Data Storage - Using Veritas Foundation Suite to stripe a H/W striped Raid set

This is Interesting: Free IT Magazines  
Home > Archive > Data Storage > December 2005 > Using Veritas Foundation Suite to stripe a H/W striped Raid set





You are viewing an archived Text-only version of the thread. To view this thread in it's original format and/or if you want to reply to this thread please [click here]

Author Using Veritas Foundation Suite to stripe a H/W striped Raid set
apples

2005-12-13, 5:49 pm

We have a SUN cluster which is SAN connected to two Sun 3510FC storage
arrays. The disks are physically striped and mirrored at the hardware
level in the arrays, however when the Volumes were created using
Veritas Foundation Suite they were concatenated as opposed to striped.

In reading some other vendors documentation for connecting to their
storage and using their LVM software, they recommend striping the
Logical Volumes at the software level even when the disks LUN's are
already striped. We've done this and have noticed significant
performance gains.

Does this make sense and can someone provide me with either Veritas or
SUN documentation which supports this.

Bill Todd

2005-12-14, 5:48 pm

apples wrote:
> We have a SUN cluster which is SAN connected to two Sun 3510FC storage
> arrays. The disks are physically striped and mirrored at the hardware
> level in the arrays, however when the Volumes were created using
> Veritas Foundation Suite they were concatenated as opposed to striped.
>
> In reading some other vendors documentation for connecting to their
> storage and using their LVM software, they recommend striping the
> Logical Volumes at the software level even when the disks LUN's are
> already striped. We've done this and have noticed significant
> performance gains.
>
> Does this make sense


Yes: instead of distributing your accesses across only half your disks,
you're distributing them over all of them. If they're large, streaming
accesses (involving presumably logically contiguous files), you're
getting the combined bandwidth of both arrays rather than that of just
one; if they're small, random accesses, then if there's any significant
locality (e.g., not much of your second array was actually populated
with data in the concatenated configuration), you're getting the
combined IOPS of both arrays rather than that of just one.

and can someone provide me with either Veritas or
> SUN documentation which supports this.


Not me - that's why I left this for a while for someone else to answer.

- bill
Sponsored Links






Free braindumps | Software forum | Database administration forum

Copyright 2003 - 2008 webservertalk.com