Area T=3D3DE9cnica de Inform=3D3DE1tica y Comunicaciones=20
> > Vicerrectorado de Campus, Inform=3D3DE1tica e Infraestructuras=20
> > Universidad de Oviedo=20
> > Edificio Severo Ochoa=20
> > C/ Fernando Bongera s/n, Campus del Cristo=20
> > 33006 - Oviedo, Asturias=20
> >=20
> > --Boundary_(ID_B1ApZu83LpmDOByskPtisQ)=20
> >=20
> >=20
> >=20
>=20
***********************************************************************=3D=
=20
********=20
> > =3D=
=20
> >=20
>=20
>=20
>=20
***********************************************************************=3D=
=20
********=20
> =3D=20
>=3D=20
***************************************************************************=
****=20
=20
*******************************************************************************
To post a response via email (IIUG members only):
1. Address it to ids@iiug.org
2. Include the bracketed message number in the subject line: [33115]
*******************************************************************************
↧