Ruling out OS - SQL Server Performance.

Hi:
I have a question hopefully the SQL Server Pros and MVPs would have an
answer for. Most of the time like any normal DBA to find performance issues
with SQL Server I make use of the performance counters through performance
monitor or system monitor and SQL Server Profiler and then rule out OS as
the bottleneck and then concentrate on the application part. My question
here is if there a simple way that without using any of these counters or
tools a DBA can rule out OS as the potential bottleneck. For example I would
assume that if my backups start taking a long time all of a sudden when they
usually take less time every day then I can probably point out OS as the
issue. I am not sure if thats the right way to tell.
I am looking for a comprehensive list of possible symptoms (like paging,
Virtual memory, Disk I/O) that can conclusively deduce that OS is/is not
the bottle neck without me making use of any of the above mentioned tools.
Any KB Article or a White paper or blogs with more information would be
really helpful. Any other information is greatly appreciated.
Thanks for your attention.
M
Meher
Well, my first question is what is your SQL Server's verion?
If you are using SQL Server 2005 you have lots of system catalogs,views that
contain such kind of info.Personally , I have not played with it
unfortunatly but as I heard it provides a comprehensive info about internal
behaviour of SQL Server as well as OS.
If you use SQL Server 2000 I have my doubt that without these tolls you can
roll out OS as the potential bottleneck at 'first glance'.
"Meher" <NOSPAM_mmsagar@.hotmail.com> wrote in message new, views
s:uIdeYK$JGHA.532@.TK2MSFTNGP15.phx.gbl...
> Hi:
> I have a question hopefully the SQL Server Pros and MVPs would have an
> answer for. Most of the time like any normal DBA to find performance
> issues with SQL Server I make use of the performance counters through
> performance monitor or system monitor and SQL Server Profiler and then
> rule out OS as the bottleneck and then concentrate on the application
> part. My question here is if there a simple way that without using any of
> these counters or tools a DBA can rule out OS as the potential bottleneck.
> For example I would assume that if my backups start taking a long time all
> of a sudden when they usually take less time every day then I can probably
> point out OS as the issue. I am not sure if thats the right way to tell.
> I am looking for a comprehensive list of possible symptoms (like paging,
> Virtual memory, Disk I/O) that can conclusively deduce that OS is/is not
> the bottle neck without me making use of any of the above mentioned tools.
> Any KB Article or a White paper or blogs with more information would be
> really helpful. Any other information is greatly appreciated.
> Thanks for your attention.
> M
>
|||Hi Uri:
My Version is SQL Server 2000. You are right that without using counters how
can a DBA Can say that OS Can be ruled out and that the Application is
possibly the bottleneck. That I assumed is my limited knowledge of
performance and if there are some kind of telltale signs that a DBA Can
point out without even using tools. If thats not possible then i guess the
tools are the only one i can rely on.
Thanks Uri for your response.
Meher
"Uri Dimant" <urid@.iscar.co.il> wrote in message
news:uXhj3U$JGHA.3912@.TK2MSFTNGP10.phx.gbl...
> Meher
> Well, my first question is what is your SQL Server's verion?
> If you are using SQL Server 2005 you have lots of system catalogs,views
> that contain such kind of info.Personally , I have not played with it
> unfortunatly but as I heard it provides a comprehensive info about
> internal behaviour of SQL Server as well as OS.
>
> If you use SQL Server 2000 I have my doubt that without these tolls you
> can roll out OS as the potential bottleneck at 'first glance'.
>
>
>
> "Meher" <NOSPAM_mmsagar@.hotmail.com> wrote in message new, views
> s:uIdeYK$JGHA.532@.TK2MSFTNGP15.phx.gbl...
>
|||Meher
Yes , based on my experience 80% of the problems ( lack of indexes, bad
written queries) caused by an application and only 10% -20% caused by a
hardware. I used to capture some events by using SQL Server Profiler and it
is absolutely reliable tool that provides very valuable info as well as
Perf. Monitor
"Meher" <NOSPAM_mmsagar@.hotmail.com> wrote in message
news:ewjsHY$JGHA.2036@.TK2MSFTNGP14.phx.gbl...
> Hi Uri:
> My Version is SQL Server 2000. You are right that without using counters
> how can a DBA Can say that OS Can be ruled out and that the Application is
> possibly the bottleneck. That I assumed is my limited knowledge of
> performance and if there are some kind of telltale signs that a DBA Can
> point out without even using tools. If thats not possible then i guess the
> tools are the only one i can rely on.
> Thanks Uri for your response.
> Meher
> "Uri Dimant" <urid@.iscar.co.il> wrote in message
> news:uXhj3U$JGHA.3912@.TK2MSFTNGP10.phx.gbl...
