Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.misc -> Re: NT vs. Unix
Walter,
I would argue that SCSI disks can be mechanically faster ie 10K rpm opposed to IDE of about 5.4K.
I think Gerrit-Jan needs to monitor the disk io queues and check the sql thats being executed. Even with a row size of 1k he should be able to cache the table, and not use the indexes.
Chris
Walter Dnes wrote:
>
> On 14 Feb 1999 16:33:32 GMT, gjlinker_at_aol.com (Gjlinker) wrote:
>
> >Environment:
> >Compac server dual processor P200
> >NT4 + Oracle 7.3 + 600 MB memory + 4 GB disk
> >10 users selecting 100 rows from a 150.000 row table using indexes.
> >15 VB programs connected via ODBC running on the server.
> >
> >Result:
> >Processors busy for most of the time 75-100%
> >Very slow performance due to high write times of Oracle.
> >
> >Analysis:
> >Apparently PC hardware does not write to disk when processor is busy over 80%
> >creating huge write times in Oracle.
>
> One thing I don't see mentioned... does the machine have a
> full-fledged Ultra-Wide SCSI disk drive, or is drive a cheap-o
> level-1 EIDE where the CPU has to babysit each byte being
> read or written from/to the drive? SCSI drives aren't
> mechanically faster, but they do have their own chips to
> handle scatter/gather, and do requested operations on their
> own. Ideally, the main CPU issues a read or write request,
> and a pointer to an area in memory, then disconnects. The
> SCSI carries out its orders and issues an interrupt to the
> main CPU when finished. In the meantime, the main CPU can be
> number-crunching. You may not see a difference in single-
> tasking DOS, but it's quite important in a true multi-
> tasking environment.
>
> Walter Dnes <waltdnes_at_interlog.com> procmail spamfilter
> http://www.interlog.com/~waltdnes/spamdunk/spamdunk.htm
> Why a fiscal conservative opposes Toronto 2008 OWE-lympics
> http://www.interlog.com/~waltdnes/owe-lympics/owe-lympics.htm
Received on Mon Feb 15 1999 - 03:42:10 CST
![]() |
![]() |