Here's what I am trying to do:
   Place 6 1000MB Sybase database devices (raw partitions) on RAID-0/1.
   I have 9 1GB drives for striping and a second set of 9 1GB drives
   to be used for the mirror.  (The remaining 3GB of space will be
   used to store some misc Sybase db partitions which have relatively
   light i/o requirements such as Rep Server RSSD databases).

   It appears that you cannot bind physical disks into a RAID stripe.
   Rather, striping is done at the volume (raw partition) level.  So,
   I created my 6 1GB db devices such that each was a volume striped
   over 5 different disks.

   Volume: dbdev1               dbdev2               dbdev3          ...
   Plex:      dbdev1-strp-pl       dbdev2-strp-pl       dbdev3-strp-pl
   Subdisks:    disk1-01 200MB        disk6-01             disk2-02
                disk2-01 200MB        disk7-01             disk3-02
                disk3-01 200MB        disk8-01             disk4-02
                disk4-01 200MB        disk9-01             disk5-02
                disk5-01 200MB        disk1-02             disk6-02

   My goal was to balance the i/o for these 6 very write-intense volumes
   over 9 drives.  And this approach has worked extremely well. :-)
   Now I need to add mirroring to my striped configuration... I am
   planning to duplicate this config on the remaining 9 disks and then
   associate the new plexes (eg dbdev1-strp-pl-2) with the volume that
   is being mirrored (eg dbdev1).  Will this work?  Will mirroring this
   configuration degrade the excellent performance I now have?

Misc Questions:
1. How do I determine the size of the NVRAM cache in each disk array?
2. How do I allocate NVRAM to specific volumes (vs disk drives)?
3. Is NVRAM safe to use with raw database partitions? Safe to use for
   all writes or only synchronous writes?  If left disabled for writes,
   does this imply it IS being used for reads?
4. What's the best way to add space to a striped filesystem?  I tried the
   Basic-Ops menu choice "Resize Filesystem" and I'm still not sure why
   it created 10 subdisks to add 500MB...  The original filesystem was
   2500MB, striped over 5 disks.  The 10 new subdisks were added to the
   existing plex...  But I have no idea if and how they were striped.
5. Is there a FAQ that has answers to questions like these?

___________________________________
Dale Benedict
MCI, Research Triangle Park, NC

Phone: 919-572-3265

|> Hello.  I'm working with SUN and its SPARCStorage Array for the first time
|> and am in need of some help.  I am quite familiar with the concepts of RAID
|> and have done extensive work in configuring RDBMS sytems with
|> Data General's CLARiiON Disk Array.  However, reading the User's Guide
|> for SPARCstorage Array has left me with the impression SUN's implementation
|> is not as robust at DG's...  The optimist in me says
|> that SUN's is probably as good, but I just haven't figured out the
|> way it's done on Sun.  Any help you offer will greatly be appreciated!
|>
It works fine, as long as you don't have PrestoServe (which it sounds
like you don't, and even if you did, couldn't use anyway)

|> Here's what I am trying to do:
|>    Place 6 1000MB Sybase database devices (raw partitions) on RAID-0/1.
|>    I have 9 1GB drives for striping and a second set of 9 1GB drives
|>    to be used for the mirror.  (The remaining 3GB of space will be
|>    used to store some misc Sybase db partitions which have relatively
|>    light i/o requirements such as Rep Server RSSD databases).
|>
|>    It appears that you cannot bind physical disks into a RAID stripe.
|>    Rather, striping is done at the volume (raw partition) level.  So,
|>    I created my 6 1GB db devices such that each was a volume striped
|>    over 5 different disks.
|>  
You assign a physical disk to a VM disk. Then take your subdisk
(which just happens to be the whole disk) out of the VM disk.
Stripe the VM disks.
Your subdisks can be as big or as little as you want. You just tell it
how big you want them. Then choose a bunch of those disks and create a
volume out of it. (Or let it pick a bunch of disks itself). It sounds
like this is what you're doing.

|>    Volume: dbdev1               dbdev2               dbdev3          ...
|>    Plex:      dbdev1-strp-pl       dbdev2-strp-pl       dbdev3-strp-pl
|>    Subdisks:    disk1-01 200MB        disk6-01             disk2-02
|>                 disk2-01 200MB        disk7-01             disk3-02
|>                 disk3-01 200MB        disk8-01             disk4-02
|>                 disk4-01 200MB        disk9-01             disk5-02
|>                 disk5-01 200MB        disk1-02             disk6-02
|>
|>    My goal was to balance the i/o for these 6 very write-intense volumes
|>    over 9 drives.  And this approach has worked extremely well. :-)
|>    Now I need to add mirroring to my striped configuration... I am
|>    planning to duplicate this config on the remaining 9 disks and then
|>    associate the new plexes (eg dbdev1-strp-pl-2) with the volume that
|>    is being mirrored (eg dbdev1).  Will this work?  Will mirroring this
|>    configuration degrade the excellent performance I now have?
|>
Mirroring should not degrade your performance IF you mirror across
controllers. (if the mirrored disks and striped disks are not on the
same controller, you should be fine)

|> Misc Questions:
|> 1. How do I determine the size of the NVRAM cache in each disk array?
|> 2. How do I allocate NVRAM to specific volumes (vs disk drives)?
|> 3. Is NVRAM safe to use with raw database partitions? Safe to use for
|>    all writes or only synchronous writes?  If left disabled for writes,
|>    does this imply it IS being used for reads?

NVRAM will be useless on raw database partitions, your writes are already
asynchronous.

|> 4. What's the best way to add space to a striped filesystem?  I tried the
|>    Basic-Ops menu choice "Resize Filesystem" and I'm still not sure why
|>    it created 10 subdisks to add 500MB...  The original filesystem was
|>    2500MB, striped over 5 disks.  The 10 new subdisks were added to the
|>    existing plex...  But I have no idea if and how they were striped.

The best way would be to increase the size of each of the subdisks (assuming
the space allocated and there was contiguous room on the physical disk to
do so) and then resize that way. The new disks were most likely striped
among themselves but concatenated to the original volume.

|> 5. Is there a FAQ that has answers to questions like these?
|>  

Answerbook has most of the answers you probably want. I'm not aware of
any FAQ.
--
____________________________________________________________________________
Doug Hughes                                     Engineering Network Services
System/Net Admin                                Auburn University