From: Oxiel Enrique Contreras Vargas (oxielc@yahoo.it)
Date: Tue Jul 05 2005 - 09:29:37 EDT
Hello Gurus:
Real thanks to all who answered :)
The final solution to my problem is:
Without touching any kernel parameters.
I split the backplane, 2 scsi bus for the disks, while one has the O.S. and
other filesystems, the other bus serves the informix data (raw partitions).
This improved the speed, and the dbimport last only 18 hours from the initial
30+.
The second thing was to add a second temporal dbspace in a different disk from
the first.
Then modify the values of the onconfig file, in order to use Parallel Database
Query, this improves the creation of indexes and execution of querys and
sorts. This improved the dbimport, finishing it in 8 hours.
Now this time is quite long, but's definitely better than the first one,
Informix people are helping me with other settings but they've not made
further difference.
Best regards to all of you
Oxiel
El Mar 07 Jun 2005 05:50, Leonard, Nick escribió:
> Good Morning Oxiel,
>
> I have had a quick look an the onconfig file:
>
> Normally tuning is based on knowing loads on the system etc how many users
> etc what the server/database is used for
>
> But on a quick view
>
> You only have one Nettype configured and only 100 threads
>
> You should look at using both vps's for this nettype or add a second one
> that your config supports
>
> E.g.
> NETTYPE ipcshm,3,330,CPU # Configure poll thread(s) for nettype
> NETTYPE soctcp,1,100,NET # Configure poll thread(s) for nettype
>
> I would try
> NETTYPE soctcp,2,100,CPU
>
> That will help performance wise and also give you 200 threads. Otherwise if
> you only have 25 users say try soctcp,2,30,CPU
>
> I would also try switching the Value NOAGE to 1 this will help performance
>
> Locks and Buffers seem ok - The buffers are using half your ram if you need
> to buffer information fine otherwise maybe reduce to 100,000 or 50,000
>
> I would also switch off the processor affinity
>
> E.g.
> NOAGE 1 # Process aging
> AFF_SPROC 0 # Affinity start processor
> AFF_NPROCS 0 # Affinity number of processors
>
> This is worth trying and see if it helps
>
> You have no NUMAIOVPS even if you are using KIO you should configure one
> for each physical disk set or add vps until one of the VPS does no work.
>
> Finally the SHMVRTSIZE is using the other half of the RAM you have ????
>
> My current values on a system running 300+ users with 6gb ram and 3x833
> cpu's
>
> SHMVIRTSIZE 196608 # initial virtual shared memory segment
> size SHMADD 8192 # Size of new shared memory segments
> (Kbytes)
> SHMTOTAL 0 # Total shared memory (Kbytes).
> 0=>unlimited
>
> This does seem very large and may explain performance if Informix is
> grabbing all the ram before anyone logs in - is the machine swapping ?
>
> Cut down these values and it will run better otherwise add another 2/3 Gb
> of ram.
>
> Any more thoughts please contact me
>
> Regards
>
> Nick Leonard
>
>
>
>
>
>
>
>
>
> -----Original Message-----
> From: Oxiel Enrique Contreras Vargas [mailto:oxielc@yahoo.it]
> Sent: 06 June 2005 17:11
> To: Leonard, Nick
> Subject: Re: Tru64v5.1B slow and Informix
>
>
> Thank you very much Nick!!.
>
> This is the environment, disk setup, all raw files and attached is the
> onstat
> -d result and onconfig.
>
> INFORMIXSTACKSIZE=128
> INFORMIXSERVER=fatima
> INFORMIXDIR=/usr/users/informix
> INFORMIXSQLHOSTS=/usr/users/informix/etc/sqlhosts
> ONCONFIG=onconfig
> DBCENTURY=C
> DBDATE=DMY4/
> LD_LIBRARY_PATH=/usr/users/informix/lib:/usr/users/informix/lib/esql:/usr/u
>s ers/informix/lib/tools:/usr/shlib:/usr/dt/lib:
>
> lrwxrwxrwx 1 root informix 15 may 25 14:41 datos01
> -> /dev/disk/dsk1d
> lrwxrwxrwx 1 root informix 15 may 25 14:41 datos02
> -> /dev/disk/dsk2d
> lrwxrwxrwx 1 root informix 15 may 25 14:41 datos03
> -> /dev/disk/dsk3a
> lrwxrwxrwx 1 root informix 15 may 25 14:41 datos04
> -> /dev/disk/dsk3b
> lrwxrwxrwx 1 root informix 15 may 25 14:41 datos05
> -> /dev/disk/dsk4a
> lrwxrwxrwx 1 root informix 15 may 25 14:41 er_cdrspace
> -> /dev/disk/dsk1e
> lrwxrwxrwx 1 root informix 15 may 25 14:41 er_dbspace
> -> /dev/disk/dsk1f
> lrwxrwxrwx 1 root informix 15 may 25 14:41 er_sbspace
> -> /dev/disk/dsk1g
> lrwxrwxrwx 1 root informix 15 may 25 14:41 logdbs
> -> /dev/disk/dsk2a
> lrwxrwxrwx 1 root informix 15 may 25 14:41 rootdbs
> -> /dev/disk/dsk1a
> lrwxrwxrwx 1 root informix 15 may 25 14:41 sfifir
> -> /dev/disk/dsk1b
> lrwxrwxrwx 1 root informix 15 may 25 14:41 tmpdbs
> -> /dev/disk/dsk4d
> lrwxrwxrwx 1 root informix 15 may 25 14:41 tmpdbs2
> -> /dev/disk/dsk5d
>
>
> Best regards.
>
> Oxiel
>
> El Lun 06 Jun 2005 11:10, escribió:
> > It will be the Informix tuning as well unless they have retuned it for
> > in /usr/informix/etc is an onconfig file for the live instance this
> > will need to be tuned KIO is faster than AIO, buffers, virtual
> > processors etc - also was the database in raw storage as this is
> > 10-20% faster than cooked storage.
> >
> > I can help with the onconfig if you want.
>
> DISCLAIMER: This e-mail and any files transmitted with it are confidential
> and intended solely for the use of the individual or entity to whom they
> are addressed. Any views or opinions expressed are those of the author and
> do not represent the views of Poole Hospital NHS Trust unless otherwise
> explicitly stated. The information contained in this e-mail may be subject
> to public disclosure under the Freedom of Information Act 2000. Unless the
> Information is legally exempt from disclosure, the confidentiality of this
> e-mail and your reply cannot be guaranteed.
___________________________________
Yahoo! Messenger: chiamate gratuite in tutto il mondo
http://it.beta.messenger.yahoo.com
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:50:20 EDT