>
|||I use performance counters to supplement my SQL performance monitoring, not
to eliminate it. Most of the time, performance problems are due excessive
resource consumption. The counters tell me which resource is being
consumed. I then look at SQL to see which processes are consuming those
resources, often by using SQL traces. Be aware that SQL will trade one
resource for another for speed. Primarily it trades higher memory and CPU
usage for less disk usage. Sometimes a problem in one area can appear to be
something else. For example, I had a system with very high disk queue
lengths on the data disk. That would seem to indicate an I/O issue.
However, it also had an extremely short page life expectency, indicating
lack of cache memory. I upped the RAM and the disk problem went away. The
I/O problem was caused by SQL having to constantly reload data that should
have stayed in cache. I still need to tune some processes that use large
intermediate result sets, but that isn't an emergency priority.
This example illustrates that it is not just when X counter exceeds Y value
that tells you there is a problem. You have to know what is normal for your
system and work from there. Counters only give you a partial picture.
Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP
"Meher" <NOSPAM_mmsagar@.hotmail.com> wrote in message
news:uIdeYK$JGHA.532@.TK2MSFTNGP15.phx.gbl...
> Hi:
> I have a question hopefully the SQL Server Pros and MVPs would have an
> answer for. Most of the time like any normal DBA to find performance
> issues with SQL Server I make use of the performance counters through
> performance monitor or system monitor and SQL Server Profiler and then
> rule out OS as the bottleneck and then concentrate on the application
> part. My question here is if there a simple way that without using any of
> these counters or tools a DBA can rule out OS as the potential bottleneck.
> For example I would assume that if my backups start taking a long time all
> of a sudden when they usually take less time every day then I can probably
> point out OS as the issue. I am not sure if thats the right way to tell.
> I am looking for a comprehensive list of possible symptoms (like paging,
> Virtual memory, Disk I/O) that can conclusively deduce that OS is/is not
> the bottle neck without me making use of any of the above mentioned tools.
> Any KB Article or a White paper or blogs with more information would be
> really helpful. Any other information is greatly appreciated.
> Thanks for your attention.
> M
>
|||Thanks Geoff. Your post is really informative. The reason why I am asking
this question is i was asked iimagining i dont have any tools to troubleshoot
SQL Server Performance issues, before driving to the conclusion that
application related issues are the cause of performance, how do I eliminate
OS as the primary cause. I am trying to get a grasp on the approach of senior
SQL DBAs in such a scenario.
Thanks again.
"Geoff N. Hiten" wrote:

> I use performance counters to supplement my SQL performance monitoring, not
> to eliminate it. Most of the time, performance problems are due excessive
> resource consumption. The counters tell me which resource is being
> consumed. I then look at SQL to see which processes are consuming those
> resources, often by using SQL traces. Be aware that SQL will trade one
> resource for another for speed. Primarily it trades higher memory and CPU
> usage for less disk usage. Sometimes a problem in one area can appear to be
> something else. For example, I had a system with very high disk queue
> lengths on the data disk. That would seem to indicate an I/O issue.
> However, it also had an extremely short page life expectency, indicating
> lack of cache memory. I upped the RAM and the disk problem went away. The
> I/O problem was caused by SQL having to constantly reload data that should
> have stayed in cache. I still need to tune some processes that use large
> intermediate result sets, but that isn't an emergency priority.
> This example illustrates that it is not just when X counter exceeds Y value
> that tells you there is a problem. You have to know what is normal for your
> system and work from there. Counters only give you a partial picture.
> --
> Geoff N. Hiten
> Senior Database Administrator
> Microsoft SQL Server MVP
>
>
> "Meher" <NOSPAM_mmsagar@.hotmail.com> wrote in message
> news:uIdeYK$JGHA.532@.TK2MSFTNGP15.phx.gbl...
>
>
|||Meher
A bottle neck can be due to either of LAN/WAN, SAN/disk, CPU or memory.
You will need to use an approach to see if a similar task can be
completed at the OS level without using SQL specific connetivity.
An example would be if you have network performance issues you will
test a command prompt file copy (say 100MB) and capture packets with
Ethereal to see how that performs compared to data transfer using SQL
commands.
regards
Pradeep Arora
Meher wrote:[vbcol=seagreen]
> Thanks Geoff. Your post is really informative. The reason why I am asking
> this question is i was asked iimagining i dont have any tools to troubleshoot
> SQL Server Performance issues, before driving to the conclusion that
> application related issues are the cause of performance, how do I eliminate
> OS as the primary cause. I am trying to get a grasp on the approach of senior
> SQL DBAs in such a scenario.
> Thanks again.
> "Geoff N. Hiten" wrote: