[Info-Ingres] Ingres 2.6 on Solaris 10: createdb slow

William Yuan yuan at ariel.its.unimelb.edu.au
Mon Apr 23 01:50:54 CDT 2007


Hi Artur,

	Thanks for your reply. I have been initially testing on a 2 Gb cooked
	log on ufs with RAID 5. Our I/O throughput and latency on a Solaris
	V890 zone appears to be 50-200% quicker than the older Dec AlphaServer,
	according to the bonnie++ disk benchmarks. We plan to trial using a
	rawlog in the next few days.

	The best time I could achieve with Ingres 2006 on a Solaris 10 zone was
	22 seconds. The timing is still a fair way off from what I had hoped for.


Kind Regards,

	William

>Hi William,
>
>How your transaction log is configured? Maybe it's on zfs?
>If it is, then it's rather bad idea, and gives similar results.
>
>Regards,
>Artur
>
>________________________________
>
>ProService Agent Transferowy Sp. z o.o. z siedzib± w Warszawie, ul.
>Pu³awska
>436, 02-801 Warszawa,
>S±d Rejonowy dla m.st. Warszawy, XIII Wydzia³ Gospodarczy Krajowego
>Rejestru
>S±dowego,
>rejestr przedsiêbiorców Krajowego Rejestru S±dowego - KRS 0000013069
>Kapita³ zak³adowy wynosi 5.050.000,00 z³otych i wp³acony jest w ca³o¶ci,
>NIP
>526-25-07-903
>
>
> > -----Original Message-----
> > From: William Yuan [mailto:yuan at ariel.its.unimelb.edu.au]
> > Sent: Monday, April 23, 2007 8:10 AM
> > To: info-ingres at kettleriverconsulting.com
> > Subject: [Info-Ingres] Ingres 2.6 on Solaris 10: createdb slow
> >
> > Hi there,
> >
> > 	I am configuring a new installation of Ingres 2.6 on
> > Solaris 10 zones and
> > 	seeing some performance issues. In particular, I'm
> > seeing some very slow
> > 	createdb times which is unusual:
> >
> >     Sun V890 (8 cpus @ 1.5 Ghz), 16 Gb RAM, Solaris 10
> > zoned):  22-70 seconds # Snail
> >     Dec AlphaServer ES47 (2 cpus @ 1 Ghz, 8 Gb RAM, DUnix
> > 5.1): 5-8 seconds   # Normal
> >
> > 	Has anyone seen this kind of slowdown on Solaris 10
> > before ? I had
> > 	similar timings on the createdb when trialling Ingres
> > 2006 as well.
> >
> > 	I wonder if "de-zoning" the box would be a viable
> > alternative in this	situation.
> >
> > Kind Regards,
> >
> > 	William



More information about the Info-Ingres mailing list