configuration of disks

I need help understanding how this works. We're using SS2000 Standard on Windows 2000 Server. I've been reading about spreading the data in a database or a table across several drives using filegroups. Since most of our work is extract, transfer and load and involves a lot of writes, it seems that raid 10 is better than 5 although more expensive. Let's say we determine that we need 3 36G drives just for data (forgetting logs for now). If we create one raid 10 using 6 drives does this mean that we address that one raid 10 with one drive letter, say drive e:? If so, does this mean that we wouldn't be able to take advantage of filegroups because everything would be on drive e:
However, if we created 3 raid 10s of two drives each then we would have drives e:, f:, and g: to work with. Correct? And then we would be able to use filegroups by creating a filegroup for each drive letter and then creating a database with 3 files, one in each filegroup
Is that correct or do I not understand how it works
Thanks
Dan D.Once you create a RAID 10, all you get is a big chuck of space, you can
still partition it into different logical drives, but everything is
spread out across those 3 disks (6 disks if you count the mirror). If
you want to spread out filegroup, then you have two options 1) forget
about RAID, put each filegroup on differnet physical disk 2) create a
set of RAID, put each filegroup on different RAID, this can be very
expensive
With 6 disks, you can't have 3 RAID 10, you have 3 RAID 1. This
basically is option 2 listed above, you can spread out filegroups into
different physical disks. Provide you have a really good SCSI card, this
may not be a bad choice.
Eric Li
SQL DBA
MCDBA
Dan wrote:
> I need help understanding how this works. We're using SS2000 Standard on Windows 2000 Server. I've been reading about spreading the data in a database or a table across several drives using filegroups. Since most of our work is extract, transfer and load and involves a lot of writes, it seems that raid 10 is better than 5 although more expensive. Let's say we determine that we need 3 36G drives just for data (forgetting logs for now). If we create one raid 10 using 6 drives does this mean that we address that one raid 10 with one drive letter, say drive e:? If so, does this mean that we wouldn't be able to take advantage of filegroups because everything would be on drive e:?
> However, if we created 3 raid 10s of two drives each then we would have drives e:, f:, and g: to work with. Correct? And then we would be able to use filegroups by creating a filegroup for each drive letter and then creating a database with 3 files, one in each filegroup.
> Is that correct or do I not understand how it works?
> Thanks,
> Dan D.|||"Dan" <anonymous@.discussions.microsoft.com> wrote in message
news:4A634A1A-1174-4048-9471-2F5EA0FD1D21@.microsoft.com...
> I need help understanding how this works. We're using SS2000 Standard on
Windows 2000 Server. I've been reading about spreading the data in a
database or a table across several drives using filegroups. Since most of
our work is extract, transfer and load and involves a lot of writes, it
seems that raid 10 is better than 5 although more expensive. Let's say we
determine that we need 3 36G drives just for data (forgetting logs for now).
If we create one raid 10 using 6 drives does this mean that we address that
one raid 10 with one drive letter, say drive e:? If so, does this mean that
we wouldn't be able to take advantage of filegroups because everything would
be on drive e:?
> However, if we created 3 raid 10s of two drives each then we would have
drives e:, f:, and g: to work with. Correct? And then we would be able to
use filegroups by creating a filegroup for each drive letter and then
creating a database with 3 files, one in each filegroup.
> Is that correct or do I not understand how it works?
Don't do RAID striping and Filegroups. Just pick one. If you have 6 disks
in RAID 10, you are already spreading out the IO over a 3 disk stripe set in
hardware. If you really want to use filegroups, you should just use RAID
for mirroring, and deploy the 6 disks in 3 2-disk mirror sets. Then you
spread out your files over the 3 disks.
David|||As far as seeking the optimal performance, can you recommend using the raid 10 over the raid 1 and filegroups or vice versa
Thanks
Dan D
-- David Browne wrote: --
"Dan" <anonymous@.discussions.microsoft.com> wrote in messag
news:4A634A1A-1174-4048-9471-2F5EA0FD1D21@.microsoft.com..
> I need help understanding how this works. We're using SS2000 Standard o
Windows 2000 Server. I've been reading about spreading the data in
database or a table across several drives using filegroups. Since most o
our work is extract, transfer and load and involves a lot of writes, i
seems that raid 10 is better than 5 although more expensive. Let's say w
determine that we need 3 36G drives just for data (forgetting logs for now)
If we create one raid 10 using 6 drives does this mean that we address tha
one raid 10 with one drive letter, say drive e:? If so, does this mean tha
we wouldn't be able to take advantage of filegroups because everything woul
be on drive e:
>> However, if we created 3 raid 10s of two drives each then we would hav
drives e:, f:, and g: to work with. Correct? And then we would be able t
use filegroups by creating a filegroup for each drive letter and the
creating a database with 3 files, one in each filegroup
>> Is that correct or do I not understand how it works
Don't do RAID striping and Filegroups. Just pick one. If you have 6 disk
in RAID 10, you are already spreading out the IO over a 3 disk stripe set i
hardware. If you really want to use filegroups, you should just use RAI
for mirroring, and deploy the 6 disks in 3 2-disk mirror sets. Then yo
spread out your files over the 3 disks
Davi|||I forgot to mention that we're moving to a SAN. I don't know if that makes any difference.
So, you're saying that the raid is created first with the six drives and then we could partition 3 logical drives on the raid 1? Isn't the object of having the 3 filegroups on 3 different drives so you while the heads of one drive are accessing some of the data, the heads another drive could be accessing some more of the data? After creating the raid 1, is it possible to create the three partitions one on each physical drive
Can you explain why with 6 disks we can't have 3 raid 10? And what advantage is there to adding a raid 0 to a raid 1
Thanks
Dan D
-- Eric.Li wrote: --
Once you create a RAID 10, all you get is a big chuck of space, you can
still partition it into different logical drives, but everything is
spread out across those 3 disks (6 disks if you count the mirror). If
you want to spread out filegroup, then you have two options 1) forget
about RAID, put each filegroup on differnet physical disk 2) create a
set of RAID, put each filegroup on different RAID, this can be very
expensiv
With 6 disks, you can't have 3 RAID 10, you have 3 RAID 1. This
basically is option 2 listed above, you can spread out filegroups into
different physical disks. Provide you have a really good SCSI card, this
may not be a bad choice
--
Eric L
SQL DB
MCDB
Dan wrote
> I need help understanding how this works. We're using SS2000 Standard on Windows 2000 Server. I've been reading about spreading the data in a database or a table across several drives using filegroups. Since most of our work is extract, transfer and load and involves a lot of writes, it seems that raid 10 is better than 5 although more expensive. Let's say we determine that we need 3 36G drives just for data (forgetting logs for now). If we create one raid 10 using 6 drives does this mean that we address that one raid 10 with one drive letter, say drive e:? If so, does this mean that we wouldn't be able to take advantage of filegroups because everything would be on drive e:
>> However, if we created 3 raid 10s of two drives each then we would have drives e:, f:, and g: to work with. Correct? And then we would be able to use filegroups by creating a filegroup for each drive letter and then creating a database with 3 files, one in each filegroup
>> Is that correct or do I not understand how it works
>> Thanks
>> Dan D.|||In article <F4FC777D-69AC-408C-A2B0-FD19A7104BAF@.microsoft.com>,
anonymous@.discussions.microsoft.com said...
> I forgot to mention that we're moving to a SAN. I don't know if that makes any difference.
> So, you're saying that the raid is created first with the six drives and then we could partition 3 logical drives on the raid 1? Isn't the object of having the 3 filegroups on 3 different drives so you while the heads of one drive are accessing some of the data, the heads another drive could be accessing some more of the data? After creating the raid 1, is it possible to create the three partitions one on each
physical drive?
> Can you explain why with 6 disks we can't have 3 raid 10? And what advantage is there to adding a raid 0 to a raid 1?
With pairs of disks you only have the option of RAID 1 or RAID 0. For a
single RAID 10 you need at least 4 disks so to do 3 of them you would need
12 disks. Do a web search on RAID and you'll understand the various
levels pretty quickly.|||"Dan D." <anonymous@.discussions.microsoft.com> wrote in message
news:43CACB17-1C0C-46B7-863D-1F27E7124962@.microsoft.com...
> As far as seeking the optimal performance, can you recommend using the
raid 10 over the raid 1 and filegroups or vice versa?
>
Personally I never use filegroups. But I can't really make a recommendation
since in principle both approaches should be pretty close in performance. I
just think doing this a the disk level is easier than integrating it into
your physical database design.
Davidsqlsql