Subject: | |
From: | |
Reply To: | |
Date: | Fri, 12 Jan 2001 09:25:55 -0800 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Sorry I wasn't clear. That's what I get for writing so quickly.
In this case the intuitive answer is correct. More spindles are
better. Much better. I saw two egregious mistakes in my original
posting. Let me post a corrected statement:
>Regarding Summit's recommendation to use the 9gb drives configured as 4gb
>drives that is likely a performance recommendation. One with which I
>agree, I might add. The Summit application SPECTRUM is highly
>spindle-sensitive. We (Lancaster Consulting) and Summit independently
>came to the conclusion that it's best to configure more disk space with
>fewer spindles than the reverse. (To keep a long story short, CSY
>benchmarks proved this out.)
That should be "it's best to configure larger disks with no more than about
four gb and not allowing your spindle count to drop."
>In fact, in a recent High Availability project for a Summit customer we
>configured two fully-populated HVD10 disk subsystems, using only have the
>actual disk, implemented Mirrored Disk and saw excellent results.
This should be "using only "half" the actual disk.
I apologize for the confusion. Abjectly.
Bill
At 08:31 AM 1/12/01, Mark Landin wrote:
>On Wed, 10 Jan 2001 18:45:49 -0600 (Central Standard Time), Bill
>Lancaster <[log in to unmask]> wrote:
>
> >The Summit application SPECTRUM is highly
> >spindle-sensitive. We (Lancaster Consulting) and Summit independently came
> >to the conclusion that it's best to configure more disk space with fewer
> >spindles than the reverse. (To keep a long story short, CSY benchmarks
> >proved this out.)
>
>Are you saying then to *reduce* spindle count? Or did I
>misunderstand....?
>
>If you are indeed saying to reduce spindle count .. this seems very
>counterintuitive. Can you give an explanation why SPECTRUM seems to
>function best in an environment which goes against the current wisdom,
>if indeed that is the case?
|
|
|