We're upgrading SQL Server from 2000 to 2005 and installing 2005 on a brand
new server...I have a question about the RAID configuration.

We're getting a Dell PowerEdge 2970 with 8 drives, all SAS 2.5" 10K and 300
GB in size.

I know 4 of those drives will be setup in RAID 10 and will hold the main
data files. My question is about the other 4 drives.

The options are:

A. Configure 2 drives in RAID 1 for OS, tempdb, backup and another 2 drives
in RAID 1 for transaction logs & backup

B. Configure all 4 drives with RAID 10 in one array and put OS, tempdb,
transaction logs and backup all together

I'd rather go with option B because that gives me flexibility in how storage
is used - 600 GB total - (as opposed to having to work with 2 separate arrays
at 300 GB each) but I'm not sure what the negative impacts will be (if any)
from having all those things together.

This is for an OLTP database (ecommerce site) and there will be just one
RAID controller installed in the server.

Is it a bad idea to combine OS, tempdb, transaction logs into a single RAID
10 array? Or do I need to use two arrays in RAID 1 (option A) and lose the
space flexibility?

Thanks.

How many transactions per second are you talking about? Are you concerned
with performance and scalability? If so then you should keep the log files
on a Raid 1 by themselves or on the Raid 1 with the OS since it is not used
that much. Backups should never be placed on the same array as the user data
or log files since if that array goes south you loose the backups and the
data or log. You might consider this:

Raid 1 OS & Log files (including tempdb log file)
Raid 1 Tempdb data and Backups
Raid 10 User data files

But it can really depend on how you use the database.


--
Andrew J. Kelly SQL MVP
Solid Quality Mentors


"devprog (AT) msbcomp (DOT) com" <devprogmsbcompcom (AT) discussions (DOT) microsoft.com> wrote in
message news:2B507DF1-FE80-4D9F-B3F2-7EAA661B8297 (AT) microsoft (DOT) com...

I looked at the Transaction/sec counter for the _Total instance and it ranges
between 50-200 - never saw it higher than that during normal operations. How
can I tell based on this what kind of load we have on the server?

It seems pretty light to me - are there some charts out there, some kind of
guiding priciples on how to interpret this data?





"Andrew J. Kelly" wrote:

Not really as a transaction is not a constant. It can depend heavily on what
the transaction is and the amount and type of data being modified. And most
of that is probably coming from Tempdb but you can't tell just from the
total counter, you need to break it out to see for sure. But I would still
recommend the configuration I posted. Normally without more specifics I
would say add the two disks from the Raid 1 for Tempdb to the Raid 10 and
just put the tempdb data on it as well. But since you need a place for the
backups anyway and tempdb is rebuilt from scratch each time it starts up
that is probably a good place to put it.

--
Andrew J. Kelly SQL MVP
Solid Quality Mentors


"devprog (AT) msbcomp (DOT) com" <devprogmsbcompcom (AT) discussions (DOT) microsoft.com> wrote in
message news:3E07DB14-0410-419B-A9F7-9766C30B7956 (AT) microsoft (DOT) com...