Choosing drives, transfert rate or IO/sec?

Hi,
what is most important to focus on when we choose a drive setup to support a
small datawarehouse?
(10GB and less)
IO/Sec
or
MB/s
?
(same question regarding the tempdb database storage)
except the SQLIO and SQLIOStress, there is any testing tool which simulate
DW loading process & DW query process?
thanks.
Jerome."Jj" <willgart@.BBBhotmailAAA.com> wrote in message
news:uPa1X68wFHA.1148@.TK2MSFTNGP11.phx.gbl...
> Hi,
> what is most important to focus on when we choose a drive setup to support
> a small datawarehouse?
> (10GB and less)
> IO/Sec
> or
> MB/s
> ?
> (same question regarding the tempdb database storage)
>
Write throughput.
For a small datawarehouse, your memory cache should be a large percentage of
your database size. So for physical IO, sequential write operations (like
the log file and checkpointing) will predominate.
David|||Hi Jerome,
for 10GB or less of data....design a reasonable model and throw a bit
more hardware at it if it is not going ok......the time and money it
will cost to think about how to tune it is more than the cost of the
HW/SW to run it....
Of course, this kind of advice does NOT apply to large DWs where we do
spend time considering the performance in some great detail..
Best Regards
Peter Nolan
www.peternolan.com

Choosing drives, transfert rate or IO/sec?

Hi,
what is most important to focus on when we choose a drive setup to support a
small datawarehouse?
(10GB and less)
IO/Sec
or
MB/s
?
(same question regarding the tempdb database storage)
except the SQLIO and SQLIOStress, there is any testing tool which simulate
DW loading process & DW query process?
thanks.
Jerome.
"Jj" <willgart@.BBBhotmailAAA.com> wrote in message
news:uPa1X68wFHA.1148@.TK2MSFTNGP11.phx.gbl...
> Hi,
> what is most important to focus on when we choose a drive setup to support
> a small datawarehouse?
> (10GB and less)
> IO/Sec
> or
> MB/s
> ?
> (same question regarding the tempdb database storage)
>
Write throughput.
For a small datawarehouse, your memory cache should be a large percentage of
your database size. So for physical IO, sequential write operations (like
the log file and checkpointing) will predominate.
David
|||Hi Jerome,
for 10GB or less of data....design a reasonable model and throw a bit
more hardware at it if it is not going ok......the time and money it
will cost to think about how to tune it is more than the cost of the
HW/SW to run it....
Of course, this kind of advice does NOT apply to large DWs where we do
spend time considering the performance in some great detail..
Best Regards
Peter Nolan
www.peternolan.com

CheckPoint Pages/Sec counter

In EM, changed the checkpoint interval to 1. I monitored the CheckPoint
Pages/Sec counter and inserted 3000 records and waiting 10 minutes, I have
not seen any activity on this counter.
SQL 2000 SP4
Thanks,
Don
I'm trying to solve a problem where during the CheckPoint Pages/Sec that SQL
initates, i see in the Profiler: EventClass- "Batch Completed" has long
Duration times and impacts customers, i.e., their data that is expected to be
retrieved in a timely fashion is many, many times longer. I'm trying to gain
control of the CheckPoint interval.
Note that the frequency of checkpoints in a database does not depend on any
time-based measure. Rather, it depends on the amount of data modifications
that have been made and SQL Server's estimate of how long it may take to
rollforward these changes. If in its estimate it would take the 'recovery
interval" number of minutes (when the recovery interval is not set to 0), it
issues a checkpoint.
So in your case, one potential explanation is that you have not made enough
changes for SQL Server to think it needs one minute to rollforward the
changes, thus no checkpoint is necessary.
Linchi
"donsql22222" wrote:

> In EM, changed the checkpoint interval to 1. I monitored the CheckPoint
> Pages/Sec counter and inserted 3000 records and waiting 10 minutes, I have
> not seen any activity on this counter.
> SQL 2000 SP4
> Thanks,
> Don
> I'm trying to solve a problem where during the CheckPoint Pages/Sec that SQL
> initates, i see in the Profiler: EventClass- "Batch Completed" has long
> Duration times and impacts customers, i.e., their data that is expected to be
> retrieved in a timely fashion is many, many times longer. I'm trying to gain
> control of the CheckPoint interval.