Bench Mark on Performance

Hi
I'm looking quite heavily at SQL Server Profiler (2000 sp3a).
I understand that the profiler looks at the logical reads and physical
writes.
What I'm trying to understand is what people use as a guide as to
whether something should be dedicated or if it could be shared on an
active/passive clustered environment.
If I'm recieving a million reads per hour, they are logical, hence in
the buffer cache, but does that mean the server is performing well? I
know that it's not always down to just read and writes but that is the
essence of a db system... so that's where I start.
Your thoughts would be most appreciated.
SimonHi
I'd also look at Performance Monitor to examine PhysicalDisk Object: Avg.
Disk Queue Length
http://www.sql-server-performance.com/performance_monitor_counters_io.asp
Per a query you can use SET STATISTICS IO command
When looking at individual queries ,the Logical Reads is more interesting
because the ibfo is consistent
Pay close attention as well to chache-hit ratio. (>=98%)
<simonaarons@.yahoo.co.uk> wrote in message
news:1129024856.929599.263150@.o13g2000cwo.googlegroups.com...
> Hi
> I'm looking quite heavily at SQL Server Profiler (2000 sp3a).
> I understand that the profiler looks at the logical reads and physical
> writes.
> What I'm trying to understand is what people use as a guide as to
> whether something should be dedicated or if it could be shared on an
> active/passive clustered environment.
> If I'm recieving a million reads per hour, they are logical, hence in
> the buffer cache, but does that mean the server is performing well? I
> know that it's not always down to just read and writes but that is the
> essence of a db system... so that's where I start.
> Your thoughts would be most appreciated.
> Simon
>|||Simon,
This is where a benchmark for your system would help. This would provide
you with how the system operates during normal operations. The performance
values derived from the benchmark can be compared against over time to
establish changes in the environment and to proactively predict future HW
upgrades/additions.
HTH
Jerry
<simonaarons@.yahoo.co.uk> wrote in message
news:1129024856.929599.263150@.o13g2000cwo.googlegroups.com...
> Hi
> I'm looking quite heavily at SQL Server Profiler (2000 sp3a).
> I understand that the profiler looks at the logical reads and physical
> writes.
> What I'm trying to understand is what people use as a guide as to
> whether something should be dedicated or if it could be shared on an
> active/passive clustered environment.
> If I'm recieving a million reads per hour, they are logical, hence in
> the buffer cache, but does that mean the server is performing well? I
> know that it's not always down to just read and writes but that is the
> essence of a db system... so that's where I start.
> Your thoughts would be most appreciated.
> Simon
>

BCVs Continuation...

I am really curious with all these SAN technologies and that all of it calls
for quiescing the disks i.e. stopping writes for a brief sec(s) to take a
snapshot..
How does one write applications or consider uptime in a situation as such ?
Even a slight blip and application tend to fall on their face . How does one
work around this on a very busy OLTP system ? Any suggestions ?
Thanks
"Hassan" <Hassan@.hotmail.com> wrote in message
news:OqR$3KrPHHA.1756@.TK2MSFTNGP05.phx.gbl...
> I am really curious with all these SAN technologies and that all of it
calls
> for quiescing the disks i.e. stopping writes for a brief sec(s) to take a
> snapshot..
I would suggest you read up on "DBCC FREEZE_IO" and associated commands, and
contact your SAN storage vendor to see if they have any layered products to
assist you in this. Many storage vendors will have packages which on one
hand integrate with SQL server (VDI interface) and on the other hand
integrate to the storage to automate these type of executions.

> How does one write applications or consider uptime in a situation as such
?
Applications should be written as normal, how you consider uptime is up to
you, if you freeze IO (to disk) for 5 seconds, but don't disconnect your
clients... is that up or down ? you need to answer that question for
yourself.

> Even a slight blip and application tend to fall on their face .
Go back to the Application developers, even when for some reason the
application gets disconnected, it should have mechanisms build in to
reconnect at lease configurable say retry 3 times and time out of x
seconds... if an application "fall on their face" upon a disconnect... then
the application will fail upon a cluster-move-group... it really should try
to reconnect... with all respect, but this is bad client app programming
|||Technically, they don't all have to quiesce the disk I/Os. Snapshots/BCVs can
be taken without any integration with SQL Server. But that is not really the
right way to work with SQL Server. Our experience is that without the SQL
Server integration piece, BCV split may not always produce good volumes for
database recovery (more so on Windows 2000 than on Windows 2003) unless
strigent operations procedures are followed.
Linchi
"Hassan" wrote:

> I am really curious with all these SAN technologies and that all of it calls
> for quiescing the disks i.e. stopping writes for a brief sec(s) to take a
> snapshot..
> How does one write applications or consider uptime in a situation as such ?
> Even a slight blip and application tend to fall on their face . How does one
> work around this on a very busy OLTP system ? Any suggestions ?
> Thanks
>
